Thread by @Joni_Coleman

It's day 3, and @sebatlab is outlining the spectrum of genetic influences that affect autism. Clear that although there are strong rare variant effects, these are not monogenic disorders #WCPG2022

Combining together results from common, rare, and structural variants allows a broader picture of genetic risk. See over-transmission of genetic risk at all levels from parents to affected children #WCPG2022

Can combine common and rare variants together into scores that are more predictive than either score alone. Inverse correlation - individuals with autism with high rare score tend to have lower common score [conditional on having autism - crosses liability threshold] #WCPG2022

See female protective effect for common and for rare variants - women seem to need a greater risk to cross the diagnostic threshold (camouflaging? Diagnostic bias?). Liability threshold differs by sex #WCPG2022

Phenotypic spectrum within autism reflects genetic spectrum. Genetic risk also correlated with apparently non-genetic risk - parental age effects are correlated with common and rare genetic risks, e.g. de novo mutations, rare variants, educational attainment PRS #WCPG2022

Parental age effects do not seem to be mediated by social effects in the parents (not e.g delaying having children) #WCPG2022

Rare variant effects in autism cluster strongly in neurons and associated cell types - common variant risk is more widely distributed, which makes sense given the effects are likely to act mostly on regulation, not protein structure #WCPG2022

Gene to mental health network. Collaboration to recruit large cohorts of rare disease. Focus on CNVs - more common than rare variants, bigger effect than common variants #WCPG2022

Often see mirror effects in CNV influences on relevant traits (head size etc) - deletions have an effect in one direction, duplications in the other. Can model these effects directly in organoids. See similar mirror effects in psychiatric risk #WCPG2022

Deep phenotyping doesn't necessarily need to be on large cohorts (and is hard to get on those scales) - targeted deep phenotyping of people with rare variants provides valuable information #WCPG2022

CNV burden contributes to all psychiatric traits to an extent - again, higher CNV burden is correlated with lower PRS due to liability threshold effects #WCPG2022

But, effect size of individual CNV alleles does not correlate well with PRS - see heterogeneity in polygenic effects across CNVs. Is PGC too heterogeneous to understand the confounds? #WCPG2022

Classic epistatic models in yeast have shown that having multiple positive effect mutations can lead to negative effects - could see this same thing with CNVs in some human traits as well (wouldn't see epistatic effects with small effect SNPs etc.) #WCPG2022

PRS might have different effects depending on CNV burden if there is a Goldilocks zone outside of which there are negative effects (so an increasing effect of PRS might push you into the zone or further out) #WCPG2022

See some initial empirical data from autism that fits the model, but effects are weak and evidence preliminary. But can look at more easily measured traits like BMI. Also need second order traits with U-distribution of risk - e.g. hand grip strength. #WCPG2022

Again, results are consistent with the model in places, but interactions are weakly powered - still unclear if the underlying model is correct. Liability threshold model doesn't fit individual CNV + PRS effects, but epistatic model isn't clearly correct either #WCPG2022

Ongoing work - looking to build larger cohorts to study these potential effects in autism #WCPG2022

Each rare variant is a unique blend of dimensional trait effects situated in a polygenic background #WCPG2022