Enter Note Done


A. M. Slavotinek1, H. Prasad1, T. Yip1, S. Rego1, P-M. Martin1, J. van Ziffle1, W. Devine1, U. Hodoglugil2, H. Hoban1, M. Kvale3; 1Univ California San Francisco, San Francisco, CA, 2Genomic Med. Lab., UCSF, San Francisco, CA, 3Inst. for Human Genetics, San Francisco, CA

The interpretation of genomic variants following exome sequencing (ES) can be aided by using HPO terms to standardize clinical features and predict gene relevance. The Pediatric arm of the P3EGS study at UCSF has performed ES on 474 probands and parents, largely from underserved populations. We studied patients who received pathogenic (P) or likely pathogenic (LP) variant results and ascertained 114 P or LP variants and corresponding genes. We used PhenoDB to manually extract HPO terms from a single clinical note and used Phen2Gene and the HPO terms to predict and rank the gene corresponding to each causative variant. Phen2Gene gene rankings were assigned to 6 rank classes, with class 1 covering ranks 1-10 and class 2 covering ranks 11-50. We found that the rank class was inversely correlated to the Phen2Gene gene score (p < 2x10-16). Phen2Gene was able to rank the causative gene into rank classes 1 or 2 in just over 25% of cases. Most of the genes in rank class 1 were associated with well-characterized phenotypes, such as Noonan syndrome. We then examined the effects of patient and gene variables to affect the Phen2Gene scores and rank classes. A linear regression analysis showed a significant association between the number of years since the gene was first published with higher Phen2Gene scores (p < 2x10-16; Bonferroni threshold p < 0.0125) and all but one of the diagnostic genes with a score >0.5 was reported 9 or more years ago. There was a significant correlation between a higher number of OMIM terms associated with the gene and a higher score (p = 0.0011) and this relationship demonstrated a threshold effect, as only one gene with <8 OMIM terms had a high score (>0.5). The number of terms with an HPO hierarchal depth greater or equal to 11 was also statistically significant (p = 2.5x10-4), implying that terms deep in the HPO hierarchy had the best chance of producing a high scoring gene. Autosomal dominant inheritance of the gene and a diagnostic category of ‘ID with MCA’ also trended towards a relationship with score but were not significant. Many causative genes were not highly ranked, possibly because of non-specific presentations, rarity of the gene, and inclusion of HPO terms distracting to the phenotype. We conclude that gene prediction using HPO terms and software tools may play a role in association with well delineated and distinctive clinical presentations, especially for established genes. The significant associations with Phen2Gene score for a higher number of OMIM terms and number of terms with HPO depth greater or equal to 11 suggests that pleiotropy, comprehensive phenotyping, and specificity of clinical terms may be important factors in successful gene prediction.
Session Type
Poster Presentations
Molecular and Cytogenetic Diagnostics