A quick heads’s up* that a recent genetic study on longevity may need further work.

Last week Science published a paper reporting a genome-wide association study (GWAS) for genes for long life, which was widely covered in the media (e.g. in the Guardian). Almost immediately some scientists queried aspects of the work, as is usual in science. Being a well-publicised work, these queries have wider reach than for other research papers.

Harry Patch at 105** (Source: Wikimedia Commons)

Harry Patch at 109** (Source: Wikimedia Commons) Wikipedia writes: ’Harry Patch, known as "the Last Tommy", was a British supercentenarian, and the last surviving soldier to have fought in the trenches of the First World War.’

Newsweek has an article by Mary Carmichael that does a good job of explaining some of the issues.

Rather than repeat the main issues Carmichael points out in condensed form here, I encourage readers to read the original.

Daniel MacArthur writing at Genetic Future adds that the high-scoring SNPs (single-nucleotide polymorphisms) have no near neighbours with similarly high scores, a pattern which is more typical of spurious peaks (i.e. noise rather than signal). This should invite checking if these are noise or signal, or if there is a wider problem.

I have a similar opinion, based on experience using markers in linkage analysis. Markers lie along the chromosome, one after the other. If you plot the linkage score for each marker and are looking for regions that are candidates to be linked to the disease you are studying, ideally you want table mountain-shaped peaks, with several adjacent markers having a high score, rather than Matterhorn-shaped peaks (Aspiring-shaped for New Zealanders), with only a single high-scoring marker.

Having several independent markers indicating an association gives more confidence that the association is likely to be real.

Recently I wrote about reporting on-going science… Wearing my scientist hat and speaking tongue-in-cheek, sometimes I feel as if we need an embargo on all reporting for a week or two until other scientists have an opportunity to look over new hot topics… (Peer-review notwithstanding.)

On the science side of things, Daniel MacArthur offers some suggestions that may help. In my (rather cynical) opinion, quality control has been an issue in areas involving large datasets and high-throughput data analysis in biology, especially when any one technique is new. To be clear: I’m not saying these types of projects are flawed or amateur, just that more emphasis could be placed on checking findings are sound and that these form a bigger part of the published work.

It reminds me of the need for standardised quality control methods, that can be reported along with the results that journals can ask as part of the publication of the particular type of data.

(Crystallography and modelling against density maps or other crystal structures comes to mind as I write this. Crystallography faced quality control issues many years ago and, over time, developed a collection of measures that indicate if issues are likely to be present. These are standardised and are considered an integral part of reporting a crystal structure.)

MacArthur writes that people interesting in working in this area should,

talk to someone who really knows what they’re doing when it comes to GWAS data

I’d extend that in cases, to suggesting that researchers avoid the temptation to do everything in their own lab. Some things really are better placed in the hands of people who already have the appropriate experience and expertise.

As a recent example, I had wanted to write in this blog about a genetics paper I read. It’s a straight-forward story that should be fairly easy to convey, involving a family of proteins I am familiar with.

Near the end of the paper these geneticists present some ’molecular modelling’, a different field to genetics, which is, to be polite, flawed. The elementary error in an otherwise satisfactory paper was an embarrassing reminder that some research groups still try take on things they shouldn’t. This is a ’add-on’ to the main work, but it lets the rest of the work down. A few minutes talking with someone familiar with molecular modelling would have put them straight.


* I’m not up to writing anything very original today. Some days are like that…

** In case you find the legend here hard to read, the legend from wikipedia reads: Harry Patch, known as “the Last Tommy”, was a British supercentenarian, and the last surviving soldier to have fought in the trenches of the First World War.

Other articles at Code for life:

Blogimmuniqué: who are you?

Boney lumps, linkage analysis and whole genome sequencing

External (bioinformatics) specialists: best on the grant from the onset

Describe your fantasy institute

Royal Society publishing free to read, 1665 – today