Phenotype Risk Score Study of Rare Variants

This study was designed to assess the phenotypic effects of rare variants. Rare variants are difficult to study in a high-throughput manner because most cohorts are underpowered to detect associations. In order to gain power to test rare variants, we use Phenotype Risk Scores (PheRS) based on features of Mendelian diseases. PheRS is calculated using claims data from an EHR that is mapped to clinical features from OMIM's clinical descriptions of Mendelian disease. We calculated PheRS for 1,204 Mendelians diseases in a cohort of 21,701 individuals genotyped on the HumanExome BeadChip. We then tested for association between these diseases and rare variants in causal genes. A phenotype risk score is calculated as the weighted sum of features for a given disease. The features are defined as a set of phecodes (consolidated ICD codes) associated with a particular Mendelian disease as described in OMIM. The weights for each phecode are calculated as the log inverse prevalence of the phecode in our cohort. For an individual, their score equals the sum of the weights for each phecode that is present in their medical record. Our study was based on a cohort of individuals with genotype data linked to de-identified electronic health records (EHR) from Vanderbilt's BioVU resource. We hope that this method will help generate phenotype-genotype correlation data on rare variants which in turn may inform rare variant interpretations use for diagnostics. ]]> STUDY ASSOCIATED OMIM TERMS - phs001516.v1.p1We conducted our analysis on a cohort of 21,701 individuals genotyped on the HumanExome BeadChip. We created phenotypes using aggregated billing codes to calculate Phenotype Risk Scores (PheRS). A PheRS is a continuous variable designed to measure the phenotypic overlap of an individual with a Mendelian disease. We only tested PheRS for variants in genes known to cause the Mendelian disease (e.g. We only tested the PheRS for cystic fibrosis in CFTR variants). We restricted our analysis to rare variants (MAF<1%). We tested for association using linear regression assuming a dominant genetic model and controlling for age and sex.In our cohort of 21,701 individuals of European ancestry, we included only individuals who were at least 18 years of age at their last visit recorded in the EHR. We excluded all individuals with evidence of hematological malignancies, blood transfusions, or stem cell transplants prior to the date the blood sample used for genotyping was drawn, as well as patients who visited Vanderbilt University Medical Center exclusively for cancer care.]]>