Thread by @Joni_Coleman

Next is Bonnie Huang discussing "Inference of fitness effects of short tandem repeat polymorphisms improves functional categorisation" #ASHG20

STRS are DNA sequences with repeated 1-6 base pair motifs. About 1.6M in human genome. Repeat number differs between individuals #ASHG20

Discussing the logic underlying the SISTR method. Challenging to determine the pathogenicity of STR mutations. STRs harder to interpret than (coding) SNVs - adding extra amino acids has less clear an effect than removing or altering them #ASHG20

SISTR is designed to assess the effects of STRs, in coding and non-coding regions. SISTR is a pop gen framework to measure negative selection at STRs. Pop common allele is set to 0. SISTR produces an S value, higher is more deleterious #ASHG20

SISTR has numerous components. STR mutation model. Generalised stepwise mutation model with two modifications - directional bias in mutation sizes (longer biased towards contraction, shorter to expansion). Length dependent mutation (longer mutates faster) #ASHG20

Fitness of each allele w is 1-s. #ASHG20

SISTR inference mehtod: observed allele freq, prior distribution on s. Sample from priors, 4 alleles at a time. Forward simulate for 50000 generations. Compare result to observed. Accept if similar. Iterate procedure. Build posterior on s #ASHG20

SISTR accurately recovers s (from simulations) down to 10-4 for most settings. Not equal across all situations [? - I missed a key part here...] #ASHG20

Applying SISTR genome-wide to the Simons Simplex Consortium identifies STRs identified as associated with neurodevelopmental disorders. Examined novel STRs in quad families (parents, proband, unaffected sib). General excess of mutations in probands, driven by rare alleles #ASHG20

Relative risk of STR mutations is enriched for high SISTR scores. SISTR can be found here: github.com/BonnieCSE/SISTR #ASHG20

[Fantastic talk from @b_b_huang - an undergrad(!) in @mgymrek lab. Clearer than most of the paid researchers at #ASHG20...!]