Explanatory hierarchy of causal structures in molecular biology

Abstract

In the debate on causal explanation in biology, in the past two decades largely influenced by the new mechanist (NM) approach, the concept of a pathway has recently reemerged as a promising research agenda (see, in particular, Ross Philosophy of Science, 85(4), 551-572, 2018; The British Journal for the Philosophy of Science, 72(1), 131-158, 2021). Ross’ account of biological explanation differentiates several autonomous types of causal structures that play explanatory and other roles across the life sciences. NM, however, prioritizes mechanisms as vehicles of biological explanations. According to this program, the causal architecture of biological pathways and other causal structures, such as cascades and processes, can be interpreted with at least one of the NM’s mechanism concepts. In other words, these alternative causal structures are not sufficiently distinctive to merit the explanatory autonomy with regard to the NM corresponding concepts. We examine the explanatory practice of molecular biology and concur with Ross that there are indeed distinct types of causal structures, not all falling under the concept of a mechanism. Nonetheless, we show that the concept of mechanism is referring to a privileged causal structure, at the center of explanatory efforts in molecular biology. Pathways and other causal concepts, while somewhat distinct from mechanisms themselves, are explanatorily relevant to the degree in which they exhibit mechanistic features, are parts of a mechanistic architecture, or may lead to a mechanistic arrangement. What emerges in that manner is a hierarchy of causal structures with mechanisms at the explanatory top, and lower levels differing in the degree in which they contribute to mechanistic arrangements.

Full text

Vol.:(0123456789)
European Journal for Philosophy of Science (2021) 11:60
https://doi.org/10.1007/s13194-021-00380-7
1 3
PAPER INTHEPHILOSOPHY OFTHELIFE SCIENCES
Explanatory hierarchy ofcausal structures inmolecular
biology
ZdenkaBrzović1· VitoBalorda1· PredragŠustar1
Received: 30 June 2020 / Accepted: 13 May 2021
© Springer Nature B.V. 2021
Abstract
In the debate on causal explanation in biology, in the past two decades largely influ-
enced by the new mechanist (NM) approach, the concept of a pathway has recently
reemerged as a promising research agenda (see, in particular, Ross Philosophy of
Science, 85(4), 551-572, 2018; The British Journal for the Philosophy of Science,
72(1), 131-158, 2021). Ross’ account of biological explanation differentiates several
autonomous types of causal structures that play explanatory and other roles across
the life sciences. NM, however, prioritizes mechanisms as vehicles of biological
explanations. According to this program, the causal architecture of biological path-
ways and other causal structures, such as cascades and processes, can be interpreted
with at least one of the NM’s mechanism concepts. In other words, these alternative
causal structures are not sufficiently distinctive to merit the explanatory autonomy
with regard to the NM corresponding concepts. We examine the explanatory prac-
tice of molecular biology and concur with Ross that there are indeed distinct types
of causal structures, not all falling under the concept of a mechanism. Nonetheless,
we show that the concept of mechanism is referring to a privileged causal structure,
at the center of explanatory efforts in molecular biology. Pathways and other causal
concepts, while somewhat distinct from mechanisms themselves, are explanato-
rily relevant to the degree in which they exhibit mechanistic features, are parts of a
mechanistic architecture, or may lead to a mechanistic arrangement. What emerges
in that manner is a hierarchy of causal structures with mechanisms at the explana-
tory top, and lower levels differing in the degree in which they contribute to mecha-
nistic arrangements.
Keywords Causal structures· Causal concepts· Biological explanation· Pathways·
Mechanisms· Genes· Transcriptional noise
This article belongs to the Topical Collection: EPSA2019: Selected papers from the biennial
conference in Geneva
Guest Editors: Anouk Barberousse, Richard Dawid, Marcel Weber
* Predrag Šustar
psustar@uniri.hr
Extended author information available on the last page of the article

European Journal for Philosophy of Science (2021) 11:60
1 3
60 Page 2 of 21
1 Introduction
Recently, Ross (2018; 2021) has reinvigorated the debate on causal explanation in
biology by stealing the spotlight from the concept of a mechanism and introducing
the concept of a causal pathway. This concept is, in her view, associated with distinct
strategies of causal investigation and consequently figures in explanations different
from mechanistic ones. Thus, according to Ross (2018: 552), the pathway concept
“refers to a group of causal factors, which are ordered in a sequence that leads to
some final outcome of interest” (e.g., gene expression pathways, cell-signaling path-
ways, metabolic pathways, ecological pathways, developmental pathways (Ross,
2021)). Think, for instance, about the glycolytic pathway in biochemistry, through
which the glucose sugar is broken down in a stepwise manner in acquiring energy.
The mechanistic explanation, as expounded by the new mechanist (NM)
approach, invokes the concept of a mechanism as paramount for explaining a diverse
range of biological phenomena. A minimal or consensus characterization of the
mechanism concept goes as follows: “A mechanism for a phenomenon consists
of entities and activities organized in such a way that they are responsible for the
phenomenon” (Illari & Williamson, 2012: 120). According to NM, pathways and
other causal structures can be captured with the mechanism concept. In this paper,
we address the question whether the factual diversity of causal structures in biol-
ogy calls for “importantly different types of explanation” (Ross, 2021: 131). We
examine Ross’ characterizations of the pathway concept and its supposed clear-cut
differences from the NM mechanism concepts. We apply, then, Ross’ strategy of
investigating different kinds of explanatory relevant causal structures to the current
scientific practice of molecular biology, specifically, to the de novo theory of gene
origin and functionality. We agree with Ross that there are indeed different kinds of
causal structures that are referred to in the extant explanatory models in the research
domain in question, and which cannot be subsumed under the single mechanism
concept. The question is, however, whether they play a role important enough to
ground a distinct type of explanation. We claim that this is not the case. Rather, vari-
ous types of causal concepts in biology play important explanatory and other roles
in as much as they describe causal structures exhibiting mechanistic features or have
the potential to lead up to a mechanistic arrangement.
In addition, we examine whether unorganized and unstructured causal processes
can play important explanatory roles, which would be a firm argument against the
mechanistic explanatory hegemony. Such cases occur (i) when there is an occasional
‘noise’ in the functioning of a mechanism giving rise to causal processes, which,
then, may lead to new mechanistic causal structures; and (ii) in the instances of a
first emergence of mechanistic architecture from non-structured causal processes.
Both types of instances, however, draw their explanatory importance from the link
to mechanisms, as we illustrate through the de novo explanatory models. Thus, it
follows that the mechanism concepts are still at the center of explanatory efforts
in molecular biology. Nevertheless, what emerges is a nuanced hierarchy of causal
structures with mechanisms at the explanatory top, with lower levels occupied by
causal structures exhibiting mechanistic features to various degrees.

1 3
European Journal for Philosophy of Science (2021) 11:60 Page 3 of 21 60
The present paper is structured as follows: in Section2, we assess Ross’ account,
in particular, as far as its list of differentiating features of the pathway concept is
concerned and, consequently, an allegedly clear-cut divide between this and the NM
explanatory patterns. In Section 3, the case study is described, which instantiates
our account of causal explanation in molecular biology. Finally, in Section4, we
delineate our account of an explanatory hierarchy of causal concepts, topped by the
mechanism concepts. We show how this hierarchy emerges from the factual diver-
sity of causal structures in biology.
2 Dierentiating causal structures andexplanatory patterns
inbiology
In her recent paper, Ross (2021) addresses a set of related traditional issues in the
philosophy of biology. The foundational issue, as claimed by the account here in
question, is concerned with the factual diversity of causal structures and, corre-
spondingly, causal concepts in biology, namely, their characterizing features. The
issue of this differentiation, then, leads to a set of further issues: one of causal inves-
tigative strategies applied in the contemporary biological sciences, the issue of bio-
logical explanatory patterns and, finally, the issue of typology of scientific reason-
ing in this domain (see Ross, 2021: 151–152). The resulting set of Ross’ responses
confronts itself with the NM program. This is the case not only with, as Ross con-
clusively emphasizes, the “nature and limits of mechanistic explanation” (Ross,
2021: 154), but also with other issues on which NM has been focusing.
Now, what is specific about Ross’ assessment of the NM general program con-
sists primarily in identifying the characterizing features of the pathway concept as
opposed to the NM mechanism concept. More derivatively, it consists in the analysis
of the ensuing explanatory patterns of pathway-based and mechanism-based causal
explanations. The corresponding assessment of NM, thus, emerges from these anal-
yses. Accordingly, we first examine the features in question and, then, the uses of
the pathway concept in explanations in the life sciences. Close inspection of Ross’
analyses reveals, in our view, that her account obtains less than expected. Namely,
the pathway concept emerges as a specific part of mechanistic explanatory scheme,
and not as its alternative.
2.1 The pathway andmechanism concepts
The pathway concept has already been examined to a considerable extent in Thagard
(2003), Schaffner (2008), and Ross (2018).1 Ross’ most recent account (2021), nev-
ertheless, further expands the analysis of this and other important causal concepts
in biology, having at the same time NM as the received view. Now, according to the
1 Differently than Ross (2021), which is specifically targeting the foundational and derivative issues con-
cerned with the pathway concept, Ross (2018) examines the causal concept in question as related to the
problem of causal selection. Despite that different focus, Ross’ earlier paper contains some important
insights on the issue of features characterizing the causal structure of metabolic pathways.

European Journal for Philosophy of Science (2021) 11:60
1 3
60 Page 4 of 21
account here under consideration, the pathway concept refers to a group of causal
factors ordered in a specific sequence leading to an outcome of scientific interest. As
a group of causal factors, take, for instance, metabolites, enzymes, enzyme cofac-
tors, and accessory substrates in a metabolic pathway. In other words, Ross, thus,
generally claims that “the notion of a pathway refers to a sequence of causal steps
that string together an upstream cause to a set of causal intermediates to some down-
stream outcome” (Ross, 2021: 136–137). Apart from the standard examples from
molecular genetics, cell biology and, especially, biochemistry, the account broadens
the application of the characterization by highlighting the explicit scientific use of
the pathway concept in developmental biology and, even, ecology.
Consider the causal structure of glycolysis, the biological process by which glu-
cose is broken down to extract energy via its conversion into two molecules of pyru-
vate. As known, this biochemical degradation provides energy in the form of adeno-
sine triphosphate (ATP) to the containing biological system. Thus, when “scientists
explain and describe glycolysis they rely on the pathway concept – they divide this
process up into 10 sequential steps that are represented along a causal chain, called
the glycolytic pathway” (Ross, 2018: 554).2
By contrast, NM has been claiming in the past two decades that biological phe-
nomena are predominantly explained via the mechanism concept. In short, biologists
search for mechanisms, because the causal structures of a mechanism are crucially
relevant for explanation and, moreover, for prediction and control or intervention
of some desired kind. The actual scientific practice of the life sciences proceeds
by tracing these causal structures, which produce, underlie, or maintain phenom-
ena of interest (see Craver & Darden, 2013: 15).3 Consider the following example
from molecular biology that illustrates the NM basic causal concept. In the process
of replication, the double helix of DNA (which here represents an entity) unwinds
(activity) and distinct new component parts (again, entities) bond (again, activity)
to both parts of the unwound DNA helix (see Craver & Darden, 2013: 17). In other
words, the causal structure of the DNA replication mechanism contains component
parts that actively produce the explanandum-phenomenon.
In what follows, we examine Ross’ characterization of the main features pertain-
ing to the general NM mechanism concept, which are, in turn, important for her
2 Ross, at this and other similar points, insists on quite a strong analogy of the pathway causal structures
of this kind and manufacturing contexts in ordinary life situations, such as the structure of an assembly
line production. In our view, the characteristics and limits of this analogy require a separate analysis,
which goes beyond the scope of the present paper.
3 The most frequently referred NM characterizations of the mechanism concept, at least, the seminal
ones, are as follows: (1) “Mechanisms are entities and activities organized such that they are productive
of regular changes from start or set-up to finish or termination conditions” (Machamer etal., 2000: 3);
(2) “A mechanism for a phenomenon of behavior is: a complex system that produces that behavior by
the interaction of a number of parts, where the interactions between parts can be characterized by direct,
invariant, change-relating generalizations” (Glennan, 2002: S344); and (3) “A mechanism is a structure
performing a function in virtue of its component parts, component operations, and their organization.
The orchestrated functioning of the mechanism is responsible for one or more phenomena” (Bechtel &
Abrahamsen, 2005: 423). We are here putting them on board in order to confront them to Ross’ charac-
terizations of the main causal concepts in biology. Apart from that, we will be using (1)-(3) and some
more recent versions both in moving objections to Ross’ account, and in arguing for our view of causal
explanation in molecular biology.

1 3
European Journal for Philosophy of Science (2021) 11:60 Page 5 of 21 60
central objective of a sharp distinction between pathways and mechanisms, plus
their respective explanatory roles. We show that Ross’ corresponding list of features
cannot be taken as a comprehensive characterization of the NM program.
2.2 Three features oftheNM mechanism concept
Ross (2021: 134–136) highlights three features of the NM mechanism concept. A
first feature regards the constitutive nature of mechanisms as causal structures, i.e.,
“involving particular systems with higher-level behaviors that can be decomposed
into lower-level causal parts” (Ross, 2021: 134). In other words, scientists proceed
by identifying mechanism’s parts, their location and their interactions that produce
or are related in a certain way to the behavior of the explanandum phenomenon.
However, NM acknowledges that mechanisms can also refer to a sequence of causal
steps occurring at the same level. Such mechanisms are invoked in etiological mech-
anistic explanations (see Salmon, 1984; Craver, 2007; Craver & Tabery, 2019). For
instance, NM emphasizes the importance of a temporal organization in the mech-
anisms. Order, rate, and duration of successive component-parts’ activities are all
crucial features of a mechanism (see Craver, 2001: 60). Furthermore, the etiologi-
cal character of mechanisms is acknowledged by pointing out that they display “a
sequence of stages from beginning to end, and it would not be possible to change
their order without gumming up the works (or making it a different mechanism
entirely)” (Craver, 2001: 61). Finally, think of the NM’s description of mechanism
as producing, underlying, or maintaining the phenomenon (see Craver & Darden,
2013). The idea of production best applies to mechanisms understood as causal
sequences terminating in a determined end-product (see Craver & Tabery, 2019).
Perhaps there is more work to be done from the side of NM to further specify
the relationship between the etiological and constitutive aspects of mechanisms.
Consider, for instance, the mechanism of protein synthesis in its most simpli-
fied version. DNA molecules being transcribed into mRNA molecules, which are,
then, translated into the corresponding sequence of amino acids making up the
resulting protein. The causal chain ‘DNA➔RNA➔amino acid sequence’ can be
taken as an etiological causal chain. The end-product of that chain, in turn, con-
stitutes a higher-level biological phenomenon, i.e., a certain phenotypic effect.
Actin proteins, for instance, form microfilaments in the cell, which constitute a
part of the cytoskeleton. That is, at the first next level of biological organization,
the phenomenon in question is an adequately folded protein’s structure, which is
critical for its functional contribution to a containing system.4 Moreover, we can
descend to lower levels and examine causal processes underlying transcription
and translation by focusing on specific molecular interactions, chemical bonds,
etc. Thus, in our view, typically, mechanisms exhibit both their constitutive and
etiological dimensions.5
4 Going up to higher levels, we arrive at the protein’s causal contribution to more palpable higher endo-
phenotypic and phenotypic effects.
5 In the next section, we see in more detail how both characteristics interact in providing causal explana-
tions in molecular biology.

European Journal for Philosophy of Science (2021) 11:60
1 3
60 Page 6 of 21
A second feature, according to Ross’ account, relates to the fact that mechanisms
are causal systems containing a significant amount of detail, therefore, lacking
the feature of abstraction. In other words, to explain a certain phenomenon, biolo-
gists need complete descriptions of mechanisms, which are thereby overloaded
with information (see Ross, 2021: 134–135). However, some NM proponents have
emphasized the importance of abstraction. Ross (2021: 143) is addressing this issue
by granting that “some mechanistic philosophers subscribe to this ‘abstract mecha-
nism view’”, but she believes that the most part of NM understands “mechanisms as
highly detailed”.
We see no reason for the view that mechanism models necessarily need to be
highly detailed. Without entering specifics of this particular debate, we will assume
here that the amount of details required by a mechanistic model depends on the
explanandum phenomenon. To illustrate that point, think again of the mechanism
of DNA replication. As well known, DNA is basically structured as a double helix
of complementary strands, which are unwound during the replication process. Each
strand of a DNA molecule serves as a template for its newly synthesized counter-
part. The major explanatory contribution of the mechanism model in question is
concerned with the tracking of how nucleotides forming the backbone of a double
helix are matched between the strands through hydrogen bonds in the corresponding
base pairings. In the context of molecular biology, it is sufficiently relevant the fact
that hydrogen bonds are relatively weak (as compared to covalent or ionic bonds),
which, in turn, represents a crucial explainer of the DNA unwinding in the overall
replication process.
Ross’ third feature of the NM mechanism concept, important for differentiat-
ing it from the pathway concept, is concerned with a predominant emphasis put by
the NM program on the notions of “force”, “action”, and “motion” in qualifying
mechanistic causal relationships. Ross proceeds to explicate this feature by using
an analogy with machines in ordinary life that have parts, such as levers and ham-
mers, which actively do things and, thus, use “force”. However, by singling out this
feature, Ross’ account neglects other important aspect of mechanisms, namely, as
we have seen in the DNA replication example, the aspect of entities’ more complex
activities performed in a determined manner within some containing living system.
Think at this point of an enzyme’s causal role. It involves something different than
implied by the abovementioned notions. That is, by qualifying in that way causal
relationships, one is losing from sight an entity’s function. For example, the motion
itself of DNA polymerase is not important, but its determined function or, even, its
malfunctioning within a system. On our reading of NM, the mechanism concepts
involve a functional sense along this line (see, e.g., Garson, 2013), not captured by
the feature in question.
From examining the above features, we have seen that Ross singles out what can
be taken, at best, as one specific understanding of the mechanism concept in the
overall NM program. Let us examine now the final part of Ross’ response to the
foundational issue, namely, the features that are distinctive of the pathway concept.
Here, inevitably, there will be some redundancy with the above description of the
mechanistic features, since one of the main points in Ross’ account is to distinguish
pathways from mechanisms (and, in turn, from other causal structures). Nonetheless,

1 3
European Journal for Philosophy of Science (2021) 11:60 Page 7 of 21 60
we follow Ross’ line of argument, but minimize the talk of the features that have
already been covered.
2.3 The causal independence ofbiological pathways
Ross’ most recent list of features characterizing pathways and our corresponding
assessment are as follows:6 first, causal pathway consists of steps (recall the glycoly-
sis example) showing a fixed order of causal relationships. In addition, that sequential
order of steps reflects which outcomes are expected to occur before and after each
step, leading to a final stage of the overall process. We will not further discuss the first
pathway feature in comparison to mechanisms, because, as Ross (2021: 140) acknowl-
edges: “this first feature does not capture a true difference between the pathway and
mechanism concepts because both can be understood in terms of causal steps”.
Let us, then, focus on a second feature, which is supposed to highlight the speci-
ficity of the pathway concept. It relates to a sense of “flow” (Ross, 2021: 139–142)
that we get from the pathway research, most interestingly, in biology. That flow may
relate to an entity or a signal passing through a certain system. For instance, if we
recall the glycolysis example, the metabolic pathway in question traces the flow of
a chemical substance via intermediate stepwise changes. Ross believes that this fea-
ture is to a highest degree bound to pathways, that is, as compared to other features
on the list and, especially, as opposed to the NM mechanism concepts. Ross, then,
at the end of this argumentative line, concludes that “something more is present in
these pathway cases that is not found in all causal relationships” (Ross, 2021: 142).
However, there are some difficulties with such a characterization of the flow fea-
ture: (i) it is, in our view, vulnerable to the same kind of difficulties as the notion
of genetic (biological) information. Namely, the fact that biologists usually qualify
the ‘flow’ of (genetic) information as involving ‘something more’ than mere causal
relationships between the main groups of biological macromolecules calls for a more
precise explication of what this locution amounts to (see Okasha, 2019: 97); (ii)
apparently, a similar feature can be found in NM. In brief, the notion of a flow spe-
cifically characterizing biological pathways seems to be closely intertwined with the
Machamer etal. (2000) notion of “productive continuity” in mechanisms. That is, to
represent how a mechanism works, biologists need to show a determined grouping of
stages, i.e., how an earlier stage leads to one or more subsequent one(s). Each of those
stages make a certain difference to what happens at the next stage(s) (see Craver &
Darden, 2013: 19). Characterized in that manner, the notion of “productive continu-
ity” also implies a sense of “flow”. However, according to Ross’ account, there is an
important difference between the two; the pathway flow “involves the permanence
or continuity of something that travels along causal connections” (Ross, 2021: 141).
6 It is worth noting here that there is a previous list of features with regard to the causal structure of
pathways, in particular in biochemistry, and referring causal concept. This Ross’ list puts forward the
following features: (i) causal control, (ii) material continuity, and (iii) fixed order. Although the list only
partially overlaps with the current one, both in number and respective characterization, we are not exam-
ining in this setting the details of their difference. The main reason for that is contextual; since the cur-
rent list of features is closely related to the issue of causal explanation in biology, whereas the preceding
list is targeting the causal selection problem (see Ross, 2018).

European Journal for Philosophy of Science (2021) 11:60
1 3
60 Page 8 of 21
Now, apart from bouncing back to difficulty (i), it is hard to see how the permanence
of that ‘something more’ besides mere causation might keep its identity.7
Consider again the metabolic pathway of glycolysis. Even if we grant the existence
of something besides causal influence, it is not clear how to individuate or establish the
permanence of any relevant biochemical entity or signal that would run through stepwise
changes of the glycolytic pathway by keeping its identity. Ross is singling out in that regard
metabolites. The question is: how can they be considered as being identical in the changes
they are going through in the overall process? Moreover, it is difficult to discern how that
point would be different from the role played by the notion of “productive continuity” con-
cerned with causal relationships. In sum, in our view, the two confronted notions indeed
overlap in some respect, whereas the "flow" feature has a more demanding burden of proof
in gaining its conceptual independence from causation than its NM counterpart.
According to a third feature, a modelled causal pathway structure abstracts from
a significant number of details. This can be seen in two ways: (1) for example, meta-
bolic pathways in biochemistry exclusively represent the “flow” itself that commonly
includes metabolites, thus, omitting other factors, such as enzymes, coenzymes,
accessory substrate, temperature, pH etc., and (2) pathways represent complex pro-
cesses by singling out a reduced number of causal steps. However, as we have already
seen, the NM overall approach has worked out different accounts in this regard. We
have proposed that the level of abstraction of the mechanistic explanatory models
depends on the corresponding explanandum. We take it that the same applies to path-
ways. Omitting certain factors is a regular procedure in causal explanations. In Ross’
example, temperature and pH would typically be taken as background conditions.
Take glycolysis, for instance, as a paradigmatic pathway. The explanation of glycoly-
sis in many cells and tissues is more complicated than usually depicted in a textbook
format. Moreover, the pathway in question can be divided in three distinct phases: (i)
the energy investment phase or priming phase; (ii) the splitting phase; and (iii) the
energy-generation phase (see Akram, 2013). Scientists are also targeting the glyco-
lytic pathway when designing a cancer treatment strategy (see Gill etal., 2016: 89).
Thus, depending on a different explanandum phenomenon, that is, the explanandum
phenomenon being a certain cancer cell-line, or a glycolysis phase, or an enzyme
activation, the level of abstraction will accordingly differ.8
8 The abstraction feature of pathways plays an important role in Ross’ assessment of why pathways are
explanatory. Namely, she argues that there are some explananda for which the pathway information is
explanatory, while the mechanistic one is not. According to Ross, if the pathway relations are fixed,
lower-level mechanistic information can vary, i.e., lower-level causal information is not explanatory. Note
that here we are not just concerned with the issue of abstraction/details that might describe a system at the
same level, but lower levels are explicitly invoked. In other words, Ross puts forward an antireductionist
argument invoking multiple realizability against the allegedly reductionist mechanist program. The NM
program, however, is hardly reductionist in the standard meaning of the term. In her assessment of the
overall NM commitments, Franklin-Hall (2016), for instance, identifies (1) somewhat reductive tenden-
cies, i.e., insistence that phenomena are explained in lower-level terms, usually at one level below; and (2)
non-reductive tendencies, i.e., resistance to the view that every phenomenon is ultimately explained at the
physical level. The fact that mechanisms deployed by etiological explanations depict a sequence of causal
steps occurring at the same level points to the conclusion that mechanistic explanatory strategies are not
necessarily reductive. Thus, the explanatory non-reductiveness of pathways cannot serve as a ground for
establishing a clear distinction between pathways and mechanisms.
7 Special thanks to anonymous reviewers for pressing this point in our assessment of Ross’ account.

1 3
European Journal for Philosophy of Science (2021) 11:60 Page 9 of 21 60
A fourth differentiating feature puts an emphasis on the “connection” character-
istically displayed by biochemical pathways, rather than on the notions of “force”,
“action” and “motion”, as mostly in NM. According to this feature, the goal of the
pathway concept “is to show what is causally connected to what, as opposed to the
fine-grained details of ‘how’ they are connected” (Ross, 2021: 144). In our view,
the fourth feature draws on all the preceding features on the list. Nevertheless, what
might be innovative in the feature under consideration is related to the use of the
pathway concept, more than to the foundational issue of the pathway concept itself.
In other words, the feature of “connection” implies a specific causal investigative
strategy that is focused on delineating “causal routes”, independently of an outcome
that is, according to this strategy, always arbitrarily circumscribed. Before turning
to the uses of the causal concepts, a final remark on this feature, which, in our view,
shortens the distance between the feature of “connection” and the NM notion of
“productive continuity”.
Consider, for instance, a linear mechanism.9 There are different stages in that type
of mechanisms, one stage leading to the next, and there must be a certain connec-
tion between the steps, independently of some final outcome. In the mechanism of
DNA replication, we have both entities (e.g., DNA) and activities (e.g., unwinding,
bonding) that are in a precise “connection” or in a relationship of “productive conti-
nuity”, quite different from the abovementioned group of notions (“force”, “action”,
“motion”). In addition, the mechanism may lead to other causal structures.
2.4 The explanatory uses ofcausal structures
Previously, we have seen that Ross claims that an obvious diversity of causal struc-
tures in biology demands more conceptual differentiations. The underlying assump-
tion being that there are clear-cut boundaries between the new pathway concept
and the NM mechanism concepts in general. However, we have pointed out that the
boundaries under consideration are much fuzzier and that the features in question
interestingly overlap between the two causal concepts.
Ross points to an important feature of mechanisms that differentiates them from
other causal structures. When discussing the example of an explanation of the
spread of a disease, she refers to an interconnected space revealing various potential
pathways. Ross’ account rightly emphasizes that such an explanation invokes causal
structures that depart from mechanisms, characterized by a discrete, isolable nature
and referring to individual causal structures (see Ross, 2021: 150). We do not think,
however, that this distinction is best spelled out by invoking the concept of pathway,
since there are examples of highly structured pathways not allowing space for dif-
ferent potential causal routes.10 As discussed in Section2.3, glycolysis is such an
example. As we will argue in the next section and, more importantly, in Section4,
9 One of the most frequent instantiations of the linear type of mechanism is protein synthesis. By that,
we are here simply pointing out that there are other, non-linear types of a mechanism, such as the Krebs
cycle in biochemistry.
10 Thagard (2003), for instance, argues that biochemical pathway is a kind of mechanism.

European Journal for Philosophy of Science (2021) 11:60
1 3
60 Page 10 of 21
the difference between mechanistic explanations and explanations invoking looser
causal structures than mechanisms is an important one and merits further analysis.
In conclusion, we side with Ross on the issue of the importance of distinguishing
different causal structures relevant to explanations in the current life sciences. None-
theless, we argue that this is not a clear-cut difference, as claimed in Ross’ account.
Rather, we show that causal structures and their respective concepts are explanato-
rily relevant to the degree in which they are related to mechanistic arrangements,
given their causal robustness, as illustrated in the next two sections. Thus, we disa-
gree with Ross that different causal concepts imply “importantly different types of
explanation” (Ross, 2021: 131). We argue, instead, that differences between biologi-
cal causal concepts amount to a difference in the degree of their explanatory power.
In the next two sections, we turn to our account of a mechanistic explanation
and explanations invoking other types of biological causal structures. We start out
by examining the recent explanatory practice in molecular biology, which appeals
to both mechanisms and other causal structures. Namely, the biological theories
dealing with the basic biological phenomenon of new genes synthesis. The corre-
sponding explanatory models illustrate how a paradigmatic example of mechanism,
according to the NM literature, but also scientists themselves, the mechanism of
protein synthesis, when examined in more detail, can suggest explanatorily signifi-
cant new causal routes. One of them is concerned with certain random and unstruc-
tured events arising from regular functioning of the mechanism itself. This ‘noise’
in the mechanism of protein synthesis may lead up to a series of interesting causal
events and, eventually, to new mechanistic arrangements. Accordingly, these scien-
tific explanatory models show how unstructured, loose causal structures may play
an important explanatory role, given their actual and potential relationships to full-
fledged mechanistic arrangements.
3 Theories ofgene origin andfunctionality: thecase ofde novo
genes
We now turn to the current biological theories of new genes synthesis. The cor-
responding explanatory models, on one hand, clearly acknowledge the existence
of different causal structures, but, on the other, suggest a unified explanatory
pattern that exhibits a hierarchy of causal structures with mechanisms positioned
at the primary explanatory level. Nevertheless, the mechanistic causal structures
are both directly and less so related to other types of causal structures, as it is
especially pointed out by the de novo explanatory models, on which we focus
in this section. Although the models of new genes synthesis and their function-
ality include evolutionary considerations in their comprehensive explanatory
efforts, they are mainly based on the mechanism of protein synthesis, its stand-
ard and non-standard functioning in molecular biology, as we further explicate
in Section4.

1 3
European Journal for Philosophy of Science (2021) 11:60 Page 11 of 21 60
3.1 Short introduction intothede novo genes origin theory
We discuss here the theories addressing the following departing ‘how-question’: ‘how
do genes arise and become functional?’ (see Tautz, 2014). The question, in this case,
refers to the ways in which new genes emerge in already existing genomes, rather
than how the very first genes have arisen in the evolutionary history. There are two
main theories that provide an answer to this question. The first one is based on the
process of gene duplication and divergence out of the already existing genes. The
second, more recent one, which we will consider as our main case study, describes de
novo gene origin from non-genic (i.e., non-coding for a protein) regions in a genome.
The inquiry into the origin of new genes is concerned with accounting explanato-
rily for the ways through which the genes that do not have detectable homologues in
other lineages come about (see Tautz etal., 2013). Such DNA segments have been
called in the scientific literature “orphan genes” and it is estimated that they constitute
up to one third of genes in eukaryotic genomes (see Tautz & Domazet-Lošo, 2011).
Let us first see more general features of both theories, i.e., their explanatory models
and their basic relationships (see Fig.1).
According to the duplication-divergence explanatory models, an already existing
gene gets duplicated. The ancestral gene (the arrow on the topleft side in Fig. 1)
continues to perform the gene’s original function, while the gene duplicate is under
reduced selective pressure and can acquire mutations (as represented by the top right
arrow with vertical lines representing mutations). The circles represent transcrip-
tion factor binding sites, and the diagonalarrows represent the progression of the
duplication events. The theory under consideration claims that the gene duplicate
is to a significant extent free to diverge and to undergo quite radical changes in its
sequence, as shown by the right arrowin the third row. Consequently, the similarity
with the founder gene is lost and, thus eventually, it is identified an orphan gene.
Contrary to that theory and its explanatory models, in the recent de novo model
of gene origin, an initially non-coding sequence (represented by the straight line in
Fig.1) can develop into a “proto-gene” (the second row in the figure, the rectangle
with vertical lines representing mutations), that is, a sequence that exhibits stable
expression and/or translation, but lacks a proper function (see Carvunis etal., 2012).
In the next step, in the third row on theright, the corresponding transcript is translated
at higher levels, and the resulting peptide starts providing an adaptive advantage to
an individual organism. Finally, by taking into account evolutionary considerations,
which apply a population-level and transgenerational setting with respect to the pre-
ceding individual one, natural selection acts to fix it in the lineage.11 In what follows,
we explicate in more detail a complete biological explanation of the explanandum-
phenomenon, thereby scrutinizing the explanatory role of different causal structures
involved in the mechanism of protein synthesis and its outcomes.
11 For a considerably long time, the duplication–divergence biological theory was the only available
explanatory answer to the crucial ‘how-question’ of a “mystery of the orphans” (Dujon, 1996). Prior
to the de novo explanatory model sketched above, researchers in this area took it to be highly unlikely
that functional genes could emerge from random, noncoding genomic sequences (see Tautz etal., 2013).
This assumption, however, has been overturned by evidence gathered in different taxa (see, e.g., Schmitz
etal., 2018; Ruiz-Orera etal., 2018).

European Journal for Philosophy of Science (2021) 11:60
1 3
60 Page 12 of 21
3.2 Mapping outcausal processes leading uptode novo genes
Let us examine more closely the progression of events leading up to the production
of a new gene, according to a recent consensus model:
In step A, we have a non-coding genomic region, that is, component of organ-
ism’s DNA that does not encode proteins. This non-coding sequence is transcribed
at low levels. As already emphasized, it is not unusual that some non-coding DNA
gets transcribed into non-coding RNA molecules. That accidental transcription or so-
called ‘transcriptional noise’, as we will see in the next steps of the progression, initi-
ates events that may end up as a fully systemic gene expression. It is, nevertheless,
important to note that the standard functioning itself of the transcriptional mecha-
nism makes that outcome possible, suggesting an interesting explanatory ordering of
causal structures in molecular biology, as we argue in the next section.
In step B of Fig.2, this non-coding intergenic region may acquire a short open
reading frame (ORF), i.e., a continuous stretch of codons (beginning with a start
Fig. 1 Two theories explaining the origin of “orphan genes” (reproduced from Plissonneau etal., 2017:
10, Fig.6)
Fig. 2 Progression of events in de novo gene origin and functionality (reproduced from Frietze & Leath-
erman, 2014: 595, Fig.1; the reproduced figure is adapted from Carvunis etal., 2012)

1 3
European Journal for Philosophy of Science (2021) 11:60 Page 13 of 21 60
codon and ending with a stop codon) that can be translated into a peptide. Even
though long ORFs are best candidates for identifying potential protein-coding
genomic regions, some short ORFs can produce functional peptides, as well. This
is exactly what happens in step C, namely, the sequence gets transcribed and trans-
lated at low levels, which is consistent with the evidence that a certain amount
of fortuitous transcripts get translated into rudimentary proteins (see Frietze &
Leatherman, 2014). The first three steps, thus, refer to somehow random outcomes
of the mechanisms of transcription and translation, the latter producing mostly
short and disordered proteins (see also Schmitz and Bornberg-Bauer, 2017).
In step D, our original sequence with a short, translated ORF increases its expres-
sion through mutations, which leads to a more stable transcription and translation at
much higher levels. Furthermore, that results in the production of more structured
peptides. At this stage of the succession of events, the proto-genes grow and may
become fully functional de novo genes. Transcripts get translated at higher levels
and the resulting peptides start to secure a certain adaptive advantage to an indi-
vidual organism. Natural selection can now contribute to distributing the new genic
sequence across the corresponding population of organisms.
However, more importantly for the mechanism of protein synthesis, transcription
and translation of the previously non-coding intergenic regions are not any more
producing fortuitous outcomes. They are now producing more stable outcomes than
those in the previous steps. Thus, step E marks the transition to a regulated tran-
scription and translation, proceeding at high levels. The newly emerged gene gets
integrated into existing regulatory networks and can eventually found a new gene
family.12 The described consensus explanatory model for de novo gene synthesis
encompasses different causal structures both at the individual organism level and,
to a lesser degree, at the populational level.13 The model also describes unstructured
causal events within the progression leading up to a new functional gene. The reason
for this, as should be clear from the above analysis, is that the sequence of events
involves, most notably, clearly structured mechanistic architecture, interspersed with
accidental, ‘noisy’ events, all arising in the process of standard functioning of the
mechanisms in question.
However, not all such chains of events are equally explanatorily important. Primar-
ily interesting are those more directly answering to the ‘how-question’ concerned with
the occurrence of orphan genes in a genome. Thus, it seems that we can talk about
12 There are two leading explanatory models based on the biological theory in question, which respond
differently to important difficulties of the above approach to the basic biological phenomenon of new
gene production; namely, the “RNA first” and “ORF first” explanatory models (for more details, see
McLysaght & Guerzoni, 2015). In our view, we may set aside in the present paper the specifics of this
issue and its potential influence on the ordering of different causal structures within molecular explana-
tions in the life sciences. We base our Fig.3 in Section4 on the model illustrated by Fig.2.
13 It is important to emphasize that when describing different causal structures and assessing the explan-
atory power of their corresponding concepts throughout the paper we primarily have in mind individual
organisms. Thus, causal arrows are representing events in which one activity directly causes the other,
rather than different steps in evolutionary process as represented by Fig.2 (for a comprehensive sche-
matic representation see Fig.3). The reference to the population level, however, is important because it
provides information about robustness of the specific causal process in question. In other words, causal
chains shielded by natural selection will more likely proceed in a regular manner.

European Journal for Philosophy of Science (2021) 11:60
1 3
60 Page 14 of 21
the explanatory significance even of highly unstructured intra-organismic events. This
is the case, according to the de novo theory, insofar as those events are related to the
more structured ones, with mechanisms, such as in the described case study, the gen-
eral mechanism of protein synthesis, being the main complete exemplars of an organ-
ized causal architecture. This brings us to the idea of an explanatory pattern in molec-
ular biology, exhibiting a hierarchy of causal structures to which we now turn.
4 Explanatory hierarchy ofcausal structures
As noted in Section2, we are not entirely convinced by Ross’ characterization of
the distinction between mechanisms and pathways. We find, however, common
ground with the view that mechanisms are characterized by their discrete and isol-
able nature (see Ross, 2021: 150). We take it that this ensures a fixed structure of
the mechanistic causal sequences, as opposed to other causal sequences with looser
structures, allowing different causal potential routes with different end-products.
Now, if we were to understand pathways as referring to an interconnected space of
different potential causal routes, as in Ross’ example of the spread of a disease, then
this would allow us to distinguish them clearly from mechanisms. However, the fact
that there are highly structured pathways not allowing different causal routes shows
that there is no clear-cut distinction between mechanisms and pathways, but, rather,
that pathways can exhibit more or less of mechanistic features.
Our aim here is to show that different causal structures and their corresponding
concepts do not lead to different explanatory patterns. Instead, we argue that a uni-
fied mechanistic explanatory pattern characterizes biological explanations refer-
ring to the diversity of causal structures. There is a difference, nevertheless, in the
degree of explanatory power, depending on the type of causal structures invoked by
an explanation. We will argue that explanations referring to the concept of a mech-
anism have the highest degree of explanatory power. Accordingly, invoking other
types of causal structures will be explanatory to the degree in which these structures
exhibit mechanistic features.
At this point, we should say more about how we distinguish mechanisms from
other causal structures in molecular biology and, more importantly for the pur-
poses of this paper, what it is about the mechanism concepts that makes mechanistic
explanations more powerful or deeper than explanations referring to other causal
structures. We assess the explanatory power of mechanistic explanations by refer-
ring primarily to their non-sensitivity (see Ylikosky & Kuorikoski, 2010), the idea
that explanation is the more powerful, the less sensitive it is to the changes of back-
ground factors value. In other words, the explanatory power hinges on the sensitivity
of causal concepts and claims referred to in the corresponding explanatory patterns.
Namely, a causal claim is insensitive if it would hold under various fluctuations of
background factors to which it relates (see Woodward, 2006). Sensitivity and insen-
sitivity of causal claims come in degrees and so does the explanatory power of
explanations referring to different types of causal structures. In the present paper,
our aim is not to offer a theoretical justification for specifying the non-sensitivity
as a function of explanatory power. Rather, we simply rely on the fact that people

1 3
European Journal for Philosophy of Science (2021) 11:60 Page 15 of 21 60
generally tend to judge highly sensitive causal claims as explanatorily defective or,
at least, as non-standard (see Woodward, 2006).
We consider mechanisms or mechanistic arrangements more generally as explan-
atorily privileged (in the above sense of providing more explanatory power than
other causal concepts in biology), exactly because they secure the insensitivity of
causal relationships comprising a mechanism. With regard to that, we single out
the following basic characteristics of a biological mechanism: (1) insulation, in the
sense that mechanisms are discrete, have relatively clear spatial and temporal bound-
aries, and a certain degree of protection or cushioning from goings-on in their envi-
ronment; (2) productive continuity, there should be a sequence of causal interactions
progressing from one step to another.14
We have started out with the insulation characteristic, rather than with the sec-
ond, more salient one in NM, because the insulation in question is responsible for
a relative insensitivity of the causal relationships making up a mechanism. A cer-
tain amount of buffering against environmental conditions ensures that the arrows
representing causal relations point to relatively fixed and resilient causal structures.
That is, it allows a causal progression without significant interruptions and perturba-
tions. Let us now try to spell out in more detail the hierarchy of causal explanations
according to the causal concepts they invoke. At the explanatory top are concepts
corresponding to causal structures with a maximum insensitivity to external pertur-
bations, i.e., the mechanism concepts. Lower levels of hierarchy are reserved for the
concepts referring to less structured causal processes.
Consider again the models examined in Section 3. A causal process is more
structured and insensitive to the degree in which it is embedded into a mechanistic
arrangement that, then, ensures a higher insulation from external perturbations. The
progression of events addressed by the explanatory models of the de novo theory
is sensitive in the abovementioned sense up to the point in which natural selection
starts to fix in a population the progression of causal events leading up to the new
adaptive advantage conferring protein, as illustrated in Fig. 2. Thus, we have an
individual causal structure occurring in an organism, which is relatively sensitive.
However, natural selection, in a certain scenario, ensures that one specific causal
sequence gets embedded into a mechanistic arrangement. In this end-step of the pro-
gression, natural selection also ensures that the causal process going on now in a
population of individual organisms is a token of the full-fledged mechanism of pro-
tein synthesis.
Now, at the lowest level of our explanatory hierarchy are explanations referring to
causal structures entirely lacking mechanistic features, but are potentially explanato-
rily interesting. In our case study, this level regards transcriptional and translational
noise in the mechanism of protein synthesis. We have singled out this type of causal
relationship, because of its ambiguous role in the biological explanatory contexts.
14 Similarly, Bechtel and Richardson (2010: 35) talk about mechanisms as “discrete systems in nature”;
Andersen (2014: 276) mechanisms as having “precise spatio-temporal boundaries”. Moreover, there is a
noteworthy debate on the modularity of mechanisms, which relates to the above insulation characteris-
tic. We are also sympathetic to the following understanding of that characteristic, according to which a
mechanism is “a chain of linearly interacting entities that are somewhat independent of the chains that
explain other phenomena in the system” (Franklin-Hall, 2008: 212; our emphasis).

European Journal for Philosophy of Science (2021) 11:60
1 3
60 Page 16 of 21
On the one side, invoking such random and unorganized chains of causal events is
typically not explanatory. Rather, we are interested in exactly the opposite, that is,
regular, mechanistic arrangements that occur repeatedly and are relatively stable. On
the other side, what distinguishes biology and the life sciences in general from the
engineering contexts, as paradigmatic mechanistic environments, is that such a sto-
chasticity is not inevitably damaging. Namely, it can generate beneficial outcomes,
as seen in the cases of de novo gene synthesis. Hence, scientific use of the concepts
that refer to such causal relationships needs not be entirely explanatorily vacuous in
the following situations: (i) when an occasional noise gives rise to a more structured
causal process, potentially leading up to mechanistic arrangements (as shown in our
case study); and (ii) in the cases of first occurrences of a mechanistic arrangement
from non-structured biological processes.
Let us illustrate the above explanatory ordering of causal structures. The acciden-
tal transcription or so-called ‘transcriptional noise’ in a genome may end up with a
systemic gene expression, as represented in more detail by Fig.3.
Note that the mechanisms in Fig.3 are represented by dash arrows to empha-
size that we are not necessarily referring to an individual causal structure, but rather
a regularity that occurs repeatedly both at the organismal and populational levels.
Fig. 3 Explanatory ordering of causal structures in the case of de novo gene origin and functionality

1 3
European Journal for Philosophy of Science (2021) 11:60 Page 17 of 21 60
However, when talking about the explanatory power of causal concepts, we refer
here to particular causal processes occurring in individual organisms.15 Individ-
ual causal processes, proceeding between a particular DNA coding sequence and
its resulting protein product, are instances of this repeating pattern. First, α causal
structure, the mechanism of transcription, can also act ‘loosely’, i.e., it is not unusual
that some non-coding DNA gets transcribed into non-coding RNA molecules. Indi-
vidual causal events, such as the relation between non-coding DNA sequence and its
RNA product is represented by solid line arrows. The transcription occurring as a
result of noise is represented by round dot arrows. Such stochastic occurrences, then,
can lead up to two different scenarios. First, it leads to a transcript degradation, i.e.,
no new genes produced, or, more importantly for our explanatory model, to a non-
gene sequence acquiring a short ORF with a low-level transcription rate (step B in
Fig.3). Second, a fortuitous transcription of the short ORF in step C can lead up to
a proto-gene step, if a translational noise, a byproduct of the translation mechanism
(represented by round dot arrow), leads to a short and disordered protein product. In
step D, if the protein in question gets more elongated and structured, it may start to
provide adaptive advantage to a containing organism.
We are now at step E of the de novo model. Here, we can talk about the emer-
gence of a new gene, if natural selection starts to act in spreading this gene through
the population. Our account is highlighting ɣ mechanistic causal structure as a result
of embedding of the new gene into the corresponding regulative network leading to
a systemic production of the endo-phenotypic and phenotypic effects; this causal
structure is safeguarded against the workings of purifying selection due to its adap-
tive outcome(s).16 In that case, we can say that a new token of the mechanism of
protein synthesis has arisen. The new mechanism is analogous to the original one,
which triggered the progression in question, given the stochasticity in the mecha-
nism’s functioning. Note that this mechanism-token as well must be maintained by
natural selection.
Finally, let us sketch a corresponding explanatory hierarchy of causal structures.
At the bottom level, we have completely unstructured causal processes. Some pro-
cesses have next to zero explanatory power. Those that do have explanatory value, as
we have already seen, acquire it to the degree in which they contribute to structured
and ordered mechanistic arrangements. At the one level above in the hierarchy, we
have more structured etiological causal chains. Observe that, given our claim about
no clear-cut distinction between pathways and mechanisms, (some) structured etio-
logical chains can be considered as pathways. Whether they will be taken as path-
ways or not depends on how structured they are, i.e., whether they are structured
enough to capture “a fixed order of causal relationships that reflect which outcomes
15 This is not to say that the evolutionary causal processes are not explanatory or cannot be captured by
the corresponding concepts of mechanism or pathway. Rather, we take it that such cases are complex
and require separate analysis, which is beyond the scope of this paper. Here, we have tried to stay close
to the type of examples already present in the debate on the explanatory importance of different causal
concepts.
16 We will not explore in this setting what kind of causal structure natural selection is, for instance, a
mechanism or a process (for the opposing views, see, e.g., Barros, 2008; and Skipper & Milstein, 2005).

European Journal for Philosophy of Science (2021) 11:60
1 3
60 Page 18 of 21
need to occur before and after others in the unfolding of a causal process” (Ross,
2021:139).17
It is the embedding of etiological causal chains into larger mechanistic structures
what makes them more explanatorily compelling due to the rise in organization and
robustness of the chain of events. Isolated etiological causal chains, when not embed-
ded into a larger organized system, as seen in the cases of fortuitous causal processes,
lack the kind of structure that makes them explanatorily interesting, until they get
embedded into a larger mechanistic system (in our case, at the point when natural selec-
tion starts to act to preserve the resulting protein product). Thus, the acquisition of
mechanistic features renders the causal structures and their respective causal biological
concepts explanatorily relevant.
Our proposal for a basic explanatory hierarchy of causal structures in molecular
biology is illustrated by Fig.4.
The hierarchy in question arranges the causal structure and their respective concepts
according to their explanatory power, from the lowest at the bottom to the highest at the
top. We can see that the explanatory power increases as the causal chains acquire more
mechanistic features and become less sensitive. The above placement of the mechanism
Fig. 4 Basic explanatory hierarchy of causal structures
17 Thus, certain aspects or parts of the mechanism of protein synthesis occurring in individual organ-
isms can be described as pathways. We take it that this does not clash with standard scientific use. For
instance, in the literature, many references can be found for “the pathway of gene expression”, or for the
“signaling pathways for gene expression”. Notice that Ross herself provides the example of gene expres-
sion pathways. On our account, they are, however, parts of the mechanistic structure of protein synthesis.

1 3
European Journal for Philosophy of Science (2021) 11:60 Page 19 of 21 60
concepts accounts for the fact that these concepts are positioned at the center of explan-
atory efforts in the current practice of the life sciences.
5 Concluding remarks
In this paper, we have obtained the following results: (i) despite rightly acknowl-
edging the factual diversity of causal structures in biology, Ross’ recent account
does not succeed in showing that there is a clear-cut distinction between the
structures, that is, their corresponding causal concepts. Here primarily, the path-
way concept and the NM mechanism concepts. Their difference results fuzzier
than expected; (ii) as a consequence, Ross’ differentiation in question does not
imply that there are importantly different types of causal explanation. We have
substantiated our results by examining the current scientific practice concerned
with the explanatory models of de novo genes synthesis; (iii) what emerges from
the present analysis is an explanatory hierarchy of causal concepts (i.e., explana-
tory patterns applying such concepts), which is, on our account of explanation in
molecular biology, still topped by the mechanism concepts.
Acknowledgements Parts of this paper were presented at the following events: 7th Biennial European
Philosophy of Science Association Conference (EPSA 19), Geneva 2019 and theInternational Society
for the History, Philosophy and Social Studies of Biology (ISHPSSB) biennial conference, Oslo 2019.
We thank the audiences at these events for their helpful questions. Special thanks to two anonymous
reviewers for their exceptionally insightful, precise and constructive comments. This paper is an output
of theresearch project “Theoretical Underpinnings of Molecular Biology” (ThUMB), funded by the Cro-
atian Science Foundation, project grant number: HRZZ-IP-2018-01-3378, and doctoral grant number:
DOK-2018-09-7078. We would also like to acknowledge the support from the University of Rijeka (pro-
ject KUBIM: uniri-human-18-265).
References
Akram, M. (2013). Mini-review on glycolysis and cancer. Journal of Cancer Education. https:// doi.
org/ 10. 1007/ s13187- 013- 0486-9.
Andersen, H. (2014). A field guide to mechanisms: Part I. Philosophy Compass, 9(4), 274–283.
Barros, D. B. (2008). Natural selection as a mechanism. Philosophy of Science, 75(3), 306–322.
Bechtel, W., & Abrahamsen, A. (2005). Explanation: A mechanist alternative. Studies in History and
Philosophy of Biological and Biomedical Sciences, 36(2), 421–441. https:// doi. org/ 10. 1016/j.
shpsc. 2005. 03. 010
Bechtel, W., & Richardson, R. (2010). Discovering Complexity: Decomposition and Localization as
Strategies in Scientific Research. MIT Press.
Carvunis, A.-R., Rolland, T., Wapinski, I., Calderwood, M. A., Yildirim, M. A., Simonis, N., Charlo-
teaux, B., etal. (2012). Proto genes and de novo gene birth. Nature, 487(7407), 370–374. https://
doi. org/ 10. 1038/ natur e11184
Craver, C, & Tabery, J. (2019). Mechanisms in science. E. N. Zalta (Ed.), The Stanford Encyclopedia
of Philosophy (Summer 2019 Edition). https:// plato. stanf ord. edu/ archi ves/ sum20 19/ entri es/ scien
ce- mecha nisms/. Accessed 8 Apr 2020.
Craver, F. C. (2001). Role functions, mechanisms, and hierarchy. Philosophy of Science, 68, 53–74.
Craver, F. C. (2007). Explaining the Brain: Mechanisms and the Mosaic Unity of Neuroscience. Clar-
endon Press.

European Journal for Philosophy of Science (2021) 11:60
1 3
60 Page 20 of 21
Craver, F. C., & Darden, L. (2013). In Search of Mechanisms: Discoveries across the Life Sciences.
The University of Chicago Press.
Dujon, B. (1996). The yeast genome project: What did we learn? Trends in Genetics: TIG, 12(7),
263–270.
Franklin-Hall, L. (2016). New mechanistic explanation and the need for explanatory constraints. In
K. Aizawa & C. Gillett (Eds.), Scientific composition and metaphysical ground. New directions
in the philosophy of science. Palgrave Macmillan. https:// doi. org/ 10. 1057/ 978-1- 137- 56216-6_2
Franklin-Hall, L. (2008). From a Microbiological Point of View. Columbia University.
Frietze, Seth, & Leatherman, Judith. (2014). Examining the process of de novo genebBirth: An Eeu-
cational primer on ‘Integration of New Genes into Cellular Networks, and Their Structural Matu-
ration.’ Genetics, 196(3), 593–99. https:// doi. org/ 10. 1534/ genet ics. 113. 160895
Garson, J. (2013). The functional sense of mechanism. Philosophy of Science, 80(3), 317–333.
Gill, K., et al. (2016). Glycolysis inhibition as a cancer treatment and its role in an anti-tumour
immune response. Biochimica Et Biophysica Acta, 1866(2016), 87–105. https:// doi. org/ 10.
1016/j. bbcan. 2016. 06. 005
Glennan, S. (2002). Rethinking mechanistic explanation. Philosophy of Science, 69(S3), S342–S353.
https:// doi. org/ 10. 1086/ 341857
Illari, P., & Williamson, J. (2012). What is a mechanism? Thinking about mechanisms across the sci-
ences. European Journal for Philosophy of Science, 2(1), 119–135.
Machamer, P., Darden, L., & Craver, C. F. (2000). Thinking about mechanisms. Philosophy of Science,
67(1), 1–25. https:// doi. org/ 10. 1086/ 392759
McLysaght, A., & Guerzoni, D. (2015). New genes from non-coding sequence: The role of de novo pro-
tein-coding genes in Eukaryotic Evolutionary Innovation. Philosophical Transactions of the Royal
Society B, 370(1678), 20140332. https:// doi. org/ 10. 1098/ rstb. 2014. 0332
Okasha, S. (2019). Philosophy of Biology. Oxford University Press.
Plissonneau, C., Benevenuto, J., Mohd-Assaad , N., Fouché, S., Hartmann, F. E., & Croll, D. (2017).
Using Population and Comparative Genomics to Understand the Genetic Basis of Effector-Driven
Fungal Pathogen Evolution. Frontiers in Plant Science, 8. https:// doi. org/ 10. 3389/ fpls. 2017. 00119
Ross, L. N. (2018). Causal selection and the pathway concept. Philosophy of Science, 85(4), 551–572.
Ross, L. N. (2021). Causal concepts in biology: How pathways differ from mechanisms and why it mat-
ters. The British Journal for the Philosophy of Science, 72(1), 131–158. https:// doi. org/ 10. 1093/
bjps/ axy078
Ruiz-Orera, J., Verdaguer-Grau, P., Villanueva-Cañas, J. L., Messeguer, X., & Mar Albà, M. (2018).
Translation of Neutrally Evolving Peptides Provides a Basis for De Novo Gene Evolution. Nature
Ecology & Evolution, 2(5), 890–896. https:// doi. org/ 10. 1038/ s41559- 018- 0506-6
Salmon, W. C. (1984). Scientific explanation and the causal structure of the world. Princeton University
Press.
Schaffner, K. F. (2008). Etiological models in psychiatry. In K. S. Kendler, & J. Parnas (Eds.), Philosoph-
ical Issues in Psychiatry: Explanation, Phenomenology, and Nosology, (pp. 1–49). John Hopkins
University.
Schmitz, J. F. & Bornberg-Bauer, E. (2017). Fact or fiction: Updates on how protein-coding genes might
emerge de novo from previously non-coding DNA. F1000 Research, 6 (F1000 Faculty Rev), 57.
https:// doi. org/ 10. 12688/ f1000 resea rch. 10079.1
Schmitz, J. F., Ullrich, K. K., & Bornberg-Bauer, E. (2018). Incipient de novo genes can evolve from
frozen accidents that escaped rapid transcript turnover. Nature Ecology & Evolution, 2(10), 1626–
1632. https:// doi. org/ 10. 1038/ s41559- 018- 0639-7
Skipper, R., & Milstein, R. L. (2005). Thinking about evolutionary mechanisms: Natural selection. Stud-
ies in History and Philosophy of Science Part C Studies in History and Philosophy of Biological
and Biomedical Sciences, 36(2), 327–347.
Tautz, D. (2014). The discovery of de novo gene evolution. Perspectives in Biology and Medicine, 57(1),
149–161. https:// doi. org/ 10. 1353/ pbm. 2014. 0006
Tautz, D., & Domazet-Lošo, T. (2011). The Evolutionary origin of orphan genes. Nature Reviews. Genet-
ics, 12(10), 692–702. https:// doi. org/ 10. 1038/ nrg30 53
Tautz, D., Neme, R., & Domazet-Lošo, T. (2013). Evolutionary Origin of Orphan Genes. Wiley. https://
doi. org/ 10. 1002/ 97804 70015 902. a0024 601
Thagard, P. (2003). Pathways to biomedical discovery. Philosophy of Science, 70(2), 235–254.
Woodward, J. (2006). Sensitive and insensitive causation. The Philosophical Review, 115, 1–50.

1 3
European Journal for Philosophy of Science (2021) 11:60 Page 21 of 21 60
Ylikosky, P., & Kuorikoski, J. (2010). Dissecting explanatory power. Philosophical Studies, 148(2),
201–219.
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published
maps and institutional affiliations.
Authors and Aliations
ZdenkaBrzović1· VitoBalorda1· PredragŠustar1
Zdenka Brzović
zdenka@uniri.hr
Vito Balorda
vito.balorda@uniri.hr
1 Department ofPhilosophy, University ofRijeka, Rijeka, Croatia