DNA-Jail logo
  1. Victim references
  2. Direct references
  3. Relative references
Tsunami victim identification considerations Jan05
Progress on identifications Feb02
Forensic mathematics home page

World Trade Center Disaster Identification

Most of the identifications of victims from the WTC attacks will be possible only by DNA, if at all. Three weeks following the attacks, much is still speculative.

The essence of victim identification by DNA is comparison of a victim DNA profile with a reference of some kind. References can be either direct or indirect, meaning either DNA from personal effects like toothbrushes which may match a victim exactly, or DNA from close relatives which of course show only similarity.

All three kinds of references are being collected. There are distinct problems or potential problems associated with each kind.

  1. Victim references
  2. There is great uncertainty as to what quantity and quality of victim profiles will be obtained. The medical examiner's office is prepared to type every tissue sample found in the wreckage. However, many bodies may have been ground to dust or thoroughly cremated, destroying all DNA. Nobody seems to have a good speculation at this time how many people will have thus disappeared with no trace at all. Of the remainder, many body parts are so burned or will be so badly decomposed by the time they are recovered that at best limited DNA typing will be possible. Mitochondrial (mtDNA) typing is the best hope for degraded samples. Marrow from partially charred bones can yield DNA, especially mtDNA. Partial STR profiles will be obtainable for some samples. The proportion that yield full STR profiles cannot be expected to be large: it must decrease over time, and it is only about half(?) among the preliminary samples (mostly peripheral victims) that are the only samples that have been collected to date.

    Hence the final results will be imperfect. A full STR profile essentially guarantees an identification providing appropriate references exist, but these may be few cases. Mitochondria is more problematic. 60-70% of mtDNA matches may be unique and thus imply and identification; the remaining 30% will be useful but not definitive. In some cases mtDNA combined with a partial STR profile will be sufficient.

  3. Direct references
  4. Direct references are unreliable for a reason that is obvious once you hear it, though I for one didn't think of it without being told. Sometimes the type obtained from a toothbrush isn't the type of the owner. Thus some of the direct references will turn out to be useless. The useless ones probably will do no harm though, barring the rare case that the person who borrowed the toothbrush is also among the missing.

  5. Relative references
  6. Indirect references – DNA samples from relatives of victims – present combinatorial and computational challenges:
    1. Given two long lists of DNA profiles, victims and potential relatives, how do you match them up?
    2. With what confidence is the matching correct?

    Neither question has an easy and straightforward answer. WTC prospective of DNA analysis (Powerpoint), is the substance of my presentation at the October 3, 2001 "Summit meeting of genetic experts" (WSJ, October 3) at the New York medical center, later considerably extended and revised. The essence is that matching up the lists is tricky because the lists will be rather long, and the number of cross-comparisons to consider is the product of the list lengths – millions, if a thousand victim profiles are recovered. Therefore the first step must be a "screening" step in which candidate relationships are found by using some sort of simplified (because it has to be fast) computer searching heuristic. However, it seems inevitable that among so large a number of comparisons there will be many unrelated people who coincidentally bear the same genetic similarity as do siblings, or as do parents and children. Therefore considerable skill may be necessary to screen the lists in such a way as to avoid being overwhelmed by false matches.

    Part of the trick must be that each victim needs at least two reference relatives. A similarity between a victim and only one reference from a given family can then be ignored as being a spurious match.

    The result of the screening step is a list of tentative or "candidate" associations between living families and victims. Testing these candidates will be a job for the Symbolic Kinship Program, mainly in the hands of people in the DNA laboratory of the New York City Medical Examiner.

    We still need to consider carefully what collection of reference relatives is likely to be sufficient to identify a victim. The estimates in my Powerpoint are overly optimistic because before the meeting I was unaware of the important point that many of the profiles will be partial profiles.

    Go to the top of this page