| Description |
Successful molecular diagnosis using an exome sequence hinges on accurate association of damaging variants to the patient's phenotype. Unfortunately, many clinical scenarios (e.g., single affected or small nuclear families) have little power to confidently identify damaging alleles using sequence data alone. Today's diagnostic tools are simply underpowered for accurate diagnosis in these situations, limiting successful diagnoses. In response, clinical genetics relies on candidate-gene and variant lists to limit the search space. Despite their practical utility, these lists suffer from inherent and significant limitations. The impact of false negatives on diagnostic accuracy is considerable because candidate-genes and variants lists are assembled ad hoc, choosing alleles based upon expert knowledge. Alleles not in the list are not considered-ending hope for novel discoveries. Rational alternatives to ad hoc assemblages of candidate lists are thus badly needed. In response, I created Phevor, the Phenotype Driven Variant Ontological Re-ranking tool. Phevor works by combining knowledge resident in biomedical ontologies, like the human phenotype and gene ontologies, with the outputs of variant-interpretation tools such as SIFT, GERP+, Annovar and VAAST. Phevor can then accurately to prioritize candidates identified by third-party variant-interpretation tools in light of knowledge found in the ontologies, effectively bypassing the need for candidate-gene and variant lists. Phevor differs from tools such as Phenomizer and Exomiser, as it does not postulate a set of fixed associations between genes and phenotypes. Rather, Phevor dynamically integrates knowledge resident in multiple bio-ontologies into the prioritization process. This enables Phevor to improve diagnostic accuracy for established diseases and previously undescribed or atypical phenotypes. Inserting known disease-alleles into otherwise healthy exomes benchmarked Phevor. Using the phenotype of the known disease, and the variant interpretation tool VAAST (Variant Annotation, Analysis and Search Tool), Phevor can rank 100% of the known alleles in the top 10 and 80% as the top candidate. Phevor is currently part of the pipeline used to diagnose cases as part the Utah Genome Project. Successful diagnoses of several phenotypes have proven Phevor to be a reliable diagnostic tool that can improve the analysis of any disease-gene search. |
