A few days ago, scientist and young-earth creationist Dr. Jay Wile wrote a post on his Proslogion blog, in which he reported that Dr. Jeff Tomkins had abandoned his claim that human and chimpanzee DNA are only about 70% similar, in favor of a revised figure of 88%. But even that figure is too low, according to the man who spotted the original flaw in Dr. Tomkins’s work.
Dr. Wile reports:
More than two years ago, Dr. Jeffrey P. Tomkins, a former director of the Clemson University Genomics Institute, performed a detailed, chromosome-by-chromosome comparison of human and chimpanzee DNA using a widely-recognized computer program known as BLAST. His analysis indicated that, on average, human and chimpanzee DNA are only about 70% similar. This is far, far, below the 95-99% numbers that are commonly cited by evolutionists, so once I read the study, I wrote a summary of it. Well, Dr. Tomkins has done a new study, and it invalidates the one he did two years ago.
The new study was done because last year, a computer programmer of financial trading algorithms (Glenn Williamson) discovered a bug in the BLAST algorithm that Tomkins used. This bug caused the program to ignore certain matches that should have been identified, which led to an artificially low similarity between the two genomes.
Here is what Glenn Williamson has to say about himself:
Yeah – 36 year old, stay-at-home father of four – including triplets, ha! 🙂
I don’t have any formal qualifications in genetics, or anything biological for that matter. I have a bachelors degree in computing science (i.e. programming) from the University of Technology in Sydney. Started my career as a programmer, but transitioned into derivatives trading, which is a lot more fun…
And for what it’s worth, I believe that my paper is more of a computing science paper than a genetics paper. It’s more my area of expertise than Jeff Tomkins’ area.
Glenn Williamson’s detailed takedown of Dr. Tomkins’s 70% similarity figure can be accessed here. Dr. Tomkins claims he submitted his paper to the creationist publication, Answers Research Journal, but it was never published. Here’s an excerpt from the paper (emphasis mine – VJT):
In this paper I carefully reproduce a subset of Dr Tomkins’ results, and show clearly and unambiguously that Dr Tomkins has fallen victim to a serious bug in the software used to obtain his results. It is this bug that causes Dr Tomkins to report the erroneous figure of 70% similarity. After correcting for both the effects of this bug and some non-trivial errors in Dr Tomkins’ methodology, I report an overall similarity of 96.90% with a standard error of ±0.21%. This figure includes indels, and the result is largely in line with the secular scientific consensus.
What happened next? Dr. Wile takes up the story:
As a result, Dr. Tomkins redid his study, using the one version of BLAST that did not contain the bug. His results are shown above… The overall similarity between the human and chimpanzee genomes was 88%.
In an update at the top of his post, Dr. Wile now admits to having cold feet, even about the revised 88% figure:
Based on comments below by Glenn (who is mentioned in the article) and Aceofspades25, there are questions regarding the analysis used in Dr. Tomkins’s study, upon which this article is based. Until Dr. Tomkins addresses these questions, it is best to be skeptical of his 88% similarity figure.
So what was wrong with Dr. Tomkins’s new study? I’ll let Glenn Williamson explain (emphasis mine – VJT):
October 16, 2015 4:22 pm
As I’ve said many times, if there is a single base pair indel in the middle of a 300bp sequence, Tomkins will say this is a 50% match.
Tomkins is most certainly aware of this, yet he chose to publish it. I think that says pretty much everything.
Another commenter named Aceofspades25 has this to add (emphases mine – VJT):
October 16, 2015 2:13 pm
The other obvious thing that Thompkins hasn’t dealt with in his BLASTN analysis, I talk about here.
There are a few cases where no match will be found because this entire sequence appears de-novo in Chimpanzees as the result of a single mutation (e.g. a novel transposable element – see here) or because humans have had a large deletion which other primates don’t. Deletions like this also likely occurred in a single mutation – see here
Thompkins (sic) would count both of these as being a 0% match (or 600 effective mutations if the sequences he was searching for were 300bp each). In reality, these probably represent just 2 mutations.
I’ll let Glenn Williamson have the final word (emphases mine – VJT):
Thanks Ace, for letting me know about this post. I reiterate here a few things about my (unpublished!) paper, and about Tomkins’ new paper.
The first thing is that he uses the “ungapped” parameter in his BLAST comparisons. As I’ve written in a few other places now, using this parameter, and calculating results in the way that Tomkins does is entirely disingenuous. If you are comparing two 300bp sequences, and one of those sequences has a single indel smack bang in the middle, Tomkins counts this as the sequences being only 50% identical.
I’ve told him at least twice that he cannot use ungapped and then calculate the result in this way. He can do one of two things:
1. Use ungapped, which ignores indels and therefore he can only report the substitution rate. (emphasis mine – VJT)If he did this, he would get a result of around 98.8%.
2. Allow gaps, and – this is what he fails to mention in his paper – get a result of around 96.9%. And this is using a very conservative method of calculation as well, since it counts a 50bp indel as having the same weight as 50 individual mutations. If you counted a 50bp indel as a single event (which it probably was), then the overall result would be pushed up towards 98%, which is the figure usually thrown around anyway.
In a comment dated 14 August 2015 (at 03:44) on an article titled, Chimp and Human DNA vs “Sophisticated Nonsense”! on a blog called Marmotism, Glenn Williamson adds:
I’ve actually written a paper on Tomkins’ 70% result, and have _ATTEMPTED_ to get it published in Answers Research Journal. Obviously they are not having a bar of it – Tomkins is the sole peer-reviewer, and he is currently refusing to provide any critique of my work – he has been silent for 8 months, while the ball is in his court ..
See the paper here:
https://www.dropbox.com/sh/dm2lgg0l93sjayv/AAATnWSJdER53EYEYZvcgiwma?dl=0
It’s the two PDFs ..
Dr. Tomkins’s latest article in Answers Research Journal (October 7, 2015) acknowledges Williamson’s work in a single sentence:
As of 2013, the issue of overall genome similarity between chimpanzee and humans seemed to be about 70% based on five different reports, three of which were based on actual data analyses. However, in 2014 , a computer programmer of financial trading algorithms discovered an apparent bug in the BLASTN algorithm and notified this author of the situation (Glenn Williamson, Tibra Capital, personal communication).
So, a submitted article only counts as a “personal communication”? Perhaps Dr. Tomkins needs to be a little more up-front about giving credit where credit is due, and acknowledging his mistakes. At any rate, the ball is definitely in his court, and his latest 88% similarity figure warrants skepticism. I have to say that Dr. Tomkins’s methodology sounds rather suspicious to me.
What do readers think?