When I Get Skeptical of Science

Found this interesting article from Instapundit, which followed through to this, more detailed description of research that shows evidence of inbreeding between modern humans and other proto-human species:

“Our work is different from the research that led to the breakthroughs in Neanderthal genetics,” he explained. “We couldn’t look directly for ancient DNA that is 40,000 years old and make a direct comparison.”

To get past this hindrance, Hammer’s team followed a computational and statistical approach.

“Instead, we looked at DNA from modern humans belonging to African populations and searched for unusual regions in the genome.”

Because nobody knows the DNA sequences of those now extinct archaic forms, Hammer’s team first had to figure out what features of modern DNA might represent fragments that were brought in from archaic forms.

“What we do know is that the sequences of those forms, even the Neanderthals, are not that different from modern humans,” he said. “They have certain characteristics that make them different from modern DNA.”

The researchers used simulations to predict what ancient DNA sequences would look like had they survived within the DNA of our own cells.

“You could say we simulated interbreeding and exchange of genetic material in silico,” Hammer said. “We can simulate a model of hybridization between anatomically modern humans and some archaic form. In that sense, we simulate history so that we can see what we would expect the pattern to look like if it did occur.”

It’s an interesting theory, but any time someone tries to tell me they can successfully model complex systems with a computer program, I get very very skeptical. From my previous job, I know the difficulties in doing this with protein-ligand interactions, which we have a lot of experience modeling in-silico, and even that’s daunting. We’re also, still, not that remarkably good at it.

I would imagine to model something like this, you’d have to make a rather large number of assumptions. Since we also do not have African proto-human DNA, in contrast to what’s available for Neanderthals, I don’t see any way you could invalidate this model. One way I can think of is to see if it can successfully tag DNA sequences from known inter-species hybrids, where we do have the DNA for both parent species available. If it can do that, I might have some faith in it. I’d want to see it work on more than just Neanderthal cross-breeding with modern humans.

I should note, that the theory sounds plausible, and the evidence that it happened with Neanderthals is pretty strong, but my bullshit alarm goes off when computer models are employed to model complex systems with a lot of unknowns.

12 Responses to “When I Get Skeptical of Science”

  1. Ronnie says:

    So, does this also mean there is a possibility that the “out of Africa” human evolution theory is actually just total bullshit?

  2. Andy says:

    Won’t get any better at it unless we try.

    Not that I don’t take the results without an entire salt mine.

  3. Sebastian says:

    I agree with trying, but it’s not an easy problem.

    From what I can tell this doesn’t dispel the “out of Africa” theory, so much as suggest we came out of Africa having sex with a lot of other hominid species, and continued to do so after we got out of Africa.

  4. JeremyS says:

    Oh docking. I come here to get away from my MS thesis and you start writing about protein-ligand interactions. Thanks! ;)

  5. Sebastian says:

    What we did isn’t docking. It’s a bit different than that. Instead of just simulating a single pose, we would simulate a distribution of poses.

  6. JeremyS says:

    Sounds interesting. Proteins are quite flexible.
    Docking on its own has a pretty low accuracy. I docked multiple derivatives of different strengths. Then clustered them and correlated to known strengths. It worked in one case… all I need.

  7. Ronnie says:

    The Chinese now have human fossil evidence, which was discovered in China, which they claim disproves the “out of Africa” theory. The Chinese went public with this claim nearly two years ago, but it seems to me that their claim fell on deaf ears with American media outlets.

  8. Interpolation good. Extrapolation bad. That’s even true of fairly well behaved and understood systems which this isn’t. Also the problem with basing your research on what doesn’t exist is that it gets unverifiable very quickly.

  9. Sebastian says:

    Jermey S:

    We have a few papers out there on the technology.

    This is another technique we used:

  10. Sage Thrasher says:

    Check out the research papers done on the Neanderthal love child question that the Max Planck institute has published. They’ve also found DNA connections between a new Central Asian hominid species (contemporary with Neanderthals, but different) and current inhabitants of New Guinea. Fascinating stuff.

  11. JeremyS says:

    I’ll check them out. I am submitting my work to the same journal in a few weeks.

  12. Bill says:

    Funny, we can’t even figure out without a doubt if blacktail deer are a cross of mulies and whitetails, or if mulies are a cross of whitetails and blacktails, and we’re trying to figure out if homo sapiens sapiens is actually a cross of multiple other hominids from many millenia ago!

    As stated above…interpolation good, extrapolation bad!!