Enter Note Done


Keywords: Bioinformatics; Variant calling11; Exome sequencing; Clinical testing; Precision medicine

K.W. Gripp 1,2; H.W. Moore 3; G. Nadav 1; Y. Hanani 1; Y. Gurovich 1; T. Hsieh 4; P.M. Krawitz 1,4; S.A. Skinner 3; M.J. Friez 3; R.J. Louie 3; J.R. Jones 3; S. Savage 1

1) FDNA Inc., Boston, MA; 2) Dept Gen, DuPont Hosp, Wilmington, Delaware.; 3) Greenwood Genetic Center, Greenwood, SC; 4) Institute of Genomic Statistics and Bioinformatics, University of Bonn, Bonn, Germany

Background: The expanding use of next-generation sequencing technologies will increase the challenges in interpretation and prioritization of resulting genomic variants. Variant prioritization and clinical correlation require the integration of rich phenotypic data, which can be accomplished through the use of next-generation phenotyping (NGP) technologies. The Face2Gene platform (F2G) leverages NGP analysis of various phenotypic signals, including DeepGestalt for facial analysis. The PEDIA algorithm is the first framework to integrate F2G-generated phenotypic scores along with molecular data to enable variant prioritization based on phenotypic relevance.

Methods: Using F2G and the PEDIA algorithm, we analyzed facial photos, HPO features, and VCFs from 126 real-world cases sent for exome sequencing at the Greenwood Genetic Center (GGC). Cases were segmented by diagnosis status and difficulty level. For each diagnosed case, we generated a PEDIA score for each variant and gene. Using these scores, we defined the PEDIA rank for the known causative gene in each case. This PEDIA rank was compared to the original GGC variant rank and a molecular rank based on the CADD score.

Results: Integration of NGP into the variant prioritization process dramatically increased diagnostic efficiency, as demonstrated by the increase in top-ranked causative genes and the overall improvement in gene ranking. PEDIA ranked the causative variant first in 27.5% of cases, compared to 2.5% based on GGC rank and 7.5% based on CADD rank. Compared to GGC and CADD ranks, PEDIA resulted in a significantly improved rank (≥ 15 positions) in 17.5% and 33% of cases, respectively. For cases originally solved only with analysis of parental data, PEDIA showed improved ranking in 25% of cases (average improvement 27.7 positions) and successfully placed the causative gene in the top 5 in 33%. In cases diagnosed with a syndrome supported by a facial model in F2G, PEDIA ranks the causative gene in the top 5 for 67% of cases, compared to 11% based on either GGC or CADD ranks.

Conclusion: Integrating NGP into the molecular diagnostic process increases diagnostic efficiency and yield. Further work may demonstrate this method’s effectiveness as an alternative to trio analysis. As more NGP technologies are developed, the impact of NGP on the molecular diagnostic process will increase substantially.