Eric Fauman (he/him)

Quick quiz. This is a plot of how often each p-value appears in this @gwascatalog.bsky.social. Actually I binned the -log10(p) values to their integer values (e.g., 8-9, 9-10, etc). I'm highlighting 3 outliers here, in orange. What are these p-values and why do they diverge so much?

Years ago I asked on Twitter why my heart rate would continue to climb even after I fixed the speed and incline on my treadmill. At the time, no one on Twitter could give me a satisfactory answer. But now GPT4o was able to answer it and suggested approaches to quantify the phenomenon.

GPT4o pointed me to a device that measures core and skin body temperature. In gold you can see my core body temp continues to rise even after I reach my max heart rate (red), thus requiring further decreases in the work output (blue). Skin temperature drops during the workout, thanks to fans + sweat

Years ago I asked on Twitter why my heart rate would continue to climb even after I fixed the speed and incline on my treadmill. At the time, no one on Twitter could give me a satisfactory answer. But now GPT4o was able to answer it and suggested approaches to quantify the phenomenon.

Specifically, the phenomenon is called cardiovascular drift and it is a consequence of the body having to do the mechanical work and then also additional work to dissipate the heat generated by the effort, especially the vasodilation to direct internal body heat to the surface.

This makes no sense to me. How does genetically predicted walking speed relate to biological aging (here as telomere length( except as some sort of biomarker of overall health? www.nature.com/articles/s42...

And if that's the case, then we haven't really established a causal link between making people walk faster and seeing them age slower.

I was today years old when I realized that the Chromosome 2 fusion event which is unique to modern humans probably coincides with a possible population bottleneck event ~900kya. Could the one have caused the other? This is my adaptation of the published figure DOI: 10.1126/science.abq7487

Any population geneticist want to weigh in on whether any of this makes sense.

I was today years old when I realized that the Chromosome 2 fusion event which is unique to modern humans probably coincides with a possible population bottleneck event ~900kya. Could the one have caused the other? This is my adaptation of the published figure DOI: 10.1126/science.abq7487

In this model the effective population size reflects some kind of weighted sum of (reduced cross viability with the larger unfused population) and (higher reproductive success with the smaller fusion carriers). In absolute terms I imagine there was a "bottleneck" of one chr2 fusion carrier.

I was today years old when I realized that the Chromosome 2 fusion event which is unique to modern humans probably coincides with a possible population bottleneck event ~900kya. Could the one have caused the other? This is my adaptation of the published figure DOI: 10.1126/science.abq7487

Let's imagine a child was born 900kya with a chromosome 2 fusion. Chromosomal fusions occur today in 1:1000 live births. Beating expectations this individual was able to reproduce, nucleating a small cohort of chr2 fusion carriers molecularcytogenetics.biomedcentral.com/articles/10....

I was today years old when I realized that the Chromosome 2 fusion event which is unique to modern humans probably coincides with a possible population bottleneck event ~900kya. Could the one have caused the other? This is my adaptation of the published figure DOI: 10.1126/science.abq7487

This paper from late 2023 got a lot of attention in both the scientific and popular press based on the claim that the effective population size of humans was reduced to 1,280 individuals about 900k years ago www.science.org/doi/10.1126/... This is the authors' model:

The most famous heuristic in mapping gwas snps to genes is "it's usually the closest gene". But only slightly less well-known is this: consider the colocalized phenotypes. That is, a genetic variant seldom disrupts exactly one phenotype. What else does tugging on that thread do?

The two themes, glucose biology and hemoglobin biology, together explain 739 of the HbA1c implicated genes even though they overlap at only 117 implicated genes. I guess HbA1c really is the perfect blend of a metabolite QTL and a protein QTL!

The most famous heuristic in mapping gwas snps to genes is "it's usually the closest gene". But only slightly less well-known is this: consider the colocalized phenotypes. That is, a genetic variant seldom disrupts exactly one phenotype. What else does tugging on that thread do?

I'll give the punchline first: Clustering all the phenotypes based on the associated loci reveals two distinct sets of phenotypes The 1st includes traits like: Type 2 diabetes, drugs for t2d, glucose. L2G links 201 unique genes. These are enriched for islet biology: maayanlab.cloud/Enrichr/enri...

The most famous heuristic in mapping gwas snps to genes is "it's usually the closest gene". But only slightly less well-known is this: consider the colocalized phenotypes. That is, a genetic variant seldom disrupts exactly one phenotype. What else does tugging on that thread do?

Are we now committed to clicking on 700 loci are perusing hundreds of colocalized phenotypes? This is where we can take advantage of one of the coolest features of the @opentargets.org platform Every table has a link to the graphql api query to programmatically reproduce that table!

The most famous heuristic in mapping gwas snps to genes is "it's usually the closest gene". But only slightly less well-known is this: consider the colocalized phenotypes. That is, a genetic variant seldom disrupts exactly one phenotype. What else does tugging on that thread do?

If you click on "view" here you learn more about that specific locus Specifically we see that all the colocalized traits are related to glucose metabolism. So maybe HbA1c, the combo of hemoglobin and glucose, behaves just like glucose.

The most famous heuristic in mapping gwas snps to genes is "it's usually the closest gene". But only slightly less well-known is this: consider the colocalized phenotypes. That is, a genetic variant seldom disrupts exactly one phenotype. What else does tugging on that thread do?

I've been exploring the new @opentargets.org platform with one of my favorite phenotypes, HbA1c. I like looking at metabolite QTLs and protein QTLs, so HbA1c is the perfect all-star combo. Glucose plus hemoglobin! A metabolite and a protein! Which genetic signals dominate?

The most famous heuristic in mapping gwas snps to genes is "it's usually the closest gene". But only slightly less well-known is this: consider the colocalized phenotypes. That is, a genetic variant seldom disrupts exactly one phenotype. What else does tugging on that thread do?