Use of a gene-based case-control association approach in exome sequencing data to elucidate the molecular basis of a mendelian phenotype

Background Over the past 5 years, exome sequencing and whole-genome sequencing have been extensively used to identify genes underlying rare mendelian disorders. These techniques have accelerated not only discovery but also false-positive reports of causality. To address this issue, we developed a statistical inference framework that evaluates the strength of findings from such studies. This method was applied to exome sequencing data from individuals with a specific retinal dystrophy, aiming to elucidate the genetic basis of their visual impairment. Methods 28 unrelated patients and 1917 controls with no retinal disease underwent exome sequencing. Patients had a progressive retinal dystrophy phenotype with early cone photoreceptor involvement, absence of retinal flecks on fundus autofluorescence imaging, and an unknown molecular diagnosis. Genetic data from cases and controls were analysed with the same bioinformatics pipeline. A gene-based case-control association study was then performed and gene-based p values were derived. Findings The initial analysis focused on rare, presumed loss-of-function variants; the most significant binomial p value (p=2 × 10−5) was obtained for TTLL5. Closer inspection highlighted biallelic loss-of-function variants in this gene as a probable cause of the studied retinal dystrophy. A second analysis using a recessive model (presence of ≥2 rare, potentially functional variants) was then performed. The most significant binomial p value (p=1 × 10−4) was obtained for DRAM2, highlighting mutations in this gene as another likely cause of this retinal dystrophy. Validation studies identified additional mutation-positive individuals (four with TTLL5-retinopathy and five with DRAM2-retinopathy in total). Overall, a molecular diagnosis was identified in 15 of the 28 patients. Interpretation Using a phenotype-driven cluster analysis, we have identified two previously unreported disease-associated genes. The success of our method highlights the key role of precise phenotyping in enhancing the utility of genomic investigations. Additionally, we have described a robust genome-wide statistical framework for objectively assigning probability of causation to new candidate genes and variants. This approach is broadly applicable to the study of rare mendelian disorders.