13073_2016_261_MOESM5_ESM.pdf (9.23 kB)

Additional file 5: Figure S5. of A visual and curatorial approach to clinical variant prioritization and disease gene discovery in genome-wide diagnostics

Download (9.23 kB)
journal contribution
posted on 02.02.2016, 05:00 by Regis James, Ian Campbell, Edward Chen, Philip Boone, Mitchell Rao, Matthew Bainbridge, James Lupski, Yaping Yang, Christine Eng, Jennifer Posey, Chad Shaw
Causal variant gene prioritization in solved clinical cohort cases and semantic algorithm score comparison. OE transitively computed a median rank of 3 (top 1 %) for host genes via maximum annotated Resnik similarity score, and 2 (top 1 %) via maximum ancestral overlap. As comparator metrics to the transitive prioritization approaches, we computed scores using direct HPO term-to-gene annotations and unions of phenotypes collapsed from the all diseases associated with each gene via the OMIM Morbidmap. The cumulative distribution curve demonstrates the quality of solutions within a given rank as the percentage of the 47 cases with variant genes correctly ranked at a given threshold. We report the median because it robustly separates the top half of a sample from the bottom half. The transitive maximum ancestral overlap method achieved the lowest median rank, while the transitive maximum OMIM catalog-weighted ancestral overlap method achieved the highest median rank percentile. (PDF 9 kb)


National Institute of Neurological Disorders and Stroke