Oops! Part of Major Ethiopia Ancient DNA Claim Was Actually An IT Glitch.

Last October, a published paper reported that there was significant Neanderthal DNA in most Africans due to Bronze Age or more recent back migration from Eurasia.

It turns out that this remarkable result was actually an IT glitch in how the ancient genome was compared to other genomes. There was significant Eurasian backmigration to Africa resulting in Mota, who preceded this migration having far less Neanderthal DNA than modern East Africans. But, it turns out that this backmigration only reached East Africa, not the rest of the African continent.

Specifically:

The results presented in the Report “Ancient Ethiopian genome reveals extensive Eurasian admixture throughout the African continent“ were affected by a bioinformatics error. A script necessary to convert the input produced by samtools v0.1.19 to be compatible with PLINK was not run when merging the ancient genome, Mota, with the contemporary populations SNP panel, leading to homozygote positions to the human reference genome being dropped as missing data (the analysis of admixture with Neanderthals and Denisovans was not affected). When those positions were included, 255,922 SNP out of 256,540 from the contemporary reference panel could be called in Mota. 

The conclusion of a large migration into East Africa from Western Eurasia, and more precisely from a source genetically close to the early Neolithic farmers, is not affected. However, the geographic extent of the genetic impact of this migration was overestimated: the Western Eurasian backflow mostly affected East Africa and only a few Sub-Saharan populations; the Yoruba and Mbuti do not show higher levels of Western Eurasian ancestry compared to Mota.

We thank Pontus Skoglund and David Reich for letting us know about this problem.

Razib has some appropriate comments:

First, scientists are humans and mistakes happen. So respect that the authors owned up to it. On the other hand, the conclusion never smelled right to many people. I was confused by it. I asked Iosif Lazaridis at ASHG. He was confused by it. I asked Pontus Skoglund. He was confused by it. . . . 

Unfortunately the result from the bioinformatics error was emphasized on the abstract, and in the press. In The New York Times[.] 

A rule of thumb in science is when you get a shocking and astonishing result, check to make sure you didn’t make some error along the sequence of analysis. That clearly did not happen here. The blame has to be distributed. Authors work with mentors and collaborators, and peer reviewers check to make sure things make sense. The idea of massive admixture across the whole of Africa just did not make sense. 

If something like this happened to me I’d probably literally throw up. This is horrible. But then again, this paper made it into Science, and Nature wrote articles like this: First ancient African genome reveals vast Eurasian migration. The error has to be corrected.

Also, despite my own tag of "bad scientists" to this post, as Razib notes, these scientists weren't actually "bad" in the usual sense of faking data or engaging in any other form of academic dishonestly.  Once or twice in my career I've made, entirely in good faith, similar vomit worthy, but subtle and non-malicious mistakes.  It is no fun, but it happens because humans make mistakes.

They were guilty of the far less culpable error of making a serious but subtle mistake, not catching it, and then believing the erroneous conclusions that flowed from the mistake even though the conclusions were paradigm changing.  Extraordinary claims require extraordinary evidence and this wasn't it.

This wasn't quite as bad as the superluminal neutrinos claim by the OPERA experiment that turned out to be due to a bit of faulty hardware in their receiver, which was a far more extraordinary claim. But, its close.

Still, it is good news and a sign of a healthy scientific establishment that the embarrassing and very public mistakes were correctly promptly upon discovery in a forthright way with humility.