Thread by @thatdnaguy

#ASHG20 DB
Common genetic variants associated with Mendelian disease severity revealed through cryptic phenotype analysis.
David Blair.

#ASHG20 DB Clinical heterogeneity is common rare Mendelian-like diseases. [ I'd go further and say that variation is rule. Just blanket variability is the rule. ]

#ASHG20 DB Looked at morbidity-dependent model for quantitative traits (MDGM). Cryptic phenotype inference (CPA).

#ASHG20 DB Model relies on 2 assumptions. Mendelian disease has to be on the extreme of a pathologic spectrum of variation. Example familial hypercholesterolemia.

#ASHG20 DB Not all Mendelian diseases are like this, i.e. little gradient of variation in the trait. Likely for many metabolic disorders. Need to differentiate from the spectrum model ( non-disease individuals have some spectrum of phenotype) vs. extreme phenotypic outliers.

#ASHG20 DB Morbidity-dependent effects can increase skew of distribution of trait without greatly affecting the mean. With linear modeling, this falls apart. With quantile modeling get P < 8x10^-32. Model choice therefore critical.

#ASHG20 DB Used UCSF clinical warehouse (n=1.2M). Tried to capture cryptic traits from their records for phenotypes with Mendelian-like disease.

#ASHG20 DB Get a matrix of symptoms derived from ICD codes. Get latent phenotypes to extract 2 scores for each patients (1 score per model). Spectrum score directly responds to cryptic phenotype. Outlier score says how much of an outlier the individual is.

#ASHG20 DB Example alpha 1 antitrypsin deficiency. Common variants (1-5% MAF). Directly genotyped in the data they were looking at. Background effects from smoking and alcohol use.

#ASHG20 DB Did quantile regression in the UKBB for alpha 1 antitrypsin deficiency. For phenotypic extreme, get A1AT locus that is a known cause. But the interquartile range pulls up additional associations.

#ASHG20 DB Used same approach in Marfan. Causal gene fibrillin (FBN1). TGF-beta signaling critical. Phenotypic extreme finds rs2177083. Interquartile range finds CDKN2A, LPA, CELSR2.

#ASHG20 DB Method requires access to large genotyped population. Looked at a subset of Marfan patients that were withheld. Severity increases with dose of minor allele for rs2177083. Doesn't affect any one symptom, but multiple Marfan-associated symptoms.

#ASHG20 DB Chromosome locus on chr 2. Also associated with BMPR2 expression. Decreases effect of TGF-beta signaling in culture models. Dose of allele inversely related to severity of Marfan syndromes, suggesting variant modifies the effect of excess TGF-B in those individuals.

#ASHG20 DB Software at
bit.ly/2HFQIcA