Read on Twitter
For my final session of the day (hopefully catch-up tomorrow morning...!), @sebatlab introduces @mehurles to discuss his and others' work on the genetics of neurodevelopmental diseases #WCPG2020
Two vignettes: the Deciphering Developmental Disorders study, and the impact of rare variation on cognitive and behavioural traits in the UK Biobank #WCPG2020
DDD: Began in 2010, leveaging 24 clinical referral centres across UK and Ireland. Worked with clincians at these centres to recruit a large cohort of children with DDs and parents. Battery of genomic assays, focus on exome sequencing #WCPG2020
Over 1500 genes already ID'd for DDs, so one aim was to feedback genetic diagnoses - can do this currently for 35-40% of diagnoses #WCPG2020
80% of patients in DDD have intellectual disability, but diverse other presentations. 75% are the only individual presenting with a DD in their family. #WCPG2020
Generally unsure at the start what genetic architecture would be found. Initially assayed known genetic disorders (mostly SNVs and indels), 35-40% diagnoses. No single diagnosis accounts for >1%. Recessive disorders are present. #WCPG2020
@mehurles highlights @ksamocha's work on mutational enrichment in different functional classes in DDD. See enrichment in some genes of excess missense and/or truncating mutations. #WCPG2020
Can infer mechanisms of novel disorders through the pattern of mutations - e.g. multiple mutations across the genes implies the mechanism is most likely to be due to a loss of function of the gene #WCPG2020
Focus particularly on burden analysis. Can compare cases and controls, or compare cases to a population genetic model determining expectations. #WCPG2020
See 42-48% of the DDD cohort has an underlying pathogenic de novo SNV or indel. Much of the missense de novos can be accounted for by loss of function (60%), rest by altered function (40%) #WCPG2020
About 50% of the cohort has a de novo, half of which is a known dominant disorder, about a quarter each loss of function and altered function disorders #WCPG2020
What is the contribution from regulatory variation? Targeted sequencing of non-coding elements in 8000 trios. Targeted intronic probes, experimentally validated enhancers, ultraconserved elements, and some heart enhancers #WCPG2020
See enrichment for highly-conserved elements active in foetal brain. From extrapolation, estimated that 1-3% would have a regulatory de novo #WCPG2020
Further work by Patrick Short (who did the regulatory work) replicates this in data from the 100,000 genomes project - 1-2.5% with a de novo non-coding SNV. Indels will increase that, but most undiagnosed patients won't have a non-coding de novo #WCPG2020
@hilsomartin has demonstrated that there are differences in causation between different groups - pattern observed in British European patients is different to that from British Pakistani patients #WCPG2020
What underlies the unidentified half? No evidence for strong polygenic component - 7% heritability in common SNPs, but distributed evenly across known and unknown causation groups. #WCPG2020
Polygenic effect clearly modulates the known effects - phenotypic variation correlated to polygenic load #WCPG2020
Expanded beyond DDD to identify what underlies the unknown causation group. Combined with GeneDx and data from Radboud gives a sample size of 31K #WCPG2020
Increased knowledge leads to increased insight. Nonsense mutations are not equally damaging - CADD score captures severity well, as does shet (which captures strong mutation selection). Genes enriched in the cohort for damaging mutations have high shet and CADD score #WCPG2020
Different types of mutations are easier to rescue than others - inframe mutations show less difference by shet than do frameshift mutations. #WCPG2020
By integrating knowledge into enrichment testing, identify 300 genes with enriched missense mutations associated with DDs. 90% already known, some disagreement over whether they are considered diagnostic genes #WCPG2020
These genes (excess missense and excess missense & truncating mutations) can be grouped to 28 novel developmental disorders. Genes are functionally similar to known diagnostic genes - not ID'ing new biology, just different disruptions to the same pathways #WCPG2020
Novel disorders are less phenotypically distinct than the known disorders. However, novel disorders only get diagnoses for ~500 more patients - still lots of unknown de novo mutations going on #WCPG2020
See a sex difference in mutational burden for autosomal mutations. Doesn't seem to be driven by X-linked disorders. Further research needed #WCPG2020
Why have we not found the other de novo disorders? Small mutational targets? Associated with prenatal loss? Incompletely penetrant? #WCPG2020
Gene discovery is far from saturation - need more samples to identify smaller diagnostic genes robustly. #WCPG2020
Haploinsufficient genes are less likely to be ranked as having a high likelihood of causing an ultrasound abnormality, suggesting known genes are enriched for those not causing prenatal loss. Need to integrate data from prenatal and postnatal studies #WCPG2020
Also evidence for lower penetrance - see decreasing enrichment in DDD associated with increasing enrichment in gnoMAD (population dataset) for protein truncating variants. Need to assess signal from inherited variants. Need to combine disease studies with pop traits #WCPG2020
Can't expect many genes with large effects, or lot of genes with small effects (wouldn't come in the methods used) - there seem to be around 500 genes with PTV losses not yet found for the reasons just discussed #WCPG2020
Switch to rare variation in @uk_biobank. Clearly there is an effect of selective constraint on association with Mendelian disorders. But most constrained genes aren't associated with a Mendelian disorder. #WCPG2020
See associations of PTVs in genes with high selective constraints with diseases, but also with population traits #WCPG202
Selection pressures - early/late prenatal lethality, pre-reproductive lethality, altered reproductive behaviours #WCPG2020
Take 500K from @uk_biobank: any evidence of PTV affecting reproductive success must be due to altered reproductive behaviours. #WCPG2020
Burden analysis of deletions and PTVs with combined shet values across genes, assuming multiplicative combination for fitness. Can see that shet burden is associated with educational attainment, household income, fluid intelligence, and with children/no children in men #WCPG2020
Why association with childlessness? Are carriers infertile, or do they not have partners? No obvious effect from removing known male infertility genes. See some attenuation when accounting for living with a partner, suggesting ability to attract a partner is relevant #WCPG2020
Discussion of @RobAPower's work on sex-differential reproductive success in psychiatric disorders. Also evidence for sex differences in mate preference. Low SES is linked to male childlessness #WCPG2020
Psychiatric disorders and lowered IQ do not seem to account for the majority of the shet-childlessness association. General phenomenon, need a more holistic explanation #WCPG2020
shet burden is a challenging phenotype - associated with not having an email address, so e.g. the mental health questionnaire is less useful, participants have lower burden. This probably biased ascertainment to UK Biobank overall as well. #WCPG2020
Focus on DDD and similar focuses on low frequency mutations. Moving into population datasets means looking at higher frequencies with less severe effects - identifying these new genetic effects means understanding better phenotype metrics for mild cognitive effects #WCPG2020
