Technical ReportPDF Available

Better ways to do research: An overview of methods and technologies that can replace animals in biomedical research and testing.



An overview of human-relevant non-animal methods in biomedical research, testing, education and training - in plain English. Further, examples are provided of how these methods have been applied and been of benefit to humans.
Better ways
to do research:
An overview of methods
and technologies that
can replace animals in
biomedical research
and testing.
2Humane Research Australia
This is an open access publication distributed under the terms of the CC – BY – NC – SA license,
which permits any non-commercial use, distribution and reproduction, provided the original
author and source are credited. If you remix, transform, or build upon the material, you must
distribute your contributions under the same license as the original.
ISBN 978-0-6485324-0-8
Humane Research Australia is a not-for-profit organisation that challenges the use of animal
experiments and promotes the use of more humane and scientifically valid non-animal methods
of research.
1800 HUMANE (1800 486 263)
Design by U-bahn design Pty Ltd
3Better ways to do research
Better ways
to do research:
An overview of methods and technologies
that can replace animals in biomedical
research and testing
Monika Merkes PhD (public health)
President, Humane Research Australia
Honorary Associate, La Trobe University, Australia
May 2019
4Humane Research Australia
Why I wrote this overview
“But isn’t animal research a necessary evil?” This is a common response when I mention that I am the
president of Humane Research Australia. For several years now, I have wished for a publication I could
recommend to people who are genuinely interested in alternatives1 to animal research2, but I have found
scientific research articles hard to locate and diicult to read. Summaries of alternative methods are
available, but they are usually limited to a few pages. This overview in plain English is intended to fill
this gap.
My sources are mostly scientific publications in peer-reviewed journals and government reports, and I
provide references so that the interested reader may find out more from the original article or report3.
I have summarised what I found in the scientific literature and use many quotations because I think
researchers who have done the work can oen explain it best.
The focus here is on animal research conducted mainly for human medicine or other human “benefit”,
which is oen questionable. Animals are also used for studies in veterinary medicine.
This overview of alternatives to animal research and testing, as well as the use of live animals in education
and training, was written in late 2018 and early 2019. In this rapidly growing area of research, existing
methods and technologies are being constantly improved and new ones added. It is also a very complex
area that is diicult to access for a layperson. I have done my best to give a sense of the available
information at the present and hope that readers may find it informative and useful.
The term “alternative” is used as relating
to methods that depart from or challenge
traditional norms. It is not used as oering
another, equally valid possibility or choice.
I use the terms animal research and animal
experimentation interchangeably with animal
referring to non-human animals.
While some articles are behind paywalls,
authors are usually willing to provide a copy
of their publication when contacted via email.
A contact email address is usually included in
the article abstract, which is available for free.
5Better ways to do research
I would like to thank the following friends and colleagues for critical and helpful comments on dras of
this document: Scott Anderson, Rob Buttrose, Dr Eleonora Gullone, Dr John van Holsteyn, Helen Marston,
Reema Rattan and Cheryl Veitch. In the spirit of peer review, three experts on novel non-animal methods
in research and testing have generously provided feedback on the manuscript: I am deeply grateful to Dr
Andrew Worth, European Commission, European Union Reference Laboratory for Alternatives to Animal
Testing; Dr Malcolm Wilkinson, Kirkstall Ltd; and Vy Tran, Center for Alternatives to Animal Testing, Johns
Hopkins University, for their thoughtful suggestions. For design and layout I thank Adam Foran and his
team at U-bahn. Finally, for company during the long hours I spent researching and writing this overview,
I thank Sheba who patiently lay under my desk, perhaps dreaming of chasing balls in the sunshine.
6Humane Research Australia
Foreword 4
Why I wrote this overview 4
Acknowledgements 5
Contents 6
Summary 7
Introduction 13
What types of research use animals? 13
How many animals are used? 16
What animal welfare measures are meant to
protect animals? 17
Why do we need alternatives to animal
research? 18
In-vitro methods 20
3D tissues and microfluidic devices 20
Organoids 20
Organs-on-chips 22
Biobanking 27
‘Omics’ technologies 28
Stem cell technologies 29
3D and 4D bioprinting 32
Robotic testing 35
In-silico methods 37
In-silico prediction methods for the
evaluation of toxicity 39
SARs and QSARs 39
Read-across 40
PBPK models 41
Expert systems 41
In-silico tools for the evaluation of toxicity 42
OECD QSAR Toolbox 42
REACHacross 42
Toxtree 43
Toxmatch 43
Other in-silico approaches for the
evaluation of toxicity 44
AOPs 44
Computer modelling of health and disease 45
Studies with human
volunteers 48
Post-mortem studies 48
Population-based studies 49
Microdosing 50
Simulators 51
Efforts by governments and the
scientific community to replace
animal experimentation 55
Policies and collaborations 55
United States 55
European Union 56
International 57
Validation 58
Why it matters 59
Glossary 62
References 65
7Better ways to do research
Better ways to do research: An overview of methods
and technologies that can replace animals in
biomedical research and testing.
Each year, millions of non-human animals4 worldwide are harmed by animal experimentation. It has been
estimated that more than 115 million animals are used per year to supply the biomedical industry5. The
countries that use the most animals include China, the US, Japan and Australia. Broadly, the types of
research that use animals consist of a) fundamental research (also called basic research), b) applied (or
human disease) research, and c) testing (or regulatory testing).
There are alternatives to using animals. New – and not so new – methods and technologies that can
replace live animals in research, testing, education and training include:
1. In-vitro methods (performed with microorganisms, tissues,
whole cells or parts of cells in test tubes, Petri dishes etc.)
2. In-silico (computer-based) methods
3. Studies with human volunteers
4. Simulators
1. In-vitro (test tube) methods
3D tissues and microfluidic devices: Organoids,
Organoids are a miniature and simplified version of a (human) organ. Organoids are grown in-vitro in
three dimensions. They allow researchers to study disease and treatments in the laboratory. Mini organs
can also be grown on microchips. Researchers have used microchip manufacturing methods to engineer
microfluidic6 culture devices that can mimic the structures and functions of living human organs.
Organoids, organs-on-chips – Useful for/ can replace animals in:
drug testing
for toxicity &
& other
Large scale &
fast studies/
Education &
access to
 
hereinaer referred to as animals Akhtar, 2015 At the micro scale, oen flowing through
8Humane Research Australia
To study human cells and tissues, researchers need a readily available supply of these human biological
samples. They are stored in so-called biobanks or tissue banks. Biobanks use tissue that is le over from
clinical procedures such as surgery, from dead bodies, or they collect tissue specifically for research. They
also store organoids.
Biobanking – Useful for/ can replace animals in:
drug testing
for toxicity &
& other
Large scale &
fast studies/
Education &
access to
Omics technologies
The term “omics technologies” refers to areas of study in biology whose names end in “omics”, such as
genomics (the study of the genome of an organism). The science of “omics” reflects diverse technologies
with a focus on studies of life processes, such as comprehensive studies of genes, proteins and metabolites
of an organism.
Omics technologies – Useful for/ can replace animals in:
drug testing
for toxicity &
medicine & other
Large scale &
fast studies/
Education &
access to
human tissues
 
Stem cell technologies
Stem cells are unspecialised or undierentiated cells with the ability to self-renew, and to dierentiate to
produce specialised cell types in the body.7
Stem cells – Useful for/ can replace animals in:
drug testing
for toxicity &
medicine & other
Large scale &
fast studies/
Education &
access to
human tissues
 
Stem Cells Australia, 2018
9Better ways to do research
Bishop, et al., 2017 Li, Zhang, Akpek, Shin, & Khademhosseini,
3D and 4D bioprinting
Bioprinting involves the precise layering of cells, biologic scaolds, and growth factors with the goal of
creating bioidentical tissue for a variety of uses.8 4D bioprinting aims to create dynamic 3D patterned
biological structures that can transform their shapes or behaviour under various stimuli. For example,
4D bioprinted materials are capable of changing their shape over time.9
3D and 4D bioprinting – Useful for/ can replace animals in:
drug testing
for toxicity &
medicine & other
Large scale &
fast studies/
Education &
access to
human tissues
 
Robotic testing
Researchers, in particular those in the pharmaceutical industry, have developed automated methods
to test biological activities of thousands of chemicals that used to be tested in animals. This is called
high-throughput testing or robotic testing.
Robotic testing – Useful for/ can replace animals in:
drug testing
for toxicity &
medicine & other
Large scale &
fast studies/
Education &
access to
human tissues
 
10 Humane Research Australia
2. In-silico (computer-based) methods
Prediction methods and tools
A range of in-silico prediction methods and tools for the evaluation of toxicity have been developed,
such as structure-activity relationships (SARs) and quantitative structure-activity relationships (QSARs),
read-across, physiologically-based pharmacokinetic (PBPK) models, expert systems, read-across structure
activity relationships (RASAR), OECD QSAR Toolbox, REACHacross, Toxtree, and Toxmatch.
Prediction methods and tools – Useful for/ can replace animals in:
drug testing
for toxicity &
medicine & other
Large scale &
fast studies/
Education &
access to
human tissues
 
Other in-silico approaches
Adverse Outcome Pathways (AOPs) provide the biological explanation for a single toxic event.10 Integrated
Approaches for Testing and Assessment (IATA) are approaches for making decisions about the toxicity of
substances that are based on multiple information sources.
AOPs and IATA – Useful for/ can replace animals in:
drug testing
for toxicity &
medicine & other
Large scale &
fast studies/
Education &
access to
human tissues
 
Computer modelling
A computer-based model or simulation is a computer program that is designed to simulate a physical or
biological system or situation. Computer models can link many processes together, something which is not
possible to achieve with animal models.
Computer modelling – Useful for/ can replace animals in:
drug testing
for toxicity &
medicine & other
Large scale &
fast studies/
Education &
access to
human tissues
 
 Taylor, 2019
11Better ways to do research
3. Studies with human volunteers
Post-mortem studies
Donated tissues aer the death of a person can be studied to gain insight into cell-level changes in
human illnesses, and cadavers can be used in training surgical skills. For example, post-mortem brain
studies are useful to gain more knowledge about psychiatric illnesses, in particular in combination with
the approaches of genomics and proteomics (the study of the structural and functional aspects of total
proteins of an organism or system using high-throughput technologies).11
Post-mortem studies – Useful for/ can replace animals in:
drug testing
for toxicity &
medicine & other
Large scale &
fast studies/
Education &
access to
human tissues
 
Population-based studies
Epidemiology is the study of diseases and other health-related states in groups (populations) of people, in
particular how, when and where they occur. Epidemiologists want to discover what factors are associated
with diseases (risk factors), and what factors may protect people against disease (protective factors).
Population-based studies – Useful for/ can replace animals in:
drug testing
for toxicity &
medicine & other
Large scale &
fast studies/
Education &
access to
human tissues
 
Microdosing involves the administration of very low doses of a substance (sub-therapeutic).
When testing a new compound or drug, microdosing can provide useful information to help decide
whether the new compound or drug should be developed further, and whether it may be safe to
progress to further human testing.
Microdosing – Useful for/ can replace animals in:
drug testing
for toxicity &
medicine & other
Large scale &
fast studies/
Education &
access to
human tissues
 Yadav, Tanveer, Malviya, & Yadav, 2017
12 Humane Research Australia
Collaborative efforts
to replace animal
Governments, the scientific community,
industry and other stakeholders,
in particular in the EU and the US,
have started to make eorts to pool
knowledge and resources to replace
animal experimentation with more
humane, more human-relevant, and
oen cheaper and faster methods. This
includes, for example, the development
of policies and tools, working together
to build large databases, developing
plans for the further development
of particular technologies (such as
organs-on-chips), and collaborating
on the validation of new methods
Why it matters
For a long time, the use of animals
in research, testing, training and
education has been considered a
necessary evil. More and more, people
question the ethics of this approach.
At the same time, the animal research
community increasingly recognises the
problems with animal research: it is
costly, lengthy and not very eective.
Also, it may have held back the
discovery of treatments and cures for
humans because they did not work well
in animals.
The main alternatives to the use of
animals in the laboratory are new in-
silico and in-vitro approaches. Studies
with human volunteers and simulators
also play an important role. Some of
these methods are used in combination
for greater eectiveness. So far,
most progress in the development of
alternatives has been made in the area
of toxicology. The new methods and
technologies are not yet perfect, and
some of the current methods that are
deemed to be alternatives might still
use animal parts.
Animal researchers argue that the
new methods can’t replace all areas of
current animal research. Considering
the ethics of using animals for the
purported benefit of humans and
the many shortcomings of animal
research, this is a compelling argument
for speeding up the development of
human-relevant research and testing
without animals.
We need urgent change. From an
animal rights perspective, it was never
okay to inflict pain and suering on
animals for the real or perceived benefit
of humans. For proponents of animal
welfare, the use of animals is justified
as long as harm is minimised. With
awareness of the many shortcomings
of animal research and testing and
increasing availability of better ways,
animal research is no longer justified.
With greater investment in innovative
and promising non-animal methods,
firm policy initiatives and robust
collaborations of all interested parties,
better treatments and cures for human
diseases can be developed. This will
also end the suering of millions
4. Simulators
Simulators are either virtual reality (VR)-based or physical model (PM)-based. Apart from replacing live
animals in education and training, VR simulators have great potential for training people in remote
locations, for example, training students and surgeons in developing countries.
Simulators – Useful for/ can replace animals in:
drug testing
for toxicity &
medicine & other
Large scale &
fast studies/
Education &
access to
human tissues
 
13Better ways to do research
This is an overview of methods and technologies that can
replace live animals in research, testing, education and
training. It is divided into four sections:
In-vitro methods (performed with microorganisms, tissues, whole cells or parts of cells in test tubes and
Petri dishes)
In-silico (computer-based) methods
Studies with human volunteers
This is followed by a section on eorts by governments and the scientific community to replace animal
experimentation, as well as concluding reflections on why the availability and further development of
alternatives matter. A glossary is also included.
But first of all, I provide a few general observations and facts about animal research.
What types of research use animals?
The types of research that use animals
are usually called fundamental or basic
research, applied or human-disease
research and testing (regulatory
testing). Basic research is curiosity
driven and, unlike applied research, it
is not necessarily designed to answer
specific questions or solve practical
problems. It is exploratory and aims
to increase and advance scientific
knowledge. Applied research aims
to solve specific practical problems,
such as using animals as a model
to seek a cure for a human disease
or condition. Animals are also used
in education and training – from
preschool to postgraduate level and
in professional development. There
is no legal requirement for animal
experimentation in basic and applied
research, nor for it to be part of
education and training. Individual
scientists decide what is worth
studying and whether or not they will
The situation is dierent for the third
type of research, regulatory testing.
Government regulators require that
new consumer products, medicines,
and industrial and agricultural
chemicals are tested to identify
potential dangers to human and animal
health, as well as to the environment.
For some product types (drugs and
vaccines, biologicals12), animal testing
includes testing for eicacy as well as
safety (toxicity). Current laws in many
countries make it diicult to avoid
using animals for regulatory testing,
although with the development of new
methods and technologies this has
started to change.
Available statistics dier in how they
describe and count procedures that
use animals. They don’t usually explain
the purposes of animal research by
these three categories. The descriptions
provided by dierent countries also
vary, as does the use of animals by
category. But generally, basic research
accounts for roughly half of all animals
used in research. The three figures
(below) are examples of statistical
reporting on the use of animals for
scientific purposes. These are the most
recent statistics available from the
European Union (EU), the United States
(US) and Australia.
 Biologicals, or biologics, are drugs that
are made from living organisms or contain
components of living organisms.
14 Humane Research Australia
Education & training
Diagnosis of disease
Biological studies
of a fundamental
Toxicological & other
safety evaluation
Research &
development - human,
veterinary, dentist
Production & quality control
for products for human
medicine and dentistry
Production and quality
control of products for
veterinary medicine
Quality control of products
and devices, production
Research &
Toxicity testing
 European Commission, 2013, p. 6  Meigs, Smirnova, Rovida, Leist, & Hartung,
Purposes of experiments, European Union 2011
Source: European Commission13
Purposes of experiments, US 2016
Source: Meigs et al.14
15Better ways to do research
Stock Breeding
Stock Maintenance
Regulatory product testing
human or animal
Achievement of
educational objective
Environmental study
Improvement of animal management or production
Maintenance and improvement of human
or animal health and welfare
Diagnostic procedures
Production of biological products
Purposes of experiments, Australia 2016
Source: Humane Research Australia15
Mice and rats are the most commonly
used animals in research. In 2011,
rodents and rabbits represented 80%
of all animals used in research in the
EU. The second most-used group were
cold-blooded animals, such as reptiles,
amphibians and fish (12.4%) while
birds accounted for 5.9%.16 However,
in recent years the use of zebrafish has
soared, partly due to their lower cost
compared to mammals. In the UK, they
are now the second most-used animals
aer mice.17
 Humane Research Australia, 2016  European Commission, 2013  Home Oice, 2015
16 Humane Research Australia
How many animals are used?
We do not know how many animals
are used worldwide for research,
testing and teaching purposes. Not
all countries release statistics about
animal use. In Australia, for instance,
only four states provide statistics while
in the US, the numbers of mice, rats,
fish and birds used are not known
because these animals are not counted.
From published data and estimates we
can ascertain that:
In Australia, around 9 million animals
were used in 201618
In the EU’s 28 member states, just
under 11.5 million animals were used
in 201119
In the US, an estimated 26 million
animals were used in 2010, with
96% of these being mice, rats, fish
and birds20
In Canada, in 2016, 4.3 million
animals were used in research,
teaching, and testing. The majority
of animals (57.3%) were used in
basic research21
In South Korea, 3 million animals
were used in 201722
An estimate of worldwide use
suggested a figure of 115.3 million
animals used for research in 200523
From what statistics are available, it
appears that, overall, the number of
animals used in research has been
stable in recent years24 although in
some countries, it has increased. In
particular, the number of genetically-
altered (transgenic) mice has risen,
especially in basic research.25
 Humane Research Australia, 2016
 European Commission, 2013
 The Hastings Center, no date
 Meigs, et al., 2018
 Meigs, et al., 2018
 Taylor, Gordon, Langley, & Higgins, 2008
 Meigs, et al., 2018
 Daneshian, Busquet, Hartung, & Leist, 2015;
Timoshanko, Marston, & Lidbury, 2017
9 million
Animals used in
Australia (2016)
Animals used
worldwide (2005)
115.3 million
An estimated The number of animals
used in the laboratory
has been constant
over recent years
Animals used in the
28 EU countries (2011)
11.5 million
Just under
17Better ways to do research
 Leonard, Thornton, & Vink, 2014
 Franco, 2013
 Franco, 2013, p. 2
 Merkes & Buttrose, 2019
What animal welfare measures are meant
to protect animals?
In animal experimentation, researchers
are permitted to use procedures
that would be illegal outside of
the laboratory, such as artificially
producing spinal cord injury in
rabbits.26 However, there are laws,
regulations and codes of practice
that direct researchers to limit harm
to the animals they use. The most
widely accepted guidance on limiting
harms to animals in biomedical
research comprises the principle of
the 3Rs: Replacement, Reduction and
Refinement. The 3Rs were proposed by
William Russell and Rex Burch in the
late 1950s and are today embedded
in many laws, regulations and
codes governing animal use.27 It was
summarised in a 2015 article as:
“In essence, they allow animals
to be used in scientific research
only when they cannot be replaced
with non-animal alternatives,
when the number of animals has
been reduced as much as possible
given the research goals, and
when procedures and housing
have been refined to minimize
welfare impacts”.28
Animal experimentation regulations
vary around the world. Many countries
require an institutional or project
license before research using animals
can be carried out. There is also usually
a requirement for a group made up
of representatives of veterinarians,
scientists, animal welfare organisations
and the general public to oversee the
ethical conduct of individual projects.
In Australia, these are called Animal
Ethics Committees (AEC) and in the EU,
all states have a national committee
for the protection of animals used for
scientific purposes. In Australia, the
members of AECs are usually appointed
by the institution that carries out the
animal experiments. The public does
not have easy access to their minutes
and reports. It is generally not possible
for a member of the public then, to
assess whether the use of animals
in a particular project could have
been avoided because alternatives
were available. In other words, the
transparency of the process that
determines decisions on the use of
animals is questionable.29
The 3Rs –
– are meant to
protect animals
in research testing.
The public does
not have enough
information to
judge whether
or how the 3Rs
protect animals.
18 Humane Research Australia
 Singer, 1989, p. 1
 Strauss, 2018
 In vivo means “in the living”. In vivo methods
are methods using a living organism/animal.
 e. g., Cummings, Morstorf, & Zhong, 2014;
Marshall, Austin, Casey, Fitzpatrick, & Willett,
2018; Pound & Ritskes-Hoitinga, 2018; Seok,
et al., 2013; Thomas, et al., 2016; Triunfol,
Rehen, Simian, & Seidle, 2018; van der Worp,
et al., 2010
 Cummings, et al., 2014
 Hutchinson & Kirk, 2011
 Savoji, et al., 2018, p. 1
Why do we need alternatives to animal research?
The use of animals in research – and the
moral status of animals more generally
– has been and remains surrounded
by ethical controversy. Answers to the
question “do we have the moral right
to use animals for our purposes?” have
been approached from many angles.
Broadly, there are two positions: animal
welfare and animal rights. Animal
welfare is concerned with minimising
suering, while the animal rights
position considers the use of animals as
our resources to be morally wrong.
Animal liberationists, for instance, call
for equal consideration of interests.
They question the right of humans to
assume that our interests must always
prevail. For example, Peter Singer noted
30 years ago:
“And the basic right that animals
should have is the right to equal
consideration. This sounds like
a difficult idea, but essentially it
means that if an animal feels pain,
the pain matters as much as it does
when a human feels pain - if that
pains hurt just as much. Pain is
pain, whatever the species of being
that experiences it”.30
On the other hand, supporters of
the animal welfare position find it
morally acceptable to use animals in
research as long as their wellbeing
is considered. The 3Rs represent an
animal welfare position, particularly
in the requirements for refinement
Recent surveys indicate a growing
concern with the suering of animals
used in research, and the public is
increasingly questioning whether
animals ought to be considered as
means to an end or as sentient beings
with inherent value. In 2018, for
instance, the Pew Research Center
reported that 52% of Americans
are opposed to using animals in
Ethics aside, there are other reasons
why the use of animals in research
and testing for human purposes
should be replaced with better ways to
In-vivo32 and traditional in-vitro
methods (also called test tube
experiments) are not good at predicting
therapeutic outcomes and possible
side eects during clinical trials with
humans. As many as 95% of drugs
that appear safe and eective in
animals fail in humans.33 Drugs for
Alzheimer’s disease, for instance, have
a 99.6% failure rate34 while those for
cancer35 and heart disease also have a
particularly high failure rate. A group
of Canadian researchers made the
following observation about the safety
and eectiveness of specific drugs:
“Cardiovascular toxicity claims the
highest incidence and severity of
adverse drug reactions in late-stage
clinical development. For example,
Vioxx (Rofecoxib), originally
designed to treat pain related to
osteoarthritis and approved by the
Food and Drug Administration (FDA)
in 1999, was linked to over 27,000
cardiovascular-related deaths
and myocardial infarctions (MI).
It was withdrawn from the market
in 2004, although later relicensed
for more specific indications, with
implementation of regulatory
and transparency safeguards. In
preliminary clinical investigations,
the drug showed effectiveness in
its target treatment and adverse
events were not significant. It was
not until four years of long-term
clinical studies that it became
evident that the risk of heart attack
and stroke was actually two-fold
higher with Vioxx compared to
the control group. Some other
compounds, such as Micturin
(Terodiline, for urinary incontince),
Fen-phen (Fenfluramine/
phentermine, anti-obesity
treatment), Seldane (Terfenadine,
allergy medication), Zelnorm
(Tegaserod, for irritable bowel
syndrome), Meridia (Sibutramine,
appetite suppressant), and
Darvon/ Darvocet (Propoxyphene,
analgesic drug), have all had a
similar record in terms of adverse
cardiovascular effects”.36
19Better ways to do research
How can these adverse drug reactions
be explained, given that the drugs
had to undergo safety (toxicity)
testing? Thomas Hartung, Director of
the Center for Alternatives to Animal
Testing (CAAT) at Johns Hopkins
University in the US oered the
“Toxicology and effective risk
assessment depend on scientific
and technological information and
should constantly adapt to these
advances in this information.
However, toxicology still largely
relies on traditional assessment
methods that were established
decades ago and that have
changed little despite scientific
and technological progress.
Consequently, safety assessments
are often based on tests of unknown
relevance and reliability and whose
predictive validity has never been
assessed objectively“.37
The extensive use of animals in basic
research raises another issue. Basic
research does not aim to result in
practical outcomes and it oen doesn’t.
For example, a group of researchers
examined articles in six highly cited
basic science journals over a five-year
period. They found that fewer than
10% of highly promising basic science
discoveries enter routine clinical use
within 20 years.38 From a utilitarian
perspective, which is oen used by
animal researchers, it can be argued
that the many animal lives lost in basic
research do not justify the benefits.
Opportunity cost, the loss of other
options when one option is chosen,
presents an additional problem. Given
that animal research and testing has
a high failure rate, we don’t know to
what extent new drugs that would
be beneficial to humans have been
overlooked, because they were harmful
to animals.
In addition to producing misleading
results, animal tests are also costly and
lengthy.39 The cost of drug development
has been increasing and the number
of new drugs approved every year
has been decreasing over the last two
decades.40 The drug development
process takes around 10 years and has
been estimated to cost up to US $2.5
billion.41 Economic considerations are
in favour of more human-relevant,
cheaper and faster methods.
“Unlike crude, archaic animal
tests, non-animal methods usually
take less time to complete, cost
only a fraction of what the animal
experiments that they replace cost,
and are not plagued with species
differences that make extrapolation
difficult or impossible”.42
“For decades, laboratory biologists
have regarded animal models
as a necessary evil. While some
activists decry their use on moral
grounds, even the most practical-
minded researchers acknowledge
fundamental problems with them.
Animals are expensive, provide
only imperfect replicas of human
biology, and introduce numerous
variables into experiments that can
be difficult or impossible to control.
These flaws aren’t purely academic.
Pharmaceutical researchers have
struggled for years with late-stage
development failures, in which
drugs that look promising in
multiple animal systems turn out to
be useless or even toxic in humans.
Nonhuman models have simply
been the least bad tool for detailed
studies on human biology”.43
 Hartung, 2009a, p. 93
 Contopoulos-Ioannidis, Ntzani, & Ioannidis,
 Meigs, et al., 2018
 Zhang, Korolj, Lai, & Radisic, 2018
 Ahadian, et al., 2018
 Ranganatha & J, 2012, p. 32
 Dove, 2018
Animals are not
good models for
clinical outcomes
because of
differences in
metabolism and
between human
and non-human
20 Humane Research Australia
In-vitro methods
3D tissues and microfluidic devices
For almost a century, scientists have
grown cells, such as bacteria or human
tissue, in Petri dishes. A Petri dish is
a shallow glass or plastic dish with
a lid that is used by scientists for
microbiological studies. It contains
a culture that feeds the cells so they
can grow. The cells rest at the bottom
of the dish and spread out as they
multiply in a two-dimensional (2D)
direction. Since this is not how organs
grow in a living body, scientists have
developed ways in which cells can be
grown in a three-dimensional (3D) way.
By studying these 3D tissues, such as
mini organs (also called organoids) and
organs-on-chips grown from human
cells, the research directly translates to
human health and saves animals from
being experimented on. It is also oen
cheaper and faster.
Lab-grown mini organs were first
developed in 2013.44 They are miniature
versions of human organs and can be
grown from many dierent organs.
Healthy or tumour cells are taken via
a small biopsy from a person’s organ,
bathed in a culture that stimulates
them to grow over a few weeks or
months and to organise themselves
into mini versions of human organs.
For example, the process of growing
a gut organoid has been described in
three steps in the following way:
1. “Take a tissue sample. A very small
biopsy is taken from the epithelium,
the tissue lining the gut.
2. Incubate. The tissue is bathed in a
mix of growth factors designed to
let gut stem cells replicate.
3. Organoids, a millimeter or less in
diameter, emerge in up to 3 weeks
and can be frozen for later use.45
Organoids have many applications.
They can be used for:
Regenerative medicine – organoids
grown from healthy tissue could be
placed back into a patient to help
repair damaged tissue.
Toxicity testing – toxicologists can
use organoids to test the eects of
chemicals on the liver and other
human organs.
Drug testing – drugs can be tested
on organoids to help predict their
eects in patients.
Microbiome studies – scientists can
study how normal human intestinal
bacteria interact with gut organoids.
Modelling infections – organoids
can be infected with viruses
or bacteria to study how these
In-vitro methods (the Latin phrase means “in the glass”) are
also commonly known as test-tube methods although this kind
of research is traditionally done in flasks or Petri dishes as
well as test tubes. In-vitro tests and experiments are largely
performed outside living organisms and involve tissues,
microorganisms, cells or other small parts of biological
This section provides an overview of recently developed in-
vitro methods that can replace experiments or tests with live
animals. They include 3D tissues and microfluidic devices,
such as organoids (mini versions of organs) and organs-on-
chips, biobanking, omics technologies, stem-cell technologies,
3D and 4D bioprinting, and robotic testing.
 Grens, 2018  Sinha, 2017, p. 4
21Better ways to do research
Personalised medicine – organoids
grown from individual patients can
help predict their response to new or
existing drugs.
Cancer studies – scientists can
study how cancer develops by
introducing mutations in organoids
grown from healthy tissues.46
These mini organs are not capable of
reproducing all biological responses
like a real human organ, but they
allow researchers to study a variety
of physiological responses to specific
manipulations and treatments. They
allow for “surrogate” trials before
conducting clinical trials, “providing
a vital, physiologically relevant bridge
between pre-clinical investigations
and clinical outcomes and bringing
the possibility of personalised
Brain organoid
Source: National Institutes of Health, US48
Inner ear organoid
Source: National Institutes of Health, US49
 Sinha, 2017
 Archibald, Tsaioun, Kenna, & Pound, 2018,
p. 2
 NIH Image Gallery, 2019
 NIH Image Gallery, 2018
22 Humane Research Australia
Mini organs can also be grown on
microchips. Researchers have used
microchip manufacturing methods to
engineer microfluidic50 culture devices
that can mimic the structures and
functions of living human organs.
These organs-on-chips are made of a
clear and flexible polymer (molecules
of a simple compound joined together)
and contain hollow microfluidic
channels lined with living human
cells. Microfluidic channels contain
tiny amounts of liquid ranging from
submicron (smaller than one millionth
of a metre) to a few millimetres.
They are equipped with mechanical
forces that can mimic the physical
environment of organs, such as
breathing motions (lung-on-a-chip)
and peristalsis-like movements
(intestine-on-a-chip). When nutrients,
air, blood or drugs are added, the cells
replicate some of the key functions of
the organ. These organs-on-chips, in
which cells can grow and fluids can
flow, are usually the size of a computer
This lung-on-a-chip serves as an accurate model of human lungs to test for
drug safety and eicacy.
Source: National Center for Advancing Translational Sciences, US51
Multiple tissue chips can be connected in a system to simulate a human-
Source: National Center for Advancing Translational Sciences, US 52
 At the micro scale, oen flowing through
 National Center for Advancing Translational
Sciences, no year-d
 National Center for Advancing Translational
Sciences, 2018d
23Better ways to do research
The term organ-on-a chip was thought
up by Donald Ingber, the Founding
Director of the Wyss Institute for
Biologically Inspired Engineering at
Harvard University.53 Together with his
multidisciplinary team he developed
in 2010 a lung-on-a-chip.54 Professor
Ingber is oen credited with having
developed the first organ-on-a-chip,
but similar work had been undertaken
some years earlier by a team from
Seoul National University and
Organs-on-chips use dierent types of
human cells: primary cells (cells directly
taken from an organ or a tissue),
immortalised cell lines (cells that have
been modified by chemicals or a virus
in order to survive and stay active
indefinitely), or stem cells.56
Organ-on-a-chip platforms that have
already been developed include liver,
skin, vascular structures (such as
arteries, veins, capillaries), cardiac
muscle, skeletal muscle, lung, bone and
bone marrow, brain, eye, gut, spleen
and kidney. There are also multi-organ-
on-a-chip platforms that mimic the
interplay between dierent organs.57
A recent workshop attended by experts
from industry, academia and the
Medicines and Healthcare products
Regulatory Agency (MHRA) held in
Liverpool UK identified the advantages
of organ-on-chip technologies:
“These dynamic and responsive
biological test platforms have the
potential to revolutionise drug
target identification and validation
studies without the need for animal
models. This will improve compound
efficacy, safety and targeted
drug delivery”.58
The experts also identified the
technical, funding and regulatory
challenges still to be overcome. Some
of the technical challenges include
physically-relevant cell interactions,
scaling ratios between organs,
and incorporation of immune or
endocrine systems. The capabilities
and interactions in multiple organ-on-
a-chip systems are only in the early
stages of development. However,
assuming future collaboration between
organ-on-chip innovators, users,
regulators and funders, the participants
of this workshop anticipated real
patient benefits by replacing poorly
predictive animal models with these
more physiologically relevant human-
Organ-on-chip technology is
still mostly used in-house by the
companies and laboratories that
developed it. However, researchers in
academic institutes and biochemical,
pharmaceutical, cosmetics and
chemical companies have started to use
the technology, and it can also be used
by hospitals for personalised medicine.
Researchers have also developed
disease models, such as chips with
tumour cells to study cancer.
“Chip-based in vitro organ models
are ranked 6th among the top ten
emerging technologies by the World
Economic Forum in 2016, thus
highlighting the potential of organ-
on-a-chips to improve lives and
transform the health care system.
In other words, organ-on-a-chip
technology is expected to speed up
pharmaceutical drug development
efforts, improve translation
of basic research to clinically
relevant patient scenarios and
provide personalized intervention
Organs-on-chips can replace many
dierent species of animals currently
used for research, such as mice, rats,
rabbits, guinea pigs and non-human
primates, which are all now used for
drug testing and vaccine development.
This technology has great potential
for pharmaceutical research and
personalised medicine. Apart from
saving the lives of millions of animals,
it is also cheaper and faster than
“Organs-on-chips are bio-
engineered devices that mimic
key aspects of the physiology
and function of human organs,
replicating some of the complexity
of the human body environment on
a microscopic scale. They are able to
mimic blood, air and nutrient flow,
as well as mechanical forces such as
peristalsis and can be continuously
monitored to obtain a profile over
time. Organs-on-chips enable the
study of basic biological processes,
the modelling of diseases and
investigation of the effects of drugs.
They can potentially identify safety
and efficacy issues earlier and more
reliably in the drug development
process, enabling the design and
selection of drug candidates that
are more likely to succeed in human
clinical trials”.60
Organ-on-chip technology is
still quite new and there is little
standardisation: “Each team is
developing its own approach, with
its own unique technology. The
players are mainly start-up companies
commercializing prototypes developed
in the local universities”.61 It is a fast
 Zhang, et al., 2018
 Huh, et al., 2010
 T. H. Park & Shuler, 2003
 Ahadian, et al., 2018
 Ahadian, et al., 2018
 Haddrick, et al., 2018, p. 4
 Rothbauer, Rosser, Zirath, & Ertl, 2019, p. 84
 Archibald, et al., 2018, p. 2
 Wilkinson, 2019, p. 648
have great
potential for
something not
possible with
animal testing.
24 Humane Research Australia
Mini guts for
personalised treatment
of people with cystic
Cystic fibrosis is a genetic condition
that primarily aects the lungs and
digestive system. There is currently
no cure for the condition, which is
caused by mutations in the CFTR
gene that aect the function of the
CFTR protein. There are drugs for
the treatment of cystic fibrosis, but
dierent drugs work for dierent
mutations of the CFTR gene. Not all
drugs work for all patients.
Researchers in the Netherlands
have shown how these drugs could
be tested for individual patients.
They took rectal biopsies from
patients, grew them into mini guts,
and tested the available drugs for
eectiveness. They showed that
the drug responses observed in the
mini guts could be used to predict
which cystic fibrosis patient would
respond to which drug. This test can
help to quickly identify the best drug
therapy even when patients carry
very rare CFTR mutations.62
Cystic fibrosis research has been
carried out on dierent species of
animals, for example, mice, pigs and
ferrets.63 But these “animal models”
can’t predict which drug therapy will
work for an individual patient with
cystic fibrosis.
 Dekkers, et al., 2016
 Lavelle, White, Browne, McElvaney, &
Reeves, 2016
 Brownell, 2018
 Maoz, et al., 2018
 National Center for Advancing Translational
Sciences, no year-b
Brain-on-a-chip on drugs
A brain-on-a-chip has been dosed with the street drug crystal meth (crystal
methamphetamine, also called “ice”). Crystal meth is a stimulant that speeds up
the messages moving between the brain and the rest of the body. Researchers
from the US, Sweden and Israel connected three chips with dierent types of
brain cells to model the blood-brain barrier and then added crystal meth to
observe how the drug aects the brain. They were able to observe previously
unknown interactions between blood vessels and neurons in the brain.
“The human brain, with its 100 billion neurons that control every
thought, word, and action, is the most complex and delicate organ in the
body. Because it needs extra protection from toxins and other harmful
substances, the blood vessels that supply the brain with oxygen and
nutrients are highly selective about which molecules can cross from the
blood into the brain and vice versa. These blood vessels and their unique
network of supporting pericyte and astrocyte cells comprise the blood-
brain barrier (BBB). When the BBB is disrupted, as happens with exposure
to drugs such as methamphetamine (‘meth’), the brain’s sensitive neurons
become susceptible to harmful damage”.64
Many dierent species have been used in brain research. While rats and mice are
more likely to be used to study simple cognitive functions, non-human primates
have been subjected to invasive brain research because scientists believe their
brain processes are more similar to those of humans. With the development
of brains-on-chips, many complex processes in the brain can now be studied
without inflicting pain and suering on animals. The new methods are also faster
and cheaper.65
Blood-brain barrier on a chip
The blue dye shows where the brain cells
would go, and the red dye shows the
route for blood circulation.
Source: National Center for Advancing
Translational Sciences, US66
25Better ways to do research
 Cells from the umbilical cord.
 Mandt, et al., 2018
 Fibrin is a protein.
 Thermoplastics are types of plastic materials
which become so when they are heated
and hard when they cool down.
 Sriram, et al., 2018, p. 338
 Lee, Hwang, & Lim, 2017
Scientists at a university in Vienna
have created a placenta model
on a microchip made up of two
areas: one represents the mother,
the other represents the foetus. A
gelatin-based material was used to
provide a structure to which human
umbilical-vein endothelial cells67 and
cells from the placenta were added.
Many studies show that conditions
of the mother, such as diabetes
and high blood pressure, can have
an eect on the unborn child.
Compared with other research
methods, this can now be studied
in greater detail on the placenta-
on-a-chip. The researchers who
developed this chip intend to use
it to study how nutrients such as
glucose are transferred from the
mother to the foetus, specifically in
situations that involve a health risk
for the foetus.68
This placenta on a microchip models
the human situation much better
than animal experiments. It can
also be used to study conditions in
individual patients, something that
is not possible with animal models.
Microchips can test pharmaceuticals and cosmetics on human skin. A team
in Singapore has developed a credit-card sized device that can overcome
a limitation of traditional skin culture systems, which use cell cultures on a
collagen matrix that easily shrinks. Instead, the team developed a new method
that prevents skin contraction, using a fibrin69-based matrix. Their microfluidic
chip has several chambers. This allowed the researchers to grow skin in the
device and conduct tests without having to transfer the skin.
The chip is made of thermoplastics70 and can be mass produced. This skin-on-
a-chip is suitable for high-throughput screening. The researchers wrote that it
has “enormous potential to revolutionize many pharmaceutical, toxicological,
and cosmetic applications, including safety and eicacy, which currently rely on
animal testing”.71
Cosmetics have been tested on many species of animals, but perhaps best known
is the Draize test. In 1944, John Henry Draize (1900-1992) and his colleagues at
the US Food and Drug Administration (FDA) developed Draize rabbit irritation
tests for identifying and evaluating toxic reactions when test materials are
in contact with the skin, penis, and eyes. The tests were originally used for
evaluating the safety of cosmetics and then extended to include insecticides,
sunscreens and antiseptics.
Draize tests are painful and many animals are killed aer the tests. These tests
have been criticised because of large variations in test results and because of
dierences in the skin and eyes of humans andrabbits.72
26 Humane Research Australia
Spinal cord nerve and blood vessel cells on a tissue chip
Source: National Center for Advancing Translational Sciences, US73
 National Center for Advancing Translational
Sciences, 2018f
 National Center for Advancing Translational
Sciences, 2018c
 National Center for Advancing Translational
Sciences, no year-c
Source: National Center for Advancing Translational Sciences, US75
Heart-on-a-chip (as compared to a dime)
Source: National Center for Advancing Translational Sciences, US74
27Better ways to do research
To study human cells and tissues,
researchers need a readily available
supply of these human biological
samples. They are stored in so-called
biobanks, which are simply tissue
banks. Biobanks use tissue le over
from clinical procedures such as
surgery and from dead bodies, or they
collect tissue specifically for research.
Biobanks also store organoids.
“Organoid biobanking
is a promising and
exciting new field with
considerable potential
for scientific research,
precision medicine,
and regenerative
While biobanking has great potential,
there are also obstacles:
“A number of key factors limit
the wide adoption of non-animal
human tissue models in cancer
research, including deficiencies in
the infrastructure and the technical
tools required to collect, transport,
store and maintain human tissue
for lab use. Another obstacle is the
long-standing cultural reliance on
animal models, which can make
researchers resistant to change,
often because of concerns about
historical data compatibility and
losing ground in a competitive
environment while new approaches
are embedded in lab practice”.77
However, these obstacles can be
overcome by improving infrastructure
and collaboration. The collection of
tissue also raises ethical issues, such
as donor consent. Unless the tissue
samples are used for personalised
medicine, they are de-identified
for biobanking. But complete de-
identification may not always be
possible and a de-identified sample
can’t be used to benefit the donor.
 Bredenoord, Clevers, & Knoblich, 2017
 Jackson & Thomas, 2017
28 Humane Research Australia
‘Omics’ technologies
The term “omics technologies” refers
to areas of study in biology whose
names end in “omics”, such as
genomics. There are too many omics
areas of study to list here. The most
common omics include:
Genomics – the field of science
focusing on the structure, function,
evolution, mapping, and editing of
Metabolomics – the study of the
set of metabolites79 in an organism,
cell, or tissue. It holds great promise
for precision medicine. Small
numbers of metabolites have
been used for decades to diagnose
metabolic diseases, such as the
development of blood glucose
test strips in the 1950s to test for
diabetes. With metabolomics, much
larger numbers of metabolites than
are presently covered in standard
clinical laboratory techniques
can be examined.80 This allows
for personalised diagnosis and
treatment of individual patients.
Proteomics – the large-scale study of
proteins, particularly their structures
and functions.
Transcriptomics – the study of
messenger RNA (ribonucleic acid)
molecules produced in an individual
or population of a particular
Recently, a number of beliefs about our
understanding of human biology have
been called into question following
studies using omics approaches.
Such beliefs include the idea that
the genome is static throughout an
organism’s lifetime, that it is identical
in all cell types, and that all of the
necessary information for cellular
function is contained within the
“Traditional toxicology evaluates
end points such as death, disease
or observable changes in the
organism or cells of the organism,
while ‘omics’ measurements are
made across multiple levels of
biological organisation and provide
information that may be used to
understand cellular processes
as an integrated system rather
than as a collection of disparate
Using several of these omics
approaches together can help scientists
learn about toxicity pathways in human
cells and study the underlying cause
“The growing use of ‘omics
technologies (e.g. transcriptomics,
proteomics and metabonomics)
in combination with in vitro test
systems allows a comprehensive
analysis of the impact of a
chemical at the molecular level
and can indicate potential toxicity
pathways that may lead to adverse
health effects”.84
 A genome is the genetic material of an
 Metabolites are small molecules necessary
for metabolism.
 Clish, 2015
 Our genetic material is encoded in DNA. RNA
is similar to DNA, but has another function:
it communicates genetic information to the
rest of the cell.
 McBride, 2017
 McBride, 2017, pp. 69-70
 European Commission, 2018b
 Rzeznik, et al., 2017
 Struillou, Boutigny, Soueidan, & Layrolle,
Metabolomics approach
to identify gum disease
in its early stages
Periodontitis is an advanced gum
disease that involves inflammation
of the gums and the supporting
structure of the teeth. If untreated,
it leads to the loss of teeth. The
condition is very common and oen
diagnosed late when substantial
damage has already occurred. It is
preventable when diagnosed early.
A group of researchers in France
tested saliva from people with
and without periodontitis, using
nuclear magnetic resonance
(NMR) spectroscopy together with
multivariate statistical analysis to
identify the metabolic signature of
active periodontitis.85 They looked
for a range of metabolites and
found that a combination of lactate,
GABA, and butyrate predicted the
presence of periodontitis. The
study showed that this simple and
non-invasive method could be used
for early diagnosis and follow-up of
Rats, mice, hamsters, rabbits, ferrets,
sheep, pigs, cats, dogs and non-
human primates have been used
for modelling human periodontal
diseases and treatments. Of
these animal species, non-human
primates and dogs are considered
to be closest to humans in their
anatomy and the way in which
dental disease develops.86 This
new method developed by the
researchers in France is more
accurate than animal models.
29Better ways to do research
Stem cell technologies
When cells are used to test drugs or
other substances, it is important that
they are reliable representatives of
cells or cellular systems in the human
body. Many traditional in-vitro methods
that have been used for decades, such
as growing cells in a two-dimensional
way on a Petri dish, do not meet these
criteria as they are using immortalised
cell lines or isolated primary cells.
Immortalised cells87 can be grown
indefinitely and they are easily cloned
and cost eective, but they are
genetically altered and may react in
dierent ways to cells that have not
been altered. Isolated primary cells
are cells from human or animal tissue.
They are of varying quality and may
not react in a consistent fashion.88 Stem
cells are a promising alternative source
of human cells for toxicity testing,
studying and treating disease.
Stem Cells Australia, an initiative
that links Australia’s leading experts
in bioengineering, nanotechnology,
stem cell biology, advanced molecular
analysis and clinical research,
explained these technologies in the
following way:
“What type of stem cells are there?
Stem cells can be divided into two
broad groups: tissue specific stem
cells (also known as adult stem
cells) and pluripotent stem cells
(including embryonic stem cells
and iPS cells). Tissue specific stem
cells are derived from, or resident
in, adult tissues, and can usually
only give rise to the cells of that
tissue, thus they are considered
multipotent. Embryonic stem cells,
derived from a small group of cells
in the early embryo (5-7 days), and
iPS cells are undifferentiated and
are considered pluripotent as they
can become every type of cell in
the body.
What are induced pluripotent stem
cells (iPS)?
Recently, scientists discovered that
a mature fully specialised cell, for
example a human skin cell, in the
right conditions could be induced
to mimic the characteristics of an
embryonic stem cell. These are
known as induced pluripotent stem
cells (iPS cells)”.89
Induced pluripotent stem cells were
first developed in 2006 through genetic
reprogramming of adult cells.90 The
Nobel Prize in Medicine in 2012 was
awarded for the discoveries that cells
can be reprogrammed to become
pluripotent stem cells, and that cells
from individual patients could be
harvested and reprogrammed to
become any tissue type found in the
human body.91
The use of stem cells has great potential
for studying and treating diseases.92
In theory, there is no limit for the
type of diseases that could be treated
with stem cell methods. However,
new methods have to be tested first
to ensure they are safe and eective.
Some stem cell therapies are already
being used in cancer treatments and
bone marrow transplantation. Human
stem cells are also now being used to
test drugs.
Parkinson’s disease
identified as a possible
autoimmune disease
A group of researchers at a German
university has gained new insights
into the development of Parkinson’s
disease by using a stem cell
approach. The researchers had
observed an unusually high number
of T cells93 in the midbrain and
the blood of Parkinson’s patients.
These T cells – a type of T cell
similar to those found in people
with autoimmune diseases – attack
and kill nerve cells that produce
dopamine in the midbrain.
With this observation in mind, the
researchers took small skin samples
from healthy people and people
with the disease, and dierentiated
these cells into stem cells that
function like midbrain nerve cells.
These were then brought into
contact with fresh T cells from the
people who had donated the skin
samples. The researchers found
that the stem cells from people with
Parkinson’s disease killed a large
number of their nerve cells, but
healthy people’s cells did not react
in this way.94
 cells that continue to divide indefinitely
 McBride, 2017
 Stem Cells Australia, 2018
 Brevini, et al., 2016
 Rothbauer, et al., 2019
 Triunfol, et al., 2018
 T cells are a type of white blood cells.
 Friedrich-Alexander Universitaet Erlangen-
Nuernberg, 2018; Sommer, et al., 2018
30 Humane Research Australia
Human stem cells could potentially
be used as a renewable source of
replacement cells and tissues to treat
diseases such as macular degeneration,
spinal cord injury, stroke, burns, heart
disease, diabetes, osteoarthritis and
rheumatoid arthritis.95 Stem cell banks
are being established worldwide and
the global market for human-induced
pluripotent stem cell technology
The use of embryonic stem cells has
raised questions about the ethics of
their use, as each embryonic stem cell
line has been grown from a human
embryo created through in-vitro
fertilisation (IVF) or through cloning
In Europe, a large public-private
partnership project has set up
biobanks for induced pluripotent
stem (iPS) cells, the European Bank
for induced pluripotent Stem Cells
(EBiSC).97 EBiSC anticipates that its
capacity will be 10,000 cell lines. At
the time of writing, EBiSC’s online
catalogue included more than 800
dierent cell lines and an additional
240 were in the process of being
tested for quality.
The use of human stem cells can
replace many experiments on animals,
in particular “humanised mouse
models”, that is mice that have been
engraed with human cells and tissues
to give them human diseases. Although
scientists have been able to cure
cancer in mice for decades98, these
animals are still being used to study
 National Institutes of Health, 2016
 Archibald, et al., 2018
 European Bank for induced pluripotent Stem
Cells, 2018
Neural stem cells in 3D culture models
Scientists at the University of North Carolina at Chapel
Hill (UNC) determined the anti-tumour eects of neural
stem cells99 in 3-D culture models. Stem cell aggregates, or
spheroids (blue in the images below), were placed next to
brain cancer cell spheroids. Fluorescent images captured over
seven days showed that the stem cell therapy decreased the
volume of cancer spheroids. The researchers are working to
advance the therapy towards first-in-human clinical trials.
Neural stem cells in 3D culture models
Source: National Center for Advancing Translational Sciences, US100
 Cimons, Getlin, & Maugh, 1998
 Neural stem cells are cells that generate
the neurons and glia of the nervous
system during embryonic development.
 National Center for Advancing
Translational Sciences, 2018e
31Better ways to do research
 Musah, Dimitrakakis, Camacho, Church,
&Ingber, 2018
 Navarro, Susanto, Falk, & Wilhelm, 2018
 Navarro, et al., 2018, pp. 7-8
 Heslop & Duncan, 2018
Glomerulus chip with
stem cells to study
kidney disease
Researchers from the Wyss Institute
for Biologically Inspired Engineering
at Harvard University have
developed a protocol to grow kidney
cells from induced pluripotent stem
cells (iPS cells) within a microfluidic
organ-on-a-chip to build a human
kidney glomerulus chip that mimics
the structure and function of the
kidney glomerular capillary wall.
The glomerulus chip can be used to
study human kidney development,
function and disease. It has the
potential to be used in regenerative
medicine as a cell therapy for people
with kidney diseases.101
Using induced
pluripotent stem cells
to understand the
development of cancer in
Reprogramming of human cells that
can be easily obtained from biopsies
or blood samples to induced
pluripotent stem (iPS) cells can help
as a prognostic tool and to discover
familial cancers early. Patient-
derived iPS cells from a tumour can
also help develop specific drugs for
that individual child.102
“The discovery of iPS cells
opens up a wide spectrum of
possible future applications
including development of new
treatments in regenerative
medicine, generation of better
and more accurate disease
models, and improving drug
discovery. Where other tools
fail, using patient-specific iPS
cells for modeling diseases
have an incredible potential to
improve our understanding of
basic mechanisms operating
during healthy and diseased
human development and
Human pluripotent stem
cells for modelling liver
development and disease
Pluripotent stem cells can help
scientists understand liver disease
much better than traditional
methods, such as studies using mice.
Mice are dierent from humans
in diet, genetics, gene expression
and physiology. Mouse studies
require time and are unsuitable
for high-throughput methods. In
contrast, iPS cells can be expanded
indefinitely and dierentiated
into liver-like cells that have many
functional characteristics of human
livers. Mutations can be introduced
through genome engineering to help
researchers study liver development
or to model rare liver diseases.
Studying liver disease, such as
hepatitis A and C, and disease
development using iPS cells can be
undertaken with semi-automated,
high-throughput methods that
make it possible to screen a wide
range of pathways and functions
32 Humane Research Australia
3D and 4D bioprinting
Bioprinting, or tissue printing, is one of
the technologies of tissue engineering
and regenerative medicine (TERM),
which involves researchers from
dierent disciplines, such as medicine,
engineering, biology and chemistry. In
recent years, they have achieved much
progress in TERM.105
Three-dimensional (3D) printing was
first conceived in 1986 and has since
influenced fields such as engineering,
manufacturing and medicine.106 While
there are dierent technologies of 3D
printing, it basically involves processes
where materials or liquids are joined,
layer by layer, under computer control,
to create three-dimensional objects.
The terms 3D printing and 3D
bioprinting have dierent meanings.
Both processes build a 3D object
layer by layer from a 3D model but
“3D bioprinting involves the use of
cell-laden bioinks and other biologics
to construct a living tissue while 3D
printing technologies do not use cells
or biologics”.107
3D printing technologies were
originally designed for non-biological
applications, such as metals, ceramics
and thermoplastic polymers. The
process used organic solvents, high
temperatures or materials that are
not compatible with living cells and
biological materials. A challenge for
TERM scientists then, is to keep the
cells alive during the printing process.
For 3D bioprinting, the researchers had
to find materials and printing processes
that are compatible with living cells and
tissues.108 They also had to find suitable
materials to build scaolds that contain
and shape the biomaterials in a desired
form. The scaold materials can be
natural or synthetic.
“Bioprinting is no longer confined
to a process for combining one
cell type with one material;
the emphasis today is to use a
variety of material types to create
bespoke scaffolds onto which
chemical cues can be tethered and
multiple cell types can be printed
with precision”.109
 K. M. Park, Shin, Kim, & Shin, 2018
 Bishop, et al., 2017
 Vijayavenkataraman, Yan, Lu, Wang, & Fuh,
2018, p. 2
 Murphy & Atala, 2014
 Mehrban, Teoh, & Birchall, 2016, p. 13
 National Center for Advancing Translational
Sciences, 2018a
 National Center for Advancing Translational
Sciences, no year-a
3D image of bioprinted skin tissue
Source: National Center for Advancing Translational Sciences, US110
3D printed eye tissue
Source: National Center for Advancing Translational Sciences, US111
33Better ways to do research
Researchers have developed many
dierent approaches to 3D bioprinting,
Laser-based bioprinting – laser
energy is used to pattern bioinks
laden with cells.
Droplet-based bioprinting – cell-
laden bioinks are ejected out of the
nozzle in the form of droplets.
Extrusion-based – bioink is pushed
out of the nozzle using pneumatic
pressure or mechanical force by
means of a piston or screw. This is the
most widely used type of bioprinting.
Extrusion-based bioprinting has been
used to bioprint cells, tissues and
organ-on-a-chip devices for tissue
engineering, cancer research, drug
testing and transplantation.
Stereolithography bioprinting
where a “layer of photopolymer resin
is cured (or polymerized) by light
(usually UV) irradiation, the light
movement controlled by a computer
code/images/CAD files, forming a
3D structure as the build stage is
translated vertically building the
object layer by layer.112
“In laser-based writing, which
consists of a laser beam, a substrate,
and a focusing system, cells are
confined in a laser beam and
deposited in a steady stream on
nonabsorbing surfaces, including
biological gels. As such, cells
can be printed continuously
and accurately without causing
significant cell death. Laser-based
writing can pattern cells with high
resolution up to the micrometer
scale and is advantageous over
other bioprinting techniques as it
can be used with many materials
and does not directly bring the
live cells into contact with the
substrate, improving cell survival.
This technique has been used to
produce vascular networks with
micrometer precision on biological
gels in vitro”.113
A disadvantage of laser-based
bioprinting is its slow printing speed
which makes this method unsuitable
for printing of large tissues or organs.
“In inkjet printing, which consists
of a reservoir tank, an orifice, and
a print head, a pressure is created
in the tank, which pushes the ink
in the orifice and out to the printer
head. As a result, cell droplets
are deposited on a surface, which
provides the advantage of limiting
contact of cells and materials on
a surface. Other advantages of
inkjet printing include controllable
resolution, high printing speed, and
relatively low material costs”.114
Inkjet printing is the most popular
printing method. It is fast, cheap and
readily available. The disadvantages
include print-head clogging,
mechanical stress and unreliability in
bioink dispensing.115
“Microextrusion printers use
pneumatic or mechanical (piston or
screw) dispensing systems to extrude
continuous beads of material and/or
cells”.116 This method has already been
used to produce, for example, aortic
valves117 and an ovarian cancer disease
model.118 But it has disadvantages, too.
They include limited spatial resolution
and, as high pressures are used, part of
the living material may not survive the
printing process.119
 Vijayavenkataraman, et al., 2018, p. 11
 Ahadian, et al., 2018, p. 19
 Ahadian, et al., 2018, p. 20
 Mehrban, et al., 2016
 Murphy & Atala, 2014, p. 775
 Duan, Hockaday, Kang, & Butcher, 2013
 Xu, et al., 2011
 Mehrban, et al., 2016
 National Center for Advancing Translational
Sciences, no year-e
Organovo bioprinter
Source: National Center for Advancing Translational Sciences, US120
34 Humane Research Australia
A research group from Singapore
described the typical steps involved in
bioprinting in the following way:
“A typical bioprinting process
consists of three major steps namely
pre-processing, processing and
post-processing. Preprocessing
involves imaging of the tissue or
organ using computed tomography
(CT), magnetic resonance imaging
(MRI), and ultrasound imaging
techniques and reconstruction
of 3D models from the imaging.
The generated 3D models are
then converted into STL file
format, which is a commonly
accepted file format by most of the
commercially available bioprinters.
The processing step starts with
harvesting primary cells from
patients, culturing and expanding it
ex vivo for the bioprinting process.
Though cancer cell lines and other
non-human cells are being used,
the ideal condition for fabricating
transplantable living tissues would
be to use the patient’s own cells.
Suitable bioinks with properties
mimicking the intended tissue to be
printed are selected and the cells are
suspended in these bioinks. The cell-
laden bioinks are then fabricated
into required 3D living tissue/
organ according to the 3D model
using a bioprinter. Post-processing
involves maintaining the bioprinted
tissue/organ in a bioprinter for
tissue maturation before being
transplanted into patients or used
as in vitro models for disease
modelling, or drug testing”.121
Dierent cell types can be used for
bioprinting. Stem cells are particularly
appealing “as they are pluripotent
and able to dierentiate into other cell
types upon exposure to the correct
physical and chemical guidance cues.
Within the human body, there are a
number of viable sources of stem cells,
such as the bone marrow, periosteum
and adipose tissue”.122
Bioprinting has many applications in
biomedical research and the health-
care sector, such as disease modelling,
drug discovery and testing, high-
throughput screening and regenerative
medicine. Bioprinting has already
contributed to much progress in skin
and cartilage repair.123
“Advanced biofabrication
techniques, such as 3D printing
and 3D microfabrication can also
be applied for the creation of tissue
models, and organoid technologies
can explore stem cells for the
formation of organ- like structures
to accurately and simultaneously
capture multiple aspects of
human physiology”.124
“Bioprinting has a great potential
to solve this ever-increasing organ
shortage crisis. Though bioprinting
of fully functional organs has a long
way to go, considerable progress
has been made to realize the greater
goal of organ printing. Bioprinted
tissues could be used as in vitro
testing beds in place of animal
testing. Given the ethical concerns
surrounding animal testing and the
high cost involved, bioprinting is a
viable alternate. In pharmaceutical
research, bioprinting could be used
as in vitro models for testing of drug
efficacy, toxicity, chemotherapy
or chemo-resistance to reduce the
high cost and shorten the time of
drug discovery”.125
Current 3D bioprinting technologies
have several limitations. Some
techniques don’t allow for the mixing
of dierent types of cells or make such
mixing very diicult, for instance,
high temperatures or voltages
may compromise the living cells or
tissues. “Thus, there is always a trade
between resolution, compatibility
with cell deposition, cell viability
as well as mechanical stability and
no one of the existing 3D printing/
bioprinting methods is able to provide
4D printing, which is a more recent
development, can overcome some
of the challenges associated with
3D printing. 4D bioprinting has been
described as 3D bioprinting with the
added dimension of time127, or shape
transformation over time.128
The new printing technique has
been defined as 3D printing of cell-
laden materials in which the printed
structures would be able to respond
to external stimulus or internal cell
forces.129 Simply, 4D bioprinted
materials are capable of changing their
shape over time. While 4D printing is
still in its infancy, it has great potential
for tissue engineering in many areas.
The current demand for donor
organs to replace damaged or lost
organs is higher than the availability
of donor organs, for instance, and
4D printing could one day provide
individualised organs or part-organs
to fulfil that demand. Current attempts
to genetically engineer animals such
as pigs to become organ donors for
humans will then be redundant.
“The significantly enhanced
usability and functionality of the
bioprinted objects due to their
capability to transform with
time will likely find widespread
applications in areas including but
not limited to tissue engineering,
regenerative medicine,
bioelectronics, robotics, actuators,
and even medical devices”.130
 Vijayavenkataraman, et al., 2018, p. 3
 Mehrban, et al., 2016, p. 12
 Mehrban, et al., 2016
 Zhang, et al., 2018, p. 258
 Vijayavenkataraman, et al., 2018, p. 2
 Ionov, 2018, p. 1
 Ashammakhi, et al., 2018
 Li, et al., 2017
 Ashammakhi, et al., 2018
 Li, et al., 2017, pp. 4-5
35Better ways to do research
A 3D bioprinted
vascularised liver tissue
model for drug testing
An international collaboration of
researchers has developed and 3D
bioprinted a vascularised liver tissue
model that allows them to mimick
drug diusion and accurately test
whether a drug is toxic to humans.
A vascularised tissue is a tissue
with vessels, especially blood
vessels. Drug diusion refers to the
movement of a substance or drug
from an area of high concentration
to an area of lower concentration.
Diusion is an important process
in living organisms. It occurs in
liquids and gases, and allows
substances to move in and out of
cells. The researchers noted that
this model could be expanded and
used to model other organs for drug
testing, such as the heart, kidney
Bioprinting can and has already been
used in combination with organs-on-
chips. Although it is diicult to recreate
complex bodily functions, the models
that are created with organ-on-chip and
bioprinting technologies allow for drug
testing and the study of physiological
functions in-vitro. The combination
of both technologies can build tissues
faster than organ-on-chip technology
alone. It creates high throughput in the
production of organs-on-a-chip, and it
helps researchers to develop new types
of organ-on-a-chip.132
Robotic testing
Researchers, particularly those in
the pharmaceutical industry, have
developed automated methods to
test biological activities of thousands
of chemicals that used to be tested in
animals. This is called high-throughput
or robotic testing.
“The last two decades have seen
innovations in technology that
have helped to evolve automated,
microprocessor controlled robotic
processes called High Throughput
Screening (HTS). This qualitative
leap in drug discovery paradigm
has been achieved via a synergy
of chemistry, biology, engineering
and informatics”.133
In the US, high-throughput testing
is being supported by a government
initiative called Toxicology in the 21st
Century (Tox21). It is a collaboration of
several government agencies that aims
“to develop better toxicity assessment
methods to quickly and eiciently test
whether certain chemical compounds
have the potential to disrupt processes
in the human body that may lead to
negative health eects”.134
“In the United States, Tox21 is
a multi-agency collaboration
to research, develop, validate,
and translate chemical testing
methods in order to characterize
chemical toxicity pathways in [sic]
interest of protecting public health.
This initiative became official
through the establishment of a
Memorandum of Understanding
between four major agencies: (1)
the US Environmental Protection
Agency (EPA)/Office of Research and
Development, (2) the US National
Institutes of Health (NIH), National
Institute of Environmental Health
Sciences/National Toxicology
Program, (3) the NIH/National
Human Genome Research Institute/
Chemical Genomics Center,
and (4) the US Food and Drug
Administration (FDA). A central
component of this initiative is to
employ high throughput screening
assays and genome analytical
methods to identify mechanisms of
chemical induced biological activity,
prioritize over 10 000 chemicals
for more extensive toxicological
evaluation, and develop
computational predictive models of
in vivo biological response. … The
major objective of Tox21 is to deliver
biological activity profiles that are
predictive of in vivo toxicities for
the thousands of substances that
are being studied over the 5-year
 Massa, et al., 2017
 J. Y. Park, Jang, & Kang, 2018
 Ranganatha & J, 2012, p. 30
 United States Environmental Protection
Agency, 2018b
 Valerio, 2014, pp. 1026-1027
36 Humane Research Australia
Testing of nanomaterials is an area
that would greatly benefit from high-
throughput testing. Manufactured
nanomaterials (materials with at
least one dimension <100 nm143) and
nanoparticles (all three dimensions
<100 nm) are used in many consumer
products, but their eects on human
health are not well known. While some
high-throughput testing methods of
nanomaterials are in development,
they have not yet been validated.144
High-throughput testing is also used
with organs-on-chips. This technology
is at present still too expensive for
widespread use but work is underway
to make its pricing more competitive so
that it can replace conventional in-vitro
and in-vivo models.145
The US Environmental Protection
Agency (EPA) contributes to Tox21
through its computational toxicology
research (CompTox), which can
evaluate chemicals for potential risks
quickly and at a small cost. These
high-throughput screening assays
can outperform the predictive ability
of models built with animal toxicity
data.136 A key part of EPA’s CompTox is
its Toxicity Forecaster (ToxCast). Data
generated by ToxCast are publicly
available online. They complement the
European Chemicals Agency (ECHA)
database and contain “results from
several thousands of in vitro tested
chemicals, measuring hundreds of
endpoints each”.137 ToxCast is a useful
resource to inform read-across .138 139
“Tox21 has established a
library of ~10,000 chemicals
for the production phase of the
program, including the NCATS
Pharmaceutical Collection (NPC),
which contains drugs used in
the clinic. This library has been
screened against 47 cell-based
assays in a quantitative high-
throughput screening (qHTS) format
generating nearly 70 million data
points to date”.140
 Archibald, et al., 2018
 Maertens, Hubesch, & Rovida, 2016
 Chesnut, et al., 2018
 See section on in-silico methods for detail
on read-across.
 R. Huang, et al., 2018, p. 2
 Judson, et al., 2010
 Judson, et al., 2010, p. 5984
 One nanometre (nm) is equal to one
billionth of a meter.
 Collins, et al., 2017
 Probst, Schneider, & Loskill, 2018; van
de Burgwal, van Dorst, Viëtor, Luttge, &
Claassen, 2018
The Deepwater Horizon
oil spill in the Gulf of
Mexico and ToxCast
The 2010 BHP oil disaster in the
Gulf of Mexico was the largest
marine oil spill in the history of the
petroleum industry. To break up
the oil slick, more than 1.5 million
gallons of the oil spill dispersant
Corexit 9500 was used. Very little
was known about the eects of
the dispersant chemicals on acute
or long-term toxicity in humans or
marine animals. When the US EPA’s
Oice of Research and Development
was asked to evaluate the potential
toxicity of oil spill dispersants, the
researchers needed a fast testing
method. They decided to use high-
throughput in-vitro tools that are
part of the EPA’s ToxCast program.141
No animals were used in these
tests. Indeed, animal tests could not
have provided such quick and low-
“ … we were able to detect specific
bioactivities in complex chemical
mixtures for time-sensitive
environmental issues and using
high throughput screening
assays. This is exciting given
that one of the challenges of real
world chemical toxicity testing
is the fact that humans and other
organisms are often exposed to
complex mixtures, rather than
the pure single compounds that
are the subject of typical toxicity
testing. The in vitro tests used
in this study rapidly profiled the
complex dispersant formulations
without the use of animals, and
screened for potential endocrine
activity, other endpoints
and cytotoxicity. In different
circumstances, a similar rapid
screening effort could be used to
make time-sensitive decisions
based on potential hazard
and risk”.142
37Better ways to do research
In-silico methods
 In silico – performed on a computer or via
computer simulation.
 Meigs, et al., 2018, p. 286
 Maertens & Hartung, 2018
 Chesnut, et al., 2018
 National Center for Advancing Translational
Sciences, 2018b
 Meigs, et al., 2018, p. 289
 Valerio, 2014, p. 1026
 Myatt, et al., 2018
Progress in information technology has enabled the development of
computer-based (in-silico146) methods for biomedical research and the
testing of chemical and biological substances. To date, most progress
with in-silico methods has been made in the area of toxicology (the study
of the adverse effects of chemical substances on living organisms). Many
computational methods have been developed to predict toxicity, and can
thus play a role in replacing and/or reducing animal testing. The main
prediction methods are described in this section. Computer modelling of
health and disease is another area where in-silico methods can replace the
use of animals.
“Chemical products are used in making
95% of all goods”.147 Chemical and
biological substances, such as newly
developed drugs, need to be tested to
determine whether they are harmful to
humans, animals or plants. But there
is limited information about most
substances because testing is very
expensive. The US Toxic Substance
Control Act inventory includes 83,000
chemicals, but has data for only 3% of
these.148 Safety testing reportedly costs
US$10 – 20 million per product and
takes several years. It is also diicult to
obtain toxicological data for the 1,000
new chemicals created every year.149
The development of new drugs and
pesticides is lengthy and extremely
expensive, as the following statements
“Therapeutic development is a
costly, complex and time-consuming
process. The average length of time
from target discovery to approval
of a new drug is about 14 years. The
failure rate during this process
exceeds 95 percent, and the cost per
successful drug can be $1 billion
or more”.150
“The probable range confronting
developers of new pesticide
chemicals appears to be $750,000
to $3.25 million – but the trend is
constantly upward. On average it
costs €2.2 billion and takes 10 years
for a new active substance to be
brought to market according to the
European Trade Association”.151
It is opportune, then, that computers
can help researchers in the safety
sciences by performing large amounts
of complex calculations at great
speed to reveal patterns, trends and
associations. In silico methods have
many applications, as a scientist from
the US Food and Drug Administration
“In silico methods for toxicology
apply modern computing
applications and informatic
technologies as scientific tools
to advance and gain efficiency
to improving our understanding
of toxicity potential, pathways,
hazards, and risks of chemical and
biological substances. Multiple
computer technologies and
methodologies serve to store,
interface, process, or transmit
information which enable scientific
advancements in the toxicological
sciences through implementation
of predictive models, databases,
detection systems, and simulations
The in-silico prediction methods for
the evaluation of toxicity outlined
below cover the most common
methods and models but they do
not represent a comprehensive list.
Several organisations and companies
have produced soware packages for
predicting toxicity or physicochemical
properties of chemicals. In general, they
contain one or more predictive models.
Rapid progress is being made in this
area and soware, models and datasets
are being constantly updated.153
testing and
methods are
booming in
the toxicology
38 Humane Research Australia
The silico methods for predicting and
evaluating the toxicity of chemical and
biological substances described in this
section include:
Structure-activity relationship (SAR)
and quantitative structure-activity
relationship (QSAR) modelling
pharmacokinetic (PBPK) models
Expert systems
This is followed by a selection of tools:
Read-across structure activity
relationships (RASAR) modelling
There are also some approaches that
are not considered a method or a tool:
Adverse Outcome Pathways
(AOPs) - a conceptual approach
to representing knowledge about
toxicity mechanisms
Integrated Approaches to Testing and
Assessment (IATA)
Lastly, this section provides an
overview of computer modelling for the
study of human organs and the body,
and how this can be used in clinical
settings154. The use of computers for
virtual reality and surgical simulation,
mainly for purposes of education and
training, will be covered in the section
on simulators.
 Obviously, such modelling can also be used
for the study of non-human animals.
 Marshall & Willett, 2018, p. 1952
 Marshall & Willett, 2018
 Bal-Price, et al., 2018
 Marshall & Willett, 2018, p. 1956
Computer-based approaches for the study of
Parkinson’s disease
Parkinson’s disease (PD) is an age-related neurodegenerative disease whose
cause or causes are largely unknown and for which a cure has not yet been found.
During the 1950s, it was discovered that the psychiatric drugs reserpine and
haloperidol induce Parkinsonian-like symptoms in humans. That discovery led
to the drugs becoming some of the first used to develop animal models. That
is, these drugs were given to animals to simulate Parkinson’s disease. Decades
and many animal models later, no animal model can fully explain the disease in
“More than 60 years have been spent attempting to produce PD-like
symptoms in various animal species as a model to study disease
pathogenesis and treatment. In addition to rodents and nonhuman
primates, Caenorhabditis elegans (roundworms), Drosophila
melanogaster (fruit flies), fish (zebrafish and goldfish), and anurans
(frogs and toads) have been used. However, each animal model generally
demonstrates only specific subsets of PD characteristics, and no single
model recapitulates all the known pathological processes associated with
PD in humans”.155
New approaches are needed that focus on humans, not on animals. These
include in-vitro and in-silico approaches. In-vitro approaches consist of, for
example, methods that use stem cells. In-silico approaches include computer
simulations, mathematical algorithms and machine learning to predict
interactions between molecules and pathways within biological systems.156
Much information about Parkinson’s disease exists already, for example in the
Organization for Economic Cooperation and Development (OECD) iLibrary
on Adverse Outcome Pathways.157 With the help of information technology,
existing knowledge can be pooled and shared. The OECD, the US Environmental
Protection Agency, and the European Union Joint Research Centre are already
collaborating on a shared resource “that covers the broad spectrum of biological
pathways that are likely to be involved in human health and ecological risk
assessment, the Adverse Outcome Pathway Knowledge Base”.158
39Better ways to do research
In-silico prediction methods for the evaluation of toxicity
SARs and QSARs
SARs and QSARs are mathematical
models used for predicting biological
activities of chemicals.
“A SAR is a qualitative relationships
[sic] that relates a (sub)structure
to the presence or absence of a
property or activity of interest.
The substructure may consist of
adjacently bonded atoms, or an
arrangement of non-bonded atoms
that are collectively associated with
the property or activity”.159
For over a century, scientists have had
some knowledge of the relationship
between the structure of a substance
and its toxicity. The knowledge of this
structure-activity relationship (SAR)
became well known, but was rarely
gathered and systematically recorded
until the 1980s. With progress in
information technology, this changed:
“Since the early 1990s, the concept
of applying compilations of
structural alerts to evaluate and
predict toxicity has progressed
rapidly, with the expansion to
other endpoints and effects. …
The development of compilations
of structural alerts has also been
stimulated by the availability of
software that can capture them, and
allow users to apply that knowledge.
Bearing in mind that, until the late
1980s, the alerts were available only
on paper, e.g. in journal articles or
book chapters, they were difficult to
use and open to misinterpretation,
and may have been applied
differently by different users”.160
Quantitative structure-activity
relationship (QSAR) modelling uses
one or more quantitative parameters
derived from a chemical structure to
a property or activity of interest. The
term “quantitative” in QSAR refers to
the nature of the parameter(s) used
to make the prediction. Quantitative
parameters enable the development
of quantitative models. QSARs are
mathematical models derived from a
training set of example chemicals. “The
training set includes the chemicals that
were found to be positive and negative
in a given toxicological study … or to
induce a continuous response … that
the model will predict”.161
The models can be used to predict a
qualitative or quantitative endpoint.162
“It is assumed that chemicals that
fit the same QSAR model may work
through the same mechanism”.163
QSAR models have been used, for
example, for skin sensitisation, eye
irritation and other human health
hazards. They require a large database
with information on the structure and
toxicity of substances.
In 2007, changes in European
legislation on chemical substances
explicitly promoted the use of in-silico
models, in particular (Q)SARs164, as an
alternative testing method to evaluate
the safety of regulated substances:
“The legislation permitting the
acceptance of in silico models to
evaluate toxicity is Regulation
EC 1907/2006, better known
as Registration, Evaluation,
Authorisation, and Restriction of
Chemical substances (REACH).
REACH specifies conditions for the
use of in silico (Q)SAR models, and
the European Chemicals Agency
(ECHA) located in Helsinki, Finland
implements REACH. ECHA offers
guidance on how the (Q)SAR in silico
method can help fulfill or support
information requirements“.165
So many dierent QSAR models are
now available, that the proliferation of
these models has been referred to as
a “zoo of QSAR”166, and this can make
it diicult to choose the best one for
the task at hand. Further, researchers
have pointed out that “many QSARs are
diicult to interpret and cannot be used
to define causal links. They represent
correlation rather than causation”.167
Progress in computer-based methods
has seen improvements in dealing
with incomplete data sets and making
models less complicated:
“Recent advances in machine
learning have resulted in models
that can handle missing data and
model multiple targets at once
(multi-label learning, in case of
toxicology for example multiple-
hazard learning). These models can
sometimes outperform single-label
models by increasing the available
data for training and by transferring
concepts applicable to one label to
predictions on another label. Multi-
label models have the potential to
simplify the QSAR space. Rather
than having a model for every
chemical property, a single model
can predict many different chemical
properties. In toxicology, many
hazards are interrelated; thus, they
can inform each others’ predictions.
For example, a skin irritant is likely
also an eye irritant, which means
that information on both labels
synergizes. So, the prediction of one
hazard (label) informs other labels
for the same and similar chemicals
can improve predictions”.168
For all in-silico predictions of toxicity,
(Q)SAR and other models, standardised
protocols need to be developed.
“Such novel in silico toxicology (IST)
protocols, when fully developed
and implemented, will ensure in
silico toxicological assessments
are performed and evaluated in a
consistent, reproducible, and well-
documented manner across industries
and regulatory bodies to support
wider uptake and acceptance of the
 European Chemicals Agency, 2008, p. 10
 Cronin & Yoon, 2018, p. 291
 Myatt, et al., 2018, p. 6
 European Chemicals Agency, 2008
 Raies & Bajic, 2016, p. 156
 SARs and QSARs are collectively referred to
as (Q)SARs.
 Valerio, 2014, p. 1027
 Luechtefeld, Rowlands, & Hartung, 2018, p. 740
 Luechtefeld, Rowlands, et al., 2018, p. 739
 Luechtefeld, Rowlands, et al., 2018, p. 741
 Myatt, et al., 2018, p. 2
40 Humane Research Australia
Read-across is a method that is
usually based on chemical structure
information and uses QSAR
approaches.170 For example, the OECD
QSAR Toolbox171 can be used to support
read-across. The read-across method
uses data from a substance for which
toxicity information is available,
to make predictions for a similar
substance about which not much is
known. In other words, a data-rich
substance is used to make toxicity
predictions for a structurally similar but
data-poor substance.
While read-across is currently based
on chemoinformatic172 approaches, it
is not considered a QSAR: “although
read-across is informed by chemical
structure, it is strictly based on local
similarity of a chemical with similar
“Read-across uses data on one
or more analogs (the “source”) to
make a prediction about a query
compound or compounds (the
“target”). Source compounds are
identified that have a structurally
or toxicologically meaningful
relationship to the target
compound, often underpinned by
an understanding of a plausible
biological mechanism shared
between the source and target
compounds. The toxicological
experimental data from these
source compounds can then be
used to “read-across” to the specific
target compound(s). Read-across is
an intellectually-derived endpoint-
specific method that provides
justification for why a chemical is
similar to another chemical (with
respect to chemical reactivity,
toxicokinetics, mechanism/mode of
action, structure, physicochemical
properties, and metabolic profile)”.174
Read-across is fast and cost-eective,
but it relies on subjective assessments
of what constitutes a “similar”
substance. When comparing chemical
information for read-across, small
changes in the structure of chemicals
can result in big dierences in the level
of toxicity. This can lead to prediction
errors. Adding other information, such
as data from in-vitro tests, can improve
the accuracy of the prediction.175 Thus,
knowledge of biological similarity
enhances read-across. An international
team of researchers oers the following
“The increasing availability
of biological data via the data
sharing depositories will augment
such support of read-across and
grouping by big data. The curation
of such datasets and the respective
data-sharing by companies,
organizations and individual
researchers needs to be further
encouraged and possibly furthered
with some incentives”.176
In the EU, manufacturers and
importers have to register information
on chemical substances (that are
produced or imported in volumes
over one tonne a year) in a central
database at the European Chemicals
Agency (ECHA) in Helsinki. This is
part of the REACH regulation system:
registration, evaluation, authorisation
and restriction of chemicals. ECHA’s
2017 report on the use of alternative
methods revealed that the most
common alternative method on
analysing substances was read-
across (63%), followed by weight of
evidence (combining information
from dierent sources), and QSAR
predictions (34%)177. Much of the
information is still based on new or old
experimental studies using animals,
but ECHA reported that out of the 6,290
substances analysed for the report,
89% had at least one data endpoint
where an alternative to animal studies
was used.
At present, regulatory agencies in
the EU consider read-across the
best method in the areas of skin and
eye irritation.178 “Read-across is an
innovative approach that can be
considered an alternative to animal
testing – and at the moment it is
probably the most eective method of
reducing the use of lab animals”.179 The
authors of a review of in-silico methods
for the prediction of chemical toxicity
summarised the practical applications
of the read-across method and
provided examples of tools used:
“Read-across was applied to predict
carcinogenicity, hepatoxicity,
aquatic toxicity, reproductive
toxicity, skin sensitization, and
environmental toxicity. Examples of
tools implementing read-across are
The OECD QSAR Toolbox, Toxmatch,
ToxTree, AMBIT, AmbitDiscovery,
AIM, DSSTox, or ChemIDplus”.180
Read-across prediction is strengthened
by the availability of high quality
data, as well as agreed principles and
guidance on how to group chemicals.181
 Zhu, et al., 2016
 OECD, 2018c
 Chemoinformatics is focused on extracting,
processing and extrapolating meaningful
data from chemical structures.
 Chesnut, et al., 2018, p. 414
 Myatt, et al., 2018, p. 7
 Chesnut, et al., 2018; Zhu, et al., 2016
 Zhu, et al., 2016, p. 178
 European Chemicals Agency, 2017b
 Archibald, et al., 2018
 Maertens, et al., 2016, p. 324
 Raies & Bajic, 2016, p. 152
 Ball, et al., 2016; Patlewicz, et al., 2014
is a fast and
method for
41Better ways to do research
PBPK models
pharmacokinetic182 (PBPK) models are
“mathematical representations of the
absorption, distribution, metabolism
and elimination (ADME) of chemicals
in humans or other animal species.
They are used for multiple purposes,
including the interpretation of in
vitro toxicity data by in vitro to in vivo
extrapolation (IVIVE) and the simulation
of internal concentrations in the
organism of interest”.183
pharmacokinetic (PBPK) models
are computational systems
now commonly used in drug
development and increasingly in
regulatory toxicology. They predict
the absorption, distribution,
metabolism and excretion (ADME)
of substances in the body, at
different doses. PBPK models based
on human rather than animal
characteristics avoid the problem
of species differences. They can
also help predict variations in
susceptibility between individuals
and at different developmental
life stages, which commonly occur
but cannot be properly addressed
by conventional animal testing.
Functional PBPK models have
developed alongside the rapid
advances made in the use of
human in vitro systems and in
understanding gene function. It
is now possible to predict ADME
outcomes in ‘virtual humans’ with
increasing confidence”.184
In short, PBPK models are translational
tools that can be used to link in-
vitro and in-silico toxicity estimates
to conditions in a living organism.
Ready to use PBPK soware tools
are available.185 PBPK models are
also more generically referred to as
PBK (physiologically based kinetic)
 Pharmacokinetics refers to the movement
of drugs into, through and out of the
body (drug absorption, distribution,
metabolism and elimination).
 European Commission EU Science Hub,
 Langley, 2012, p. 24
 Cronin & Yoon, 2018
 for example, Paini, et al., 2019
 European Chemicals Agency, 2008
 Slikker, et al., 2018
 Myatt, et al., 2018, p. 6
Expert systems
Expert systems are a varied group
of models covering a combinations
of SARs, QSARs and databases.187
They derive toxicity predictions and
estimates from a range of in-silico
models. With access to large chemical
and toxicological databases, increasing
computational power, statistical
algorithms for structure-activity
modelling and powerful datamining
tools, this area of research and testing
has seen much progress in recent years
and has become increasingly relevant
and important for risk assessment
required by government regulators.188
In an overview of in-silico toxicology
methods, expert systems were
described as follows:
“Expert rule-based (or expert/
structural alerts). This methodology
uses structural rules or alerts
to make predictions for specific
toxicological effects or mechanisms
of toxicity. These rules are derived
from the literature or from an
analysis of data sets generated by
scientists. Structural alerts are
defined as molecular substructures
that can activate the toxicological
effect or mechanism. … The purpose
of an in silico expert review is
to evaluate the reliability of the
prediction. The outcome of the
review provides information to
include in the assessment of the
toxicological effect or mechanism.
As part of this review, the expert
might agree with, or refute,
individual in silico predictions”.189
42 Humane Research Australia
In-silico tools for the evaluation of toxicity
The REACH registration requirement
has resulted in the ECHA database,
a large database with information
about thousands of chemicals. ECHA
expected 60,000 registrations in 2018.
This information is publicly available,
but it is not presented in a standardised
way that would make it easy to be read
by a computer. A group of researchers
has used natural language pattern
matching to make the information
machine-readable. They found, for
example, that many chemicals have
been repeatedly tested on animals:
“Interestingly, many chemicals
have been tested more than once,
some shockingly often: For example,
one of the often challenged animal
tests is the Draize rabbit eye test,
where for more than 70 years now,
test chemicals are administered
into rabbit eyes. Two chemicals
were tested more than 90 times,
69 chemicals were tested more
than 45 times. … Notably, the 9
most frequently done animal tests
analyzed here consumed 57% of
all animals for toxicological safety
testing in Europe 2011”.190
Aer making the information machine-
readable, they combined the ECHA
database with several other large
databases that are publicly available
and developed algorithms that enable
automatic read-across to model
chemical properties. They named
their innovation read-across structure
activity relationships—RASAR.
With RASAR, the researchers mapped
the relationships between chemical
structures and toxic properties, based
on 74 characteristics (such as whether
a substance will cause eye irritation) to
predict the properties of a substance.
They found that this method can
predict toxic properties of any chemical
substance more accurately than animal
tests. Toxic substances were correctly
predicted in 89% of cases. They
estimated that by using the RASAR data
fusion method, they would have saved
2.8 million animals and US $490 million
testing costs, and would have received
more reliable data.
“It has recently been demonstrated
that machine-learning software
combined with big data can now
be used to create sophisticated
read-across-based tools that
greatly outperform animal studies
in predicting chemical safety,
with an accuracy of 80%–95%,
compared to 50%–70% for the
respective animal tests”.191
In addition to avoiding animal testing
and obtaining superior results, RASAR
can help avoid duplication of testing,
achieves higher throughput, and is
faster than conventional testing. But at
this stage, it can’t be used for complex
human health eects, such as cancer.
At the time of writing, RASAR research
had just been published, and it is not
yet clear whether regulators will accept
this new testing method. Also, other
researchers will carefully examine
the claims made about RASAR.192
Thomas Hartung, one of the team that
developed RASAR, commented that:
“In the future, a chemist could
check RASAR before even
synthesizing their next chemical
to check whether the new structure
will have problems. Or a product
developer can pick alternatives
to toxic substances to use in
their products. This is a powerful
technology, which is only starting
to show all its potential”.193
 Luechtefeld, Marsh, Rowlands, & Hartung,
2018, p. 199
 Archibald, et al., 2018, p. 3
 for example, Alves, et al., 2019
 Hartung, 2018
 OECD, 2018c
 Ford, 2016
 OECD, 2018c
 Luechtefeld, Rowlands, et al., 2018, p. 732
 Underwriters Laboratories, no year
The OECD QSAR Toolbox194 is a large,
curated database developed by the
European Chemicals Agency and the
OECD. It is freely available online.195 The
OECD also provides training in the use
of this resource, whose most important
features are described as:
“Identification of relevant structural
characteristics and potential
mechanism or mode of action of a
target chemical.
Identification of other chemicals
that have the same structural
characteristics and/or mechanism or
mode of action.
Use of existing experimental data to
fill the data gap(s)”.196
Building on other in-silico methods,
such as read-across and QSAR, and the
availability of large databases, a group
of scientists added machine-learning
techniques. “A first implementation of
machine learning-based predictions
termed REACHacross achieved
unprecedented sensitivities of >80%
with specificities >70% in predicting
the six most common acute and
topical hazards covering about two
thirds of the chemical universe”.197
However, this new tool has not yet
been independently validated. The
soware tool is, however, available to
help companies to meet their REACH
requirements. As the REACHAcross
website notes,
“it predicts 8 required endpoints
for REACH submissions. Users can
produce Toxicology Assessment
Reports for 6 key human health
endpoints, including skin
sensitization, acute dermal
irritation, acute eye irritation, acute
oral toxicity, acute dermal toxicity,
mutagenicity and 2 key ecotoxicity
endpoints of acute aquatic toxicity
and chronic aquatic toxicity”.198
43Better ways to do research
Toxtree is a free in-silico tool “that places chemicals into
categories and predicts various kinds of toxic eects by
applying decision tree approaches. The soware is made
freely available as a service to scientific researchers and
anyone with an interest in the application of computer-
based estimation methods in the assessment of chemical
toxicity”.199 The Joint Research Centre of the EU Commission
commissioned the development of the soware, and
various contributors have since collaborated in its further
Toxtree tool
Source: European Commission EU Science Hub200
 European Commission EU Science Hub,
 European Commission EU Science Hub,
 European Commission EU Science Hub,
 European Commission EU Science Hub,
 United States Environmental Protection
Agency, 2018a
“Toxmatch is an open-source soware application that
encodes several chemical similarity indices to facilitate the
grouping of chemicals into categories and read-across.
The core functionalities include the ability to compare
datasets based on various structural and descriptor-based
similarity indices as well as the means to calculate pair wise
similarity between compounds or aggregated similarity of a
compound to a set.
The soware is made freely available as a service to scientific
researchers and anyone with an interest in the application of
computer-based estimation methods in the assessment of
chemical toxicity”.201
Toxmatch tool
Source: European Commission EU Science Hub202
US EPA’ s GenRA tool helps predict whether a chemical is toxic
Current read-across relies on a subjective assessment of information about one drug to make predictions about another drug
that is similar. Researchers at the US EPA have developed an automated read-across tool called Generalized Read-Across
(GenRA) that aims to encode many expert assessments to make read-across a more systematic and data-based method of
making predictions about the toxicity of specific drugs:
“In its current form, GenRA lets users find analogues, or chemicals that are similar to their target chemical, based
on chemical structural similarity. The user can then select which analogues they want to carry forward into the
GenRA prediction by exploring the consistency and concordance of the underlying experimental data for those
analogues. Next, the tool predicts toxicity effects of specific repeated dose studies. Then, a plot with these outcomes
is generated based on a similarity-weighted activity of the analogue chemicals the user selected. Finally, the user is
presented with a data matrix view showing whether a chemical is predicted to be toxic (yes or no) for a chosen set of
toxicity endpoints, with a quantitative measure of uncertainty”.203
This read-across resource is freely available on the EPA’s website.
44 Humane Research Australia
Other in-silico approaches for the evaluation of toxicity
An adverse Outcome Pathway (AOP)
is a conceptual framework for risk
assessment. It provides a biological
explanation for a toxic event,204 and has
been described in the following way:
“An AOP is a sequence of events
that starts by a chemical effect at
the molecular level (termed a
Molecular Initiating Event) and
progresses through changes
(termed Key Events) in cells,
tissues, and organs to produce an
adverse effect in the body. AOPs
act as a bridge between emerging
methods of safety testing and,
ultimately, what happens in the
body in response to a particular
AOPs are designed to enable targeted,
fast, low-cost and tailored assessments.
They can be continually improved,
as new information is added to the
pathways. “The eventual goal is to
create a network, or web, of pathways
that is suiciently well described that
the eventual eects of a chemical can
be predicted with a limited spectrum of
molecular information”.206
The OECD, which evaluates
international chemical testing
protocols, launched a program on the
development and review of AOPs in
2002. The new AOPs are available on
the AOP Wiki207, which allows scientists
anywhere in the world to share,
develop and discuss their knowledge
of AOPs. Taking the comments on
the AOP Wiki into account, the OECD
publishes the endorsed AOPs on its
website. These published AOPs can be
updated on the Wiki.208 Eectopedia209,
an online encyclopaedia of AOPs, is
another example for the collaborative
development and review of AOPs. The
OECD’s Adverse Outcome Pathway
Knowledge Base (AOP-KB) brings
the AOP Wiki, Eectopedia and two
other AOP platforms together.210 The
following quote comments on the
importance of AOP development:
“The safety evaluation of
environmental and industrial
chemicals is currently propagating
the concept of adverse outcome
pathways (AOPs). These
developments have repercussions
with drug development and safety
assessments of drugs. Under
the umbrella of Organisation
for Economic Co-operation and
Development (OECD) and closely
linked to their chemical safety
testing guideline program, AOPs
are organizing in a crowdsourcing
movement the mechanistic
knowledge on how chemicals
impact on human health and
the environment. This happens,
e.g. in the AOP-Wiki and for
more quantitative AOPs in the
Effectopedia platform. This is
organizing existing knowledge on
shared molecular initiating events
and the subsequent key events
along the chain from cellular
to tissue, organ, organism, and
population effects”.211
Integrated Approaches to Testing and
Assessment (IATA) are approaches
for making decisions about the
toxicity of substances that are based
on multiple information sources,
such as physicochemical properties,
non-testing methods such as QSAR
models and read-across, and testing
methods such as in-vitro and in-vivo.
The identification of an AOP is a
key component of this approach.212
IATA approaches are mainly used for
regulatory purposes. IATA work and are
used in the following ways:
“Integrated Approaches to Testing
and Assessment (IATA) are a flexible
tool for chemical safety assessment,
based on the integration and
translation of data derived from
multiple methods and sources. In
addition to traditional in vitro and
in vivo tests, IATA are increasingly
incorporating new approach
methods, such as high-throughput
screening and high-content imaging
methods, along with computational
approaches that are used as a means
of data generation, interpretation
and integration”.213
“Models are no longer applied in
isolation to determine chemical
safety; there has been a growing
global trend towards the
development and use of multiple
strands of information within
Integrated Approaches to Testing
and Assessment (IATA) for safety
assessment. Nearly all IATA involve
the use of existing data, (Q)SAR
predictions and/or read-across that
are amenable to being integrated
into computational workflows”.214
As this approach includes subjective
expert judgement (weight-of-evidence),
it is not easy to standardise across
industry sectors and countries.
However, IATA frameworks for
skin irritation, skin corrosion, eye
irritation, serious eye damage and
skin sensitisation have already been
adopted internationally.215
 Taylor, 2019
 Ram, 2019, p. 367
 Marshall & Willett, 2018, p. 1956
 AOP-Wiki team, no year
 OECD, 2018a
 OECD, 2016
 OECD, 2019
 Hartung, 2017b, p. 1
 Clippinger, et al., 2018
 Worth & Blaauboer, 2018, p. 301
 Cronin, Madden, Yang, & Worth, 2019, p. 40
 Casati, 2018; Zuang, et al., 2017
45Better ways to do research
Computer modelling of health and disease
Computer models are also used
to simulate (virtual) organs or the
human body, and to explore various
aspects of diseases. Computer models
can link many processes together,
something which is not possible to
achieve with animal models. For
example, atherosclerosis is a common
cardiovascular disease that is caused
by a combination of factors and
can be studied with the help of
computer modelling:
“As a disease that depends on
multiple factors operating on
different length scales, the
natural framework to apply to
atherosclerosis is mathematical
and computational modelling. A
computational model provides
an integrated description of the
disease and serves as an in silico
experimental system from which
we can learn about the disease and
develop therapeutic hypotheses”.216
Computer modelling can integrate
electrical and mechanical processes
into electromechanical models.
These models are useful to study, for
example, implanted cardiac devices
such as pacemakers.217 Modelling of the
respiratory system is another area of
study.218 Such models can be created
for individual patients. Computer
modelling makes an important
contribution to the discovery of new
knowledge. Below, a team of experts
in anaesthesiology219 comment on this
contribution from their perspective:
… complex in silico models have
been applied to pathophysiological
problems to provide information
which cannot be obtained
practically or ethically by traditional
clinical research methods. These
experiments have led to the
development of significant insights
in subject matters ranging from
pure physiology to congenital heart
surgery, obstetric anaesthesia
airway management, mechanical
ventilation and cardiopulmonary
bypass/ventricular support
Other applications of computer
modelling include virtual reality and
surgical simulation. These will be
covered in the section on simulators.
 Parton, McGilligan, O’Kane, Baldrick, &
Watterson, 2016, p. 562
 Pluijmert, et al., 2015
 Clark, Kumar, & Burrowes, 2017
 A medical speciality that encompasses
anaesthesia, intensive care medicine, critical
emergency medicine, and pain medicine
 Colquitt, Colquhoun, & Thiele, 2011, p.499
 Slesnick, 2017, p. 1160
 Slesnick, 2017, p. 1168
Computer modelling in planning and performing
surgery in children and young adults with congenital
heart disease
Congenital heart disease (CHD) is the most common form of birth defect in the
US and Canada. Nearly all forms of CHD lead to long-term complications, and the
cardiovascular autonomy of patients varies. Hence, there is great interest and
potential in the application of biomedical engineering (BME)-based modelling
and simulations to predict the outcomes of surgical and other interventions.
For example, advanced imaging and 3D printing technology allows the surgeon
to print a 3D model of an individual patient’s heart and then perform surgery
on the printed model. Computational fluid dynamics (CFD) involves numerical
analysis to solve problems that involve fluid flows, such as blood flow.
“Application of advanced CFD analyses on a patient’s preoperative anatomy and
on the proposed postoperative solution allows a higher level of confidence that
the intended approach will yield the desired result”.221 A medical doctor reported
his observations regarding the opportunities of computer modelling for his work
as follows:
“Despite the intensity of the work involved, surgical planning with
computational modelling offers unique insights and tremendous
potential for the care of patients with CHD. Several centres have dedicated
laboratories working in this field, and a few multicentre collaboratives
have formed. Congenital cardiology is moving into an era of ‘personalized
medicine,’ and surgical planning with computational modelling and
CFD offers hope that each patient’s interventional procedure can be
planned and analyzed preoperatively on the basis of their patient-specific
characteristics to ensure optimal long-term outcomes”.222
46 Humane Research Australia
Computational psychiatry
“Translating advances in neuroscience into benefits for patients with
mental illness presents enormous challenges because it involves both
the most complex organ, the brain, and its interaction with a similarly
complex environment. Dealing with such complexities demands powerful
techniques. Computational psychiatry combines multiple levels and
types of computation with multiple types of data in an effort to improve
understanding, prediction and treatment of mental illness”.223
Computational psychiatry, similar to other computational predictive approaches,