Open Thread, 01/28/2019

If I haven’t made it clear, I highly recommend The First Farmers of Europe: An Evolutionary Perspective. A very readable book. One thing I haven’t emphasized is that the early European farmers seem to have been big consumers of cheese. This is curious as it doesn’t look like they have the modern European lactase persistence allele. Cheese is different from milk because the proportion of sugar is lower, as the fermenting process exhausts some of it. But cheese-base agro-pastoralism seems to have been common in many places before the arrival of Indo-Europeans.

David Reich is on giving talks in India. He has stated that the draft of the Indian ancient DNA paper is complete. This doesn’t speak to when it will be posted on bioRxiv, but we’re seeing the light at the end of the tunnel. Also, I assume the Iberian paper showing mass male-mediated migration ~2000 BC will arrive at some point this year.

A friend pointed me to a new book by Edward Dutton on J. Phillipe Rushton. After watching clips of Dutton speak about the contents of the book, it is not a flattering portrayal. Dutton depicts Rushton as a bit of a general sociopath (though he seems to couch it as part of Rushton’s individual “life history strategy” of being a “user”). More concerning than his personal life is that Rushton was clearly willing to fudge facts in regards to his science, which was of such a controversial nature regarding race and life history that ultimately he should have been much more careful than is the norm. Dutton documents some instances apparently in his book, but I can give another specific example.

In 2007 it came to my attention he was citing History and Geography of Genes to present a phylogenetic model of the relatedness of West and South Eurasia, where Europeans are one clade, and West Asians and South Asians another clade. In the book, L. L. Cavalli-Sforza explicitly mentions that this classification was due to the relatively smaller number of samples and data from West and South Asians. That is, phylogenetically they did not form a natural clade, but for the purposes of statistical power he pooled them together artificially. The reality had already become clear by the middle 2000s with microsatellite and the earliest high-density SNP arrays that West Asians and Europeans were a more natural clade.

I emailed Rushton to correct his misimpression. He explained that he was only citing what was in the book, and I quoted back to him what Cavalli-Sforza said in his note to show he was transmitting a misunderstanding, at which point Rushton sputtered and admitted it was beyond his ken as a psychologist to evaluate. Less than one year later I saw Rushton assert the exact same thing I had told him was just wrong, again citing History and Geography of Genes. Dutton has uncovered much more egregious scientific misconduct, but this is one case I have first-hand knowledge of. This is not the behavior of a scientist following the facts where they led….

This is an important paper. I’ve heard rumors of this finding for a long time. Direct estimation of mutations in great apes reconciles phylogenetic dating:

The human mutation rate per generation estimated from trio sequencing has revealed an almost linear relationship with the age of the father and the age of the mother, with fathers contributing about three times as many mutations per year as mothers. The yearly trio-based mutation rate estimate of around 0.43 × 10−9 is markedly lower than previous indirect estimates of about 1 × 10−9 per year from phylogenetic comparisons of the great apes calibrated by fossil evidence. This suggests either a slowdown in the accumulation of mutations per year in the human lineage over the past 10 million years or an inaccurate interpretation of the fossil record. Here we inferred de novo mutations in chimpanzee, gorilla, and orangutan parent-offspring trios. Extrapolating the relationship between the mutation rate and the age of parents from humans to these other great apes, we estimated that each species has higher mutation rates per year by factors of 1.50 ± 0.10, 1.51 ± 0.23, and 1.42 ± 0.22 for chimpanzee, gorilla, and orangutan, respectively, and by a factor of 1.48 ± 0.08 for the three species combined. These estimates suggest an appreciable slowdown in the yearly mutation rate in the human lineage that is likely to be recent as genome comparisons almost adhere to a molecular clock. If the nonhuman rates rather than the human rate are extrapolated over the phylogeny of the great apes, we estimate divergence and speciation times that are much more in line with the fossil record and the biogeography.

Some readers of this weblog will be curious about my other podcast, the “BrownCast.” Check it out on LibsyniTunes and Stitcher. Next week a podcast with Carl Zha of CLASH! will drop.

In 5G Race With China, U.S. Pushes Allies to Fight Huawei.

The transferability of lipid-associated loci across African, Asian and European cohorts.

A GWAS in Latin Americans highlights the convergent evolution of lighter skin pigmentation in Eurasia.

The Cultural Brain Hypothesis: How culture drives brain expansion, sociality, and life history.

Genetic architecture of human thinness compared to severe obesity.

Summary statistic analyses do not correct confounding bias.

Plan to enact the leaky gated model for February.

40 thoughts on “Open Thread, 01/28/2019

  1. Cultural Brain Hypothesis paper: Ambitious! Impressed at initial glance by synthetic argument/ratiocination-tooling (lotta models invented, looks to my naive eye).

    If people who are reasonably knowledgeable in adjacent fields care to comment here about this paper, I would be a keen-listening fly on the wall. (If I go to the sub-specialty sites directly, I would have to find such sites first, and then the talk would of course be too cranky and abstruse, and I lack time for that sort of fine-grained contentioning.) Thanks.

  2. tggp, don’t recall. the 2000s was a long time ago ;0)

    Ah. But will comments then go to paying customers only? That’s a common way to do it. Makes sense.

    i assumed comments would be open to anyone on open posts.

  3. Lots of discussion in the news these days over the name of the Former Yugoslav Republic of Macedonia. Now that the Greek parliament has ratified a key piece of legislation, it will be known as the Republic of North Macedonia.

    Greek nationalists are outraged and see it as an usurpation of the legacy of Alexander the Great.

    My recollection is that the whole area was subject to in-migration of Slavic groups in late Antiquity / early Middle Ages. “Greek” Macedonians and “North” Macedonians may not be as genetically different as they believe.

  4. My recollection is that the whole area was subject to in-migration of Slavic groups in late Antiquity / early Middle Ages. “Greek” Macedonians and “North” Macedonians may not be as genetically different as they believe.

    So Slavs masquerading as Thrako-Illyrians masquerading as Greeks?

  5. “is anyone else getting internal server errors or anything? getting more in the admin, wonder if a new plugin is doing it.

    No, nothing.

  6. Thanks to a tweet by Robin Hanson, I just learned of, and read, your post about variation in intelligence. The comments are closed on that post though so here I am.

    Human brains are exceptionally powerful, is it a coincidence that our bodies are exceptionally allocative? Being able to simultaneously carry a wide variety of resources over greater distances meant having to solve harder allocation problems. This put exceptional pressure on the selection of intelligence and voila! Here we are.

    Up until fairly recently, people who were exceptionally good at solving hard allocation problems exerted exceptional influence on the gene pool. This is no longer the case. We’ve reached peak intelligence.

    This would change if humanity started colonizing outerspace in the same fashion that we colonized the planet. The intelligence of your packing decisions could very well determine your influence on the gene pool.

    Intelligence is all about correctly estimating the usefulness of things. To be clear, this really isn’t something that IQ tests measure.

    Right now you’re using democracy to try and measure a comment’s usefulness. The thing is, voting can only measure how popular something is. If you truly want to measure usefulness then you need to use a market. I might be wrong. The proof is in the pudding. If you want to make an informed decision, then check out Honest Cash and Cent.

  7. “Mutant blue-eyed coyotes spreading through California: Five coyotes with baby blues, an eye color not seen elsewhere, were photographed in California—suggesting the rare trait is proliferating.” By Callie Broaddus

    PUBLISHED January 28, 2019

    Last spring, a photographer and guide in northern California’s Point Reyes National Seashore named Daniel Dietrich spotted an unusual animal: A female coyote with blue eyes. …

    Now, it appears the trait is spreading. In the last few months, at least four other coyotes with blue peepers have been seen and photographed within about a 100-mile radius of the seashore. …

    Due to the localized nature of the phenomenon, and the complete absence of it elsewhere in the country, scientists now hypothesize that a mutation to the genes that control eye color likely appeared several generations ago, and these newly-documented canines are recent descendants of one original “one-in-a-million” mutant. …

    While the occurrence of such a mutation is rare, the alternative—introduction of the gene through hybridization with domestic dogs—is more unlikely in this case. Coyotes and dogs can interbreed to form “coydogs.” These animals typically have significant changes in coat color and altered facial structure and proportions, says Stan Gehrt, a wildlife ecologist at Ohio State University and a National Geographic explorer. But Gehrt has never seen a change in eye color. …

    Golden-brown eyes have been selected for through millions of years of coyote evolution. “They have the color which is best for their environments and their way of living,” says Juan Negro, a scientist with expertise in animal eye hue. Thus, blue irises are almost certainly not advantageous to these animals, and it’s even possible that they could have downsides like interfering with camouflage or causing increased light sensitivity, Negro says.

    In other animals, like dogs, blue eyes have been selected through artificial breeding. Wolves, for example, do not have blue irises.

    In many ways, coyotes have it better now than in the past. Humans have killed or driven out many of their predators, such as wolves and mountain lions … Thus, there are different—and more relaxed—selective pressures on the animals. …

  8. In other animals, like dogs, blue eyes have been selected through artificial breeding.

    I do not think this is the case with Siberian Huskies. Some Siberians have blue eyes, some brown eyes, and some bi-color.

  9. is anyone else getting internal server errors or anything

    I got one a couple of days ago. Just once. And first time in a long while.


    In some contrast to ‘superlinear scaling’ models of city growth where large cities are more productive than the sum of their parts by 15% for every 200% increase in size, models find large amount of this (40%) driven by selection bias. E.g. cities probably more like 9% more productive for every doubling in size, controlling for intake characteristics. Driven by higher premiums for high income potential individuals. Supports similar finding in UK Biobank data (which suggest a degree of genetic influence).

    Not part of the study but would also support that this centralisation is welfare reducing in areas of welfare where you benefit from having a spread of high income folks through a country, rather than a concentration (for one, downside of increased social interconnectivity in a city is decreased social interconnectivity throughout a whole nation).

  11. Twinkie: I am not a dog person. Aren’t English sheepdogs supposed to have one blue and one brown eye? Also, aren’t Huskies backcrossed to wolves frequently? These may be myths dog people have told me.

  12. I think stories like this will become a staple of human interest stories in the media over the next few years.

    “Two Sisters Bought DNA Kits. The Results Blew Apart Their Family. In an age of ubiquitous direct-to-consumer genetic testing, family secrets are almost impossible to keep.” By Amy Dockser Marcus Feb. 1, 2019

    It is too long and rambling to quote and is paywalled.

    The upshot is that the subject family consisted of father(D), mother(M) and 4 kids, 2 girls and 2 boys (F1,F2,M1,M2).

    F1 gets DNA test from Ancestry. Results tell her that she is closely related (half sibling) to a man outside the family. She gets F2 to take test. F2 is not related to outside man and is only a half sibling of F1. So, D had fathered male outside of family, and M had also strayed, D was not F2’s father.

    Emotional turmoil follows as connections are made with F1’s half brother, and F2’s biological father. Heartstrings are tugged. Everyone is sadder Budweiser.

  13. The response to the Kirkegaard paper is interesting. There are perhaps some concerns about whether stratification is adequately controlled for, but the response (if any) has been disappointingly poor or entirely free of content, showing the strength of political biases when certain topics come up.

  14. There’s an absolutely huge Siberian Husky living near me that has blue eyes.

    Bigger than a cold climate wolf, and more heavily built – the perfect sort of dog to keep in a small apartment in a sub-tropical climate (sarcasm). Very scary looking animal, but it’s terrified of bicycles, which cause it to freeze, cowering, plus it can barely drag itself around.

  15. ““Greek” Macedonians and “North” Macedonians may not be as genetically different as they believe.”

    By looking at individual samples, they’re much more different than their geographic distance would suggest, though they both derive the majority of their ancestry from pre-Slavic groups. The former much more than the latter on average.

  16. Aren’t English sheepdogs supposed to have one blue and one brown eye? Also, aren’t Huskies backcrossed to wolves frequently? These may be myths dog people have told me.

    I don’t know anything about English sheepdogs. Siberian’s are reputedly an ancient breed and do have certain wolf-like tendencies, but cross-breeding with actual wolves is rare. Their blue eyes are a “natural” development.

    There’s an absolutely huge Siberian Husky living near me that has blue eyes.

    Bigger than a cold climate wolf, and more heavily built

    Siberian Huskies are medium-sized dogs. That doesn’t sound like a Siberian. Alaskan Malamutes are much bigger, but they have brown eyes. Could be some sort of a mix.

  17. n=53 is not an issue on its own if the sample size is representative of the wider Ashkenazi population (as the appendix suggests). a > 1SD difference in PGS is very unlikely if there is no real difference.

    and they have replicated the result in a larger sample.

  18. @Twinkie – I think you could well be right. I’m no expert. Eyeballing photos, it does look a lot more like an Alaskan Malamute than a Siberian Husky, with the exception of the blue eyes.

  19. Hey Razib, I have a question that I’m curious to see your answer on (I’ll copy my question from Anthrogenica):

    Quite recently, a pre-print discussing “Neanderthal introgression reintroduced functional alleles lost in the human out of Africa bottleneck” came out and it is a quite interesting pre-print that I recommend reading. The re-introduced alleles estimated are alleles with perfect r^2=1.0 linkage disequilibrium with known Neanderthal derived alleles (NDAs) that are either believed to be a high-confidence ancestral allele from 1000 genomes or an allele present in modern African populations.

    In particular to South Asian genetics, it was interesting that South Asians (SAS) had the highest number of estimated re-introduced alleles at 139,270, followed by East Asians (EAS) at 125,257 with Europeans (EUR) at 90,121. It should be noted however that this estimate is naturally conservative as some reintroduced alleles may have shifted away from perfect linkage disequilibrium with NDAs.

    This result seems surprising to me, however, given both Wong et. al 2014 and Wall et. al 2013 estimate that East Asian Neanderthal introgression is higher than that of South Asians and Europeans (it is worth noting that D stats test 2 in Wall et. al 2014 was inconclusive w.r.t South Asians and East Asians) . Likewise, if we take into account a more complicated picture of West Eurasian gene flow into West Africans given by Petr et. al 2019, the Neanderthal ancestry dilution hypothesis due to Basal Eurasians may not be true. This ultimately suggests that Neanderthal % is the same as those of Asians.

    What possible explanations could result then in South Asians having a similar to higher number of reintroduced alleles compared to East Asians and significantly so compared to Europeans?

  20. What possible explanations could result then in South Asians having a similar to higher number of reintroduced alleles compared to East Asians and significantly so compared to Europeans?

    honestly, i have no idea. and i’ve been wondering about it too! could it be another archaic we missed?

  21. @Traject, I’ll kind of have a go at thinking about the issues involved, though I have no answer.

    I’m considering possibly might be relating to their strategy of identifying Neanderthal tag SNPs may have some bias related to the indirect ratio (which is what gave the estimates that Petr’s paper seems to have refuted).

    The process to identify RAA (Reintroduced Ancestral, e.g. Homo Sap), RHA (Reintroduced Hominin), and NDA (Neanderthal Derived) is here:

    It looks like roughly proportionate reductions happen at most steps, so the main driver is identification more alleles at step 1 in their process for SAS, than for EAS and EUR.

    Then there are also slight differences in how the ratios that they call of RAA, RHA, NDA and unclassified – mostly the same, but EUR seem to have proportionately many unclassified, and proportion of NDA is about the same between all panels, so this is mainly trading off RAA+RHA against unclassified, in proportions (proportions of Neanderthal Derived Alleles relative to the filtering step 1 is basically identical between EUR, EAS, SAS).

    They note on these unclassified that:

    “Our approach identifies more than 200,000 RAs, yet more work is needed to comprehensively identify all RAs in Eurasians. For example, our conservative approach misses true RAs that no longer have perfect LD with the original Neanderthal tag SNPs. Furthermore, thousands of candidate RAs were not classifiable because they lacked a high confidence ancestral assignment or were not observed in modern Africans. Some of these unclassified variants are undoubtedly ancient, but thus far defy confident characterization due to their complex histories.”

    So I guess the questions are: 1) why are more alleles identified in SAS panel than EUR and EAS in step 1?, 2) why are proportionately so many more alleles unclassified in EUR?

    There’s not much info on how they’re identifying N tag snps, so not clear on what role that has to play. Possibly they may need to rethink how they’re identifying those to exploit the direct Neanderthal ratio, depending on how they’re doing it?

    Accepting results as accurate for a sec (ignoring unclassified) though, naively same proportion of ND alleles and more reintroduced allele in East Eurasian just implies same amount of total Neanderthal admixture, but pulse in East Eurasian happens after more radical bottleneck (more variants lost to be recovered?). However then that’s hard to reconcile with same amount of mixture (two independent pulses lead to same overall amount?).

  22. @Trajet, alternatively, ignoring the unclassified factor, greater Reintroduced but same Neanderthal alleles in EAS vs EUR might be explained by Denisovan/other Hominin pulse into EAS.

    At far lower size than Neanderthal introgression assuming about 2/3 of those reintroduced alleles (matching EUR) are from Neanderthal, and only the difference from EUR is from this introgression.

    But then you still have to explain the uniformly high numbers in SAS at every step…

  23. @Traject, final (I think) comment, looking again again, further note in their paper:

    “NDA totals include both I Neanderthal “tag SNPs” (EAS:132405, EUR:132296, SAS:179662) as well as the NDAs predicted from the present pipeline.”

    So this begins with significantly more tag SNPs (SNPs tagging introgressed Neanderthal haplotypes) in SAS.

    “Sequence data – Genomic variants were taken from 1000 Genomes Phase 3v5a data. Introgressed Neanderthal tag SNPs were downloaded from: (12)”

    12 is – “B. Vernot et al., Excavating Neandertal and Denisovan DNA from the genomes of 460 Melanesian individuals. Science (80-. ). 352, 235–239 (2016).”

    So passing over how Vernot identified tag SNPs (I don’t really grasp how the S* statistic Vernot uses to identify introgressed haplotypes actually works, sorry), normalizing to initial tag SNP count gives:

    PCA on the normalized amount in each class discovered through their pipleine (RAA, RHA, discovered NDA, unclassified):

    Confirms that normalized pattern is basically a unidimensional differentiation between EAS and EUR on reintroduced vs unclassified, with SAS intermediate. Reflecting SAS genetic history as very broadly considered ENA+West Eurasian groups?

  24. @matt

    Thanks for the answer! Looking at how Vernot identified tag SNPs from the 2014 data set, from Vernot and Akey 2014,

    We estimate that S* can identify ~30% of all true positives at a low false positive rate, and that this result is largely invariant across demographic models

    Finally, because Neanderthal admixture is expected to have occurred only in non-African populations, variants on the introgressed haplotype should not be found in African individuals (an assumption that can be relaxed to allow for the possibility of gene flow between African and non-African individuals). Thus, S* is designed to detect divergent haplotypes whose variants are in strong linkage disequilibrium and are not found in a “reference” population. It is important to use a reference population that contains a minimal amount of archaic (i.e., Neanderthal) introgressed sequences. We chose 13 randomly selected Yoruban genomes as a reference population; a recent analysis found that levels of Neanderthal variation in Yorubans are not statistically enriched (as measured by the D statistic), as opposed to Sandawe, Maasai, and African Americans (45).

    The S* statistic appears to find the maximum scoring subset of SNPs in a locus where for each chromosome in the dataset, if pairwise SNP in the subset is the same across all genotypes, the score is 5000+bp distance between the SNP pair. The score is the sum of the scores of each adjacent variant in the subset. If there are 1 to 5 differences for a pair of SNPs, a negative penalty of -10000 is given while -inf is given for more than 5. SNP tags are generated then by comparing S* scores of a window to the S* score of that region under a null S* distribution generated by population simulations filtering all tags with p<0.01. The final call-set appears to have been generated by filtering tags that have p<0.05 based on how similar identified potential introgression compared to a randomly selected area of the genome that had no introgression in reference to the archaic (Neanderthal here) genome.

    So, in this case, it looks like that the data this model used from Vernot and Akey 2014 did not take into account influx from West Eurasians into West Africans possibly explaining smaller EUR tag numbers (but that the S* statistic can be modified to handle an assumption of backmigration). But, regardless, it seems as if even with that no introgression assumption, EAS and EUR are already roughly equal tag wise…

  25. @matt

    As an additional note, the Petr paper seems to have mentioned that the f4 statistic is especially vulnerable to violations of population assumptions but given the comments Vernot and Akey provide above, it is possible the S* statistic may not be as vulnerable to such violations.

  26. @matt
    Got flagged for a 500 but don’t see the repeat comment so I’ll try again:

    As an additional note, the Petr paper seems to have mentioned that the f4 statistic is especially vulnerable to violations of population assumptions but given the comments Vernot and Akey provide above, it is possible the S* statistic may not be as vulnerable to such violations.

Comments are closed.