Forgot your password?
typodupeerror
Biotech Science

Mute Witness: Forensic Sketches From Nothing But DNA 68

Posted by timothy
from the why-is-your-yearbook-too-squishy? dept.
First time accepted submitter Todd Palin (1402501) writes "Researchers at Penn State university are trying to reconstruct images of faces based only on a DNA sample of the individual. As far out as this sounds, they did a pretty good job at matching the actual appearance of the faces. This is a pretty good start on a whole new use for DNA samples. Imagine a mug shot of a rapist based only on a DNA sample."
This discussion has been archived. No new comments can be posted.

Mute Witness: Forensic Sketches From Nothing But DNA

Comments Filter:
  • by Mortimer82 (746766) on Saturday March 22, 2014 @07:13AM (#46550239)
    I'm somewhat surprised that neither the article or summary mentioned how this seems inspired by Gattaca [imdb.com].
    • Maybe that's because it wasn't, and is actually just the latest development in the continuing advancement of the study of genetics, and would have happened regardless of the existence of a 17-year-old science fiction film.

      And last time I checked, New Scientist hadn't yet had to stoop to opening most articles with "It sounds like something out of [insert vaguely relevant pop culture reference here], but scientists now say..."

  • by Idou (572394) on Saturday March 22, 2014 @07:15AM (#46550249) Journal
    Well, they always had access to it, but they just could not make it useful by mapping it to a specific identity.

    I wonder how many unique individual DNA can be extracted on average by taking a sample of rain run-off from a busy city street? Let me coin the process here as "Gutter Diving."
    • Yes, took a while to find, but I couldn't have asked for better. You look about my body build and proportion. Same skin, eye and hair color. Seen in the dark or in passing in a hoodie, I doubt someone wouldn't be equally likely to pick either of us from a lineup. Ah, that was a fine cup of coffee, wasn't it. Well, in addition to dropping mine in that refuse bin too, I'm snatching yours out; Saliva, [x] Check. Oh, silly me, I just meant to drop in this napkin, not my keys and cup. How embarrassing! See

  • Imagine when they investigate 100 sperm donor siblings or "cousins"...
  • would it be necessary to come up with an approximate facial image? If you have the DNA sample, that is far more definitive than the approximated face. I don't understand how this is useful. It might be interesting, but I don't see any practical use in forensics.
    • to find the person in the first place, the DNA might not be on any databases.
      • You will never find the person with a crude approximation of their face, which of course is subject to all sorts of modifications and changes throughout life. You would end up bringing all sorts of innocent people in for questioning because their face looked a little bit like the reconstruction. This is no way to solve a crime.You solve crimes other ways than putting up a wanted poster with a face that may or may not look like the person in question.
        • so why do they have police sketch artists and photofit systems? just to please the thriller writers? it is just one added weapon in the armoury, i'm surprised you cannot see this. its not going to convict anyone, the DNA will do that if it matches anyone they find via the sketch
    • by Jmc23 (2353706)
      What do you think is more useful on a wanted poster.

      A person's face ->which pretty much everybody can understand

      A dna print ->which nobody can understand.

      Seriously slashdot?

    • If you have the DNA sample, that is far more definitive than the approximated face.

      But what if you don't have the DNA sample from a suspect? Or even any suspects?

      • then you can't do this process, doh
        • *facepalm*

          I meant a situation where you have DNA from the crime scene, but don't have and/or can't get DNA from a suspect.

          • the point of this process is to create a starting point for finding suspects. How they get the DNA from a suspect is however they do it now and not really relevant to this process.
    • by s1sfx (1883880)
      It's a way to do racial profiling.
  • Plastic surgery FTW, eh?

    Ok, in seriousness, I could see this as a great tool for anthropologists, but so far as crime-solving goes, it's just about guaranteed to have a false- positive rate about 100x the true positive rate. OTOH, that works for the TSA...

    • but so far as crime-solving goes

      It's not for crime-solving. It's for crime-investigating.

      it's just about guaranteed to have a false- positive rate about 100x the true positive rate

      Based on the numbers pulled out of whose behind?

      If you've got a shortlist of 10, or even 100, suspects and no other evidence - yet - don't you think even a rough idea of what that person probably looks like might be helpful?

      Didn't you play Guess Who? [wikipedia.org] as a child?

      • If you've got a shortlist of 10, or even 100, suspects and no other evidence - yet - don't you think even a rough idea of what that person probably looks like might be helpful?

        If you've got a shortlist of 10 people, plus DNA evidence, you don't need to guess what he looks like.

        You probably don't need to guess what he looks like if you have 100 people and DNA.

        On the other hand, if you have a shortlist of 10 or 100 people and NO other evidence, you have no DNA to fake up a picture.

        • If you've got a shortlist of 10 people, plus DNA evidence, you don't need to guess what he looks like.

          You do if you have no DNA for comparison. Maybe none of the suspects won't give a sample. Maybe you can't get a court order forcing them to provide. Maybe you don't even know who the suspects are - maybe all you have is evidence of their presence near the scene (CCTV, for example). With a rough idea of what the culprit looks like, you can concentrate any subsequent CCTV trawl on the most likelies.

          On the other hand, if you have a shortlist of 10 or 100 people and NO other evidence, you have no DNA to fake up a picture.

          Err, yes. Obviously. So?

        • to guess what he looks like will help to give you the 10 people in the shortlist then you match DNA. Police already use artists to create an image so they can create a short list so if this process really works then its another tool in the armoury.
    • by stoploss (2842505)

      it's just about guaranteed to have a false- positive rate about 100x the true positive rate. OTOH, that works for the TSA...

      Just wanted to let you know that there is a term for that: positive predictive value [wikipedia.org]. In the TSA's case, the PPV would be effectively 0.

      Thought I would share in the hope that others will return the favor for me in the future.

  • I fished around the links in the article, including http://www.plosgenetics.org/ar... [plosgenetics.org] and it looks fascinating. But we really need to see more samples. The one of the reporter is quite good, but let's see them do it again.
  • by iktos (166530) * on Saturday March 22, 2014 @09:08AM (#46550571)

    Interview with Kayser ("we've only found the first five genes"): http://www.scientificamerican.... [scientificamerican.com]
    In short: Hair and eye colour prediction: 0.9, height: 0.75, everything else "much lower" than 0.75 with 0.5 being totally random.

    And from the article itself: "The next step is to run larger studies in different populations to confirm that the variants found so far are statistically reliable." which explains why there aren't any more test examples.

    A bit about how it works ("Fine Tuning of Craniofacial Morphology by Distant-Acting Enhancers"): http://www.evolutionnews.org/2... [evolutionnews.org]

  • It's a nice story, and they provide a MatLab environment to play around with their model, but ultimately I don't believe this work is reproducible given the materials provided. All we're really given is a sandbox to play in where we can adjust model parameters, and so the work should never have been published.

    What would convince me? For starters, the ability to take an arbitrary set of values for these SNPs, punch them in, and see the result change. If I put in SNPs from one of the CEU HapMap samples, I wou

  • OK, the police have a drawing of what the rapist might look like. Turns out I'm an exact match to the picture, but I decline to give a DNA sample. Do they have probable cause to allow the courts to force my hand?
    • if that was me and i was innocent, i'd have no issue with volunteering my DNA.
      • If you ever find yourself in that situation, consult with your lawyer first! If you are innocent it is not your responsibility to prove it. You can only make things worse.
        • how can i make it worse if I know my DNA will not match the crime scene? The only concern would be if they preserve my DNA on a database somewhere if I have volunteered it and I'm not convicted of any wrong doing.
  • ... tattoos, scars, meth teeth vs a nice white set of porcelain caps, orange Mohawk or a marine buzz cut, piercings, a broken nose.

  • She had cells on her from unknown sources. It would be great to have a look at the murderer that did this.
    Think of how fast it would be to match up.

    Interestingly, would courts require you submit a cell in future, if your pix appears to match one?
    Also, I wonder if this can be used to capture others that are sitting in jail, but did the crime and simply did not admit to it.
  • "Researchers at Penn State ... Imagine a mug shot of a rapist based only on a DNA sample."

    Such a case seems ironic given that this is the same university that was (recently) involved in a scandal related to such crime.

Life. Don't talk to me about life. - Marvin the Paranoid Anroid

Working...