Our latest: "GWAS of three molecular traits highlights core genes and pathways alongside a highly polygenic background": biorxiv.org/content/10.110…
We use urate, IGF-1, and testosterone as model molecular traits to learn general principles about the architecture of complex traits.
We use urate, IGF-1, and testosterone as model molecular traits to learn general principles about the architecture of complex traits.
Most importantly, this project is thanks to wonderful work by Nasa Sinnott-Armstrong and Sahin Naqvi @snaqvi1990 in my lab; thanks also @manuelrivascruz whose work with Nasa doing GWAS of UK Biobank biomarkers led to this project.
We picked urate, IGF-1 and testosterone as models for complex trait architecture.
All 3 traits show strong enrichment of genes involved in the relevant synthesis, transport or signaling pathways, depending on the trait. However, most h^2 comes from a huge polygenic background.
All 3 traits show strong enrichment of genes involved in the relevant synthesis, transport or signaling pathways, depending on the trait. However, most h^2 comes from a huge polygenic background.
1. Urate in the blood depends on a balance between input from tissues, and export into the urine. This is controlled by the kidneys. The top hits are near kidney transporter genes. 
Remarkably, 8 out of 10 annotated urate transporters are close to genome-wide significant signals (37-fold enrichment over background). [In this thread, when I say a gene is a hit I generally mean a genome-wide significant variant is within 100kb.]
In contrast, the biosynthesis pathway which produces urate as a byproduct shows relatively modest enrichment (2.1x, p=0.02). [Pathway on the left, hits on the right]
Vignette 2: IGF-1 is a key signaling protein linking growth hormone released from the pituitary to growth effects in peripheral tissues. The top hits are highly enriched with the IGF-1 pathway.
The upstream signaling pathway from pituitary to liver to IGF-1 balance in serum all show huge signal enrichments. Weaker enrichment downstream; presumably implicate feedback loops.
Vignette 3: Testosterone is an anabolic steroid in both sexes, and the primary male sex hormone. Like other recent papers, we see essentially no overlap between male and female hits. (The sex differences are super interesting and I'll make another thread on this later.)
Unlike for urate, there is huge enrichment of testosterone hits within the relevant biosynthetic pathway. Notably male and female hits tend to fall in different parts of the pathway, probably because the major products of the steroid pathway differ.
For males (but not females) there is strong enrichment of signals within key signaling genes in the HPG (hypothalamic-pituitary-gonadal) axis, and in SHBG which binds testosterone. [CBAT is estimated bioavailable testosterone - eg not SHBG-bound].
So for all three traits we see very clear enrichment of signals in core genes and pathways. Additionally, most of the lead hits are interpretable (a point that @Eric_Fauman is also making for molecular traits). In this regard, these traits seem far simpler than most diseases.
So we wanted to know how much of the SNP heritability comes from the core pathways, vs the polygenic background. Indeed for all 3 traits, the core genes make modest contributions (analysis by HESS)
Instead, we estimate that the bulk of the heritability comes from a large polygenic background: around 5k-10k causal variants, under conservative assumptions. (Black is data; lines are sims. Fractions out of 4.4M variants. See the paper for details. Estimate for height is 80k.)
In summary these 3 traits are relatively simpler traits than diseases. The lead variants are highly interpretable, and known core pathways show huge enrichment. But most heritability comes from a polygenic background of ~10k variants spread broadly across the genome.
These data are broadly consistent with the "omnigenic" model we put forth previously: Most heritability comes from huge numbers of trans-regulatory effects at peripheral genes that have no direct relationship to the trait in question: ncbi.nlm.nih.gov/pubmed/31051098
And lastly, most disease traits depend on inputs from many distinct biological processes, including sometimes the 3 studied here. So then they absorb the polygenicity of every trait they depend upon.
[For more about sex-differences in genetics of testosterone, see here: ]
