July 8, 2015

Zipf's law against 'Genghis Khan' sensationalism

Very interesting new short paper at BioRxiv:

Elsa G. Guillot & Murray P. Cox, High Frequency Haplotypes are Expected Events, not Historical Figures. BioRxiv 2015 (pre-pub, freely accessible) → LINK [doi: http://dx.doi.org/10.1101/022160]

Abstract

Cultural transmission of reproductive success states that successful men have more children and pass this greater fecundity to their offspring. Balaresque and colleagues found high frequency haplotypes in a Central Asian Y chromosome dataset, which they attribute to cultural transmission of reproductive success by prominent historical men, including Genghis Khan. Using coalescent simulation, we show that these high frequency haplotypes are expected simply by chance. Hence, an explanation invoking cultural transmission of reproductive success is statistically unnecessary.


Not surprisingly it is, once again, the hyper-sensationalist, hyper-recentist, over-simplifying and evidence cherry-picker geneticist Patricia Balaresque who is the object of these very legitimate criticisms.

The basic argument is very simple: in neutrality conditions haplotype distributions follow Zipf's power law, while a single-founder effect of the type of the alleged Genghis Khan one would never cause that: one lineage would be outstanding, while the rest would show no hierarchy. 

However the authors, to make their argument even more certain, simulated genetic data under the standard coalescent, a neutral model that does not include cultural transmission of reproductive success. As you can expect, the simulations confirmed that what surprised Balaresque and others is just absolutely normal by mere chance: no Genghis Khan effect ever took place.

15 comments:

  1. I totally agree with you Maju!!.

    ReplyDelete
  2. Thanks for highlighting this paper
    I wonder if these cautions could extend to the Copper / Bronze Age "superfathers".

    ReplyDelete
    Replies
    1. I see no reason to imagine otherwise: I never took such ideas seriously (none the less because of the impossibility of properly estimate the age of a lineage).

      Delete
    2. This comment has been removed by the author.

      Delete
    3. Nirjhar: is that a question or a statement? It is written as the latter but has a question mark, so perplex.

      Delete
  3. This comment has been removed by the author.

    ReplyDelete
  4. "no Genghis Khan effect ever took place."

    "Ever"?

    Nothing in this says it *couldn't* happen in the context of a mass invasion and multiple massacres like the Mongol invasion or if some small group had a TFR advantage; it just says it could happen by chance also.

    Do the calculations in the paper predict the levels of homogeneity that would result from a random event compared to a founder effect?

    ReplyDelete
    Replies
    1. Read the study and tell me. My understanding is that the GK effect would behave unnaturally and not within Zipf's law. I.e. there would be a highly successful lineage, then an "unnatural" gap, and then the normal distribution of other lineages that do follow Zipf's law. That's not observed.

      Delete
    2. I can well believe the random chance effect - all daughters etc.

      I'm thinking more of the Euro example with the star-shaped y dna clades and unusual homogeneity. It seems to me things like that could act as a cross-check.

      Delete
  5. Pongo esto aquí por ser el último artículo (lo borras luego).
    ¿No crees que sería útil integrar un buscador de google en el blog para poder buscar contenidos?
    Gracias.

    ReplyDelete
    Replies
    1. Arriba a la izquierda. A no ser que lo bloquees con algo desde tu navegador.

      Delete
    2. Gracias
      La advertencia de cookies de blogger lo tapaba XD

      Delete
  6. My read is that they ran 1,000,000 simulations, with their choice of model parameters (whatever those are), and found 27% contained 15 or more haplotypes carried by more than 20 men. It happened a quarter of the time.

    Run a simulation that actually allows at least infrequent randomised cultural transmission events and I wonder what they'd find? Probably more than 27% (even with whatever their model parameters are) while being consistent with the overall neutral model.

    A more complex, but also higher probability model, would influence our estimate of what actually happened. You'd set that greater probability against whether or not people actually think it is more parsimonious to assume that there is no cultural transmission of numbers of children fathered or not.

    There's more to the idea of "super fathers" (to use that term) than the overall frequency as well, but the spatial structure and time structure (and subsistance structure) of where and when they are found. That's another element their model needs to be able to mirror.

    I'm not particularly wed to the idea that this super fathers effect is what happened of course, I've never really been terribly interested in the Y chromosome lineages, this is just as it strikes me.

    ReplyDelete
  7. Interesting paper, Luis.

    Thanks.

    ReplyDelete

Please, be reasonably respectful when making comments. I do not tolerate in particular sexism, racism nor homophobia. Personal attacks, manipulation and trolling are also very much unwelcome here.The author reserves the right to delete any abusive comment.

Preliminary comment moderation is... ON (your comment may take some time, maybe days or weeks to appear).