,
17 tweets,
4 min read
Read on Twitter
The announcement that all of @uk_biobank is going to be sequenced (Whole Genome Sequenced) has been long trailed but very welcome nevertheless. All credit to Rory Collins, Mark Effingham (@uk_biobank) and Sara Marshall (@wellcome), MRC / @UKRI_News and the companies involved.
UK BioBank is currently - and will remain for some time - the world's best prospective cohort. This both due to scale (0.5 million people), due the phenotyping consistency at baseline + depth (100K Imaging, all metabolomics), NHS data linkage and the age of the project start
A key part of this success was the recruitment and engagement of the 500,000 volunteers across the UK (includes my dad! So proud!) who have consented to have their data used for research, and not expecting anything back but the knowledge is being used for all humanity's benefit
UK BioBank does very well in informing and keeping these volunteers involved (they come back from imaging for example) and having a two way dialog about the science and the use of data. Exemplary.
You can quibble about some details (ascertainment by volunteers gives biases, recall by genotype not baked in); the world needs more than UK BioBank (sub Saharan African cohorts in particular for genetics) but ... it's a great project.
The richness of the baseline / intermediate phenotypes (everyone with metabolomics! 100K Torso, head and heart MRIs! 90K OCTs!) coupled with the patient accumulation of health events (Hospital events via NHS Digital; Primary Care records) changes the landscape
It changes the landscape in epidemiology (the main driver to the sample size) as every common disease in the UK population can be associated with these baseline measurements, including full blood metabolomics. Imaging ... less powered but still amazing.
It changes the landscape in genetics as every common disease in the UK and *every intermediate phenotype* will be have a well powered common+rare variant study (rare variants ... for the most common diseases) across the entire genome.
The combination is also Mendelian Randomisation heaven - Mendelian Randomisation is a repositioning of genetics not to find the genetic influences or predictors of disease and phenotypes, but rather to understand the causal links between intermediate phenotypes + disease
And, for sure, this will trigger all sorts of serendipitous discoveries that are hard to predict now - about biology, probably about history of the UK. Things which are hard to predict. Here the unfussy, responsible data sharing of @uk_biobank taps into global collective smarts
So - is it "all" done? Human geneticists, should you down tools and just analyse UK BioBank? Other cohorts, pack up? Hell no! Sooooo many things here. First off the actual analysis I've packed into 3 tweets is insanely complex.
Just getting out "intermediate phenotypes" from MRI / OCT images is ... a real ask which needs lots more innovation. Rare variants in non coding regions needs entirely new methods. Converting primary care records to sensible multi-dimensional health phenotypes
What about other cohorts? Even in the UK one could do with... more numbers actually to drive into the rare variant spectrum on less common diseases better. Every big country will need a big enough cohort to know whether the UK result holds up in their healthcare setting
(Editorial note: human biology is remarkably universal but human healthcare systems and societies are really quite remarkably varied. One shouldn't imagine the UK findings mapping 1 to 1 into other systems, though the fundamental biology almost certainly will).
In terms of genetics, the @uk_biobank population is a pretty good sampling of the blend of Yamma / Anatolina farmer / neolithic hunter-gather that we sometimes label "European ancestries" :) - but more genetic diversity would *really* help - this leads you to sub saharan cohorts
But the *BIG* point is that biology is really complex, and even with ... UK BioBank sequenced, great intermediate phenotypes, automated MR meta-analysis ... we will still have plenty of head-scratching, what on-earth-is-going-on-here questions
To solve this we wont get this from "just" genetics - this is molecular biology and human physiology writ large. We need... gene expression, cell biologists, proteomics, tissues, organ, and there is no need to get this insight from human - any relevant model works.
