18 NOVEMBER 2005 VOL 310SCIENCEwww.sciencemag.org
tion in mass casualties (e.g., plane crashes)
efore the September 11, 2001, World
Trade Center (WTC) attack, the use of
DNA profiling for victim identifica-
was typically lim-
ited to situations
with fewer than
500 persons (1–5).
Often, the condi-
tion of the remains allows rapid recovery of
intact bodies. However, the number and con-
dition of the remains in the WTC attack
After many deaths, especially in the after-
math of a mass fatality, there is an element of
disbelief. Accepting the loss is a component
of a complex grief process (6). Our motiva-
tion was to provide a tangible artifact of
remains to survivors to facilitate coping and
grief processes. Enabling family participa-
tion in some of the identification decisions
was a critical component of the effort.
At the time of the attack, no infrastruc-
ture existed for rapid, effective victim iden-
tification in large-scale disasters (>1000
victims). Processes had to be scaled up to
collect and analyze massive amounts of
data in order to return identified remains to
the families of almost 3000 victims. We
summarize some challenges of the DNA-
based component of the victim identifica-
tion process, how these were met, and con-
siderations for the future.
Some mass fatality identification projects
begin with a list of victims (e.g., airline flight
manifests listing passengers and crew). In
contrast, the WTC mass fatality was initially
“open” because the number of victims was
unknown. Concerns about unreported or
fraudulently reported victims made estimates
difficult. The condition of the remains ranged
from a few nearly complete bodies to multi-
tudes of tiny fragments of charred bone, often
difficult to distinguish from inorganic mate-
rial. The fires affected the remains with tem-
peratures exceeding 1000°C (7) that burned
for more than 3 months. The towers’collapse
fragmented and commingled victim remains
and admixed building material. Many tissue
fragments were retrieved months after the
crashes, and bacterial and other processes fur-
ther compromised the DNA. These factors
made it difficult to isolate and genotype the
DNA from the specimens.
Identification of human remains by
DNA typing requires reference samples for
comparison. These and other sources of
information formed a deluge of material
and data to be cataloged, archived, and ana-
lyzed. Preexisting sample collection and
identification methods were insufficient
for these needs.
Identification of WTC victim remains
was the responsibility of the New York
City Office of Chief Medical Examiner
(OCME), which is one of the largest and
most sophisticated in the country. Yet, its
resources and scope of experience had to be
expanded. The New York State Police
Department (NYSP) was responsible for
DNA analysis of reference samples. Several
private DNA laboratories also tested sam-
ples, and software vendors helped to
develop data analysis and compatibility
tools. The NYSP and the OCME asked the
National Institute of Justice (NIJ) to
convene a group of scientific and medical
experts [the Kinship and Data Analysis
Panel (KADAP)] to advise them in
the DNA identification effort. KADAP
included experts in forensics; bioinformat-
ics; and molecular, medical, statistical, and
population genetics. The KADAP’s charge
was to assist the OCME in the development
of procedures, standards of evidence, and
processes related to the DNA identification
effort. The final determination of a specific
identification rested with the OCME (8).
DNA identification technologies. A bat-
tery of genetic identification markers, the
Combined DNA Index System 13-locus
short tandem repeat (STR) panel (CODIS)
(9), was initially selected because it was
established in forensic and legal systems and
was compatible with forensic software pack-
ages. The first round of STR genotyping had
a relatively low yield, because of DNA dam-
age and other factors. Therefore, KADAP
recommended several other approaches.
Because there are many more copies of
mitochondrial DNA (mtDNA) than nuclear
DNA, mtDNA analysis can be successful
when DNA is limiting (10, 11). Although
mtDNA typing is generally labor-intensive,
mtDNA typing with semiautomated analy-
sis was provided for this project. Alone,
mtDNA typing is insufficient to meet the
threshold of identification and could only
be used in conjunction with STR profiles.
KADAP also recommended use of “mini”
STR markers (12–16) that encompass the
same CODIS STRs, but use shorter ampli-
cons, which makes them more likely to be
successful on fragmented DNA. Finally,
technology for typing single-nucleotide
polymorphisms (SNPs) was considered.
Similar in concept to the use of mini-STRs,
SNP typing can work with fragmented
DNA because the amplicons are small.
Statistical criteria. About 5000 persons
were initially assumed missing, so we used
1/5000 as the prior probability that a tissue
fragment was from a particular missing per-
son. This prior was lowered to 1/3000 as the
victim list was refined. Direct DNA profile
comparisons were used to test for identity
DNA Identifications After the
9/11 World Trade Center Attack
Leslie G.Biesecker,* Joan E.Bailey-Wilson,Jack Ballantyne,Howard Baum,
Frederick R.Bieber,Charles Brenner,Bruce Budowle,John M.Butler,
George Carmody,P.Michael Conneally,Barry Duceman,Arthur Eisenberg,
Lisa Forman,Kenneth K.Kidd,Benoît Leclair,Steven Niezgoda,Thomas J.Parsons,
Elizabeth Pugh,Robert Shaler,Stephen T.Sherry,Amanda Sozer,Anne Walsh
*Authors were members of the World Trade Center
Kinship and Data Analysis Panel.Affilations are pro-
vided in Supporting Online Material.The opinions
expressed here are those of the authors.
*Author for correspondence. E-mail: leslieb@helix.
PLANNING FOR FUTURE DISASTERS
Further research and development of
forensic DNA typing systems is needed.
Software must be able to integrate
Information technology infrastructure
must be adequate to interconnect data-
reporting functions (22,23).
A kit should be developed for mass-
fatality reference sample and kinship
data collection,as well as for family
education and counseling.
External prioritization requests should
Processes should be designed to test and
to validate novel identification processes
concurrent with their development.
Criteria for determining end points
should be designated early in the
identification process (24).
Enhanced online at
Published by AAAS
P O L I C Y F O R U M Download full-text
www.sciencemag.orgSCIENCE VOL 310 18 NOVEMBER 2005
of tissue samples to reference samples from
victims when these could be obtained. The
CODIS marker set typically produces a ran-
dom match probability (the probability of
finding the same DNA profile in a ran-
domly selected, unrelated individual) that is
much less than 10–10in most populations.
This random match probability, when com-
bined with the prior probability, yields the
final match probability (17).
KADAP chose an initial minimum ran-
dom match probability threshold (10–10). This
was stringent, but there was consensus that the
approach would permit the OCME to make
the easier identifications promptly and with
high confidence. As the project progressed,
fewer novel DNA profiles were generated,
and the victim count decreased. Therefore, the
required random match probability was
reduced to 2.5 ×10–10when DNA-based gen-
der testing gave no results, to 5 × 10–9for
females, or 5 × 10–10for males when gender
markers were available (18). Details of these
methods are described elsewhere (19).
Although they represented a wide range of
continental origins, many victims’ back-
grounds were unknown, so data on allele fre-
quencies were not available for all groups.
KADAP recommended that random-match
probability be calculated using frequency data
from each of four major population groups.
The result that gave the most conservative ran-
dom match probability was used (20).
Kinship analyses (comparison of a WTC
remains DNA profile to those of biological
relatives of the victims) were used when vic-
tim reference samples were not available or
to confirm or increase confidence in direct
comparisons. A kinship likelihood ratio was
calculated for a victim sample by using fam-
ily reference data, calculated jointly when
multiple kin samples were available (17).
This likelihood ratio compares the probabil-
ities of observing a given DNA profile if a
victim sample belonged to a particular miss-
ing individual (based on the stated genetic
relationship of kin providing reference sam-
ples) to the chance of observing the profile
if it was from an unrelated person. This like-
lihood ratio, when multiplied by the prior
odds, yields the posterior odds (of kinship).
To reach KADAP’s goal of high confidence
of correct identification, sufficient loci and
family members were needed to reach a like-
lihood ratio of at least 3 × 106:1 (in favor of
the stated genetic relationship).
Software issues. Matching thousands of
victim sample genotypes (and partial geno-
types) to next-of-kin and/or reference sam-
ples meant new software tools were needed.
Data format incompatibilities and difficul-
ties interconnecting the data set were major
technical challenges. Contractors wrote
custom “middleware” interfaces to unify
several database and analytic functions.
This provided a virtual tool linking many
previously independent functions and freed
the analysts from manually moving data
among different platforms. Prioritizing
samples was critical to maintaining high
throughput. Work flow and tracking soft-
ware modules were also developed.
Protocols and software for evaluating and
performing quality control of the software
and analytic processes were designed and
The OCME recognized that its comput-
ers and data communication facilities were
inadequate for this project. Data transfer
between the NYSP and the OCME required
new information technology infrastructure
and support. A primary data repository
allowing shared access to analysts outside
OCME was set up on a secure server at
the National Center for Biotechnology
Information, in Bethesda, Maryland.
Because collection of personal reference
and kinship samples was implemented rap-
idly, ~1/6 of the initial data had to be cor-
rected or resampled. The KADAP developed
new kinship and personal reference collec-
tion forms and components for standardized
sample collection kits for future collections.
Summary of the effort and thoughts for
the future. The OCME cataloged 19,913
putative victim tissue fragments from 2749
individuals reported missing. The fragment
count increased to 20,120 because anthropo-
logical review identified commingled frag-
ments (confirmed by DNA profiles). The
DNA identification project generated more
than 52,000 STR, 44,000 mtDNA, and
17,000 SNP profiles. As of September 11,
2005, about 850 of the 1594 victim identifi-
cations established for the 2749 WTC vic-
tims were based solely on DNA results. Most
DNA identifications used standard CODIS
genotypes. Although many CODIS geno-
types originally failed, technical improve-
ments leading to better DNA yields from
damaged samples gave useful DNA profiles
in 40% of the samples for which standard
procedures failed. Modifications included a
two-stage drilling process to isolate unconta-
minated bone powder from the compromised
specimens and modifications to the wash
incubations, buffer concentrations, and elu-
tion times for the DNA isolation kit (16).
Beyond standard CODIS STR typing tech-
nologies, more than 20% of the DNA identi-
fications were made solely from mini-STRs.
SNP analysis alone identified about 10 indi-
viduals with 10 more identifications made
when SNP genotypes were used to supple-
ment partial STR profiles. Additional identi-
fications were made when mtDNA typing
results were used to screen for potential
matches, followed by DNA re-extraction and
mini-STR retyping. No DNA-based identifi-
cations were accomplished by mtDNA
analysis alone, as expected. The rate of new
identifications has become negligible. The
OCME and the KADAP believe that addi-
tional large-scale efforts are scientifically
unwarranted at this time.
In looking toward the future, the KADAP
panel recognized several major needs for
improvements in technology and infrastruc-
ture (see table, p. 1122) (22). There is no
doubt that improved preparedness and
enhanced mass fatality forensic infrastruc-
ture would lead to more rapid and efficient
identifications in the event of future mass
disasters or terrorist attacks.
References and Notes
1. J.Ballantyne,Nat.Genet. 15,329 (1997).
3. F. R. Bieber, in DNA and the Criminal Justice System, D.
4. B. Leclair, C. J. Fregeau, K. L. Bowen, R. M. Fourney,
J. Forensic Sci. 49,939 (2004).
5. A.Alonsoet al.,Croat.Med.J. 46,540 (2005).
6. F.R.Bieber,Harvard Med.Alumni Bull. 76,34 (2002).
7. National Institute of Standards and Technology
8. Z.M.Budimlija et al.,Croat.Med.J. 44,259 (2003).
9. B. Budowle et al., in Second European Symposium on
Human Identification (Promega Corporation, Madison,
10. T. J.Parsons,M.D.Coble,Croat.Med.J. 42,304 (2001).
11. B. Budowle et al., Annu.Rev.Genomics Hum.Genet. 4,
Legal Med. 114,269 (2001).
13. P.Wiegand, M. Kleiber, Int. J. Legal Med. 114, 285
14. K.Tsukada et al., Leg.Med.(Tokyo) 4,239 (2002).
15. J. M. Butler,Y. Shen, B. R. McCord, J.Forensic Sci. 48,
16. M.M.Holland et al.,Croat.Med.J. 44,264 (2003).
17. I. W. Evett, B. S. Weir, Interpreting DNA Evidence:
Statistical Genetics for Forensic Scientists (Sinauer,
18. About four times more males than females worked at
the WTC,so the statistical threshold of a female profile
(by the amelogenin marker) was higher than for a male.
The reduction in stringency also meant if one fragment
sample was matched to a reference sample by the crite-
ria, a second fragment with a partial profile could also
be matched if the random match profile probability of
the two victim fragment samples was at least 1 × 10–8.
19. C. H. Brenner, B. S.Weir, Theor. Popul. Biol. 63, 173
20. The population group reference DNA panel comprised
525 samples previously collected by the OCME;
African American (n = 126), Caucasian (n = 123),
Hispanic (n = 151),and Asian (n = 125).
21. National Center for BioTechnology (www.ncbi.nlm.nih.
22. A “Lessons learned” document will be available
through the National Institute of Justice:
23. B. Budowle, F. R. Bieber, A. J. Eisenberg, Leg. Med.
(Tokyo) 7,230 (2005).
24. KADAP attempted to determine an end point early in
the process, but this could not be done because of
unknown variables and the overwhelming nature of
the attacks.The development of principles to facilitate
such determinations would be useful.
25. We dedicate this to the memory of the victims of the
WTC attack, to the families, and to everyone whose
efforts made the identifications possible.We gratefully
acknowledge Mark Dale for his advocacy and support.
Supporting Online Material
Published by AAAS