Refined preferences of prioritizers improve intelligent diagnosis for Mendelian diseases

• Published in: Scientific reports Vol. 14; no. 1; p. 2845
• Main Authors: Yuan, Xiao, Su, Jieqiong, Wang, Jing, Dai, Bing, Sun, Yanfang, Zhang, Keke, Li, Yinghua, Chuan, Jun, Tang, Chunyan, Yu, Yan, Gong, Qiang
• Format: Journal Article
• Language: English
• Published: London Nature Publishing Group UK 03.02.2024; Nature Publishing Group; Nature Portfolio
• Subjects: 631/114/794; 631/1647/48; 631/208/2489/1512; 692/4017; Developmental disabilities; Genes; Humanities and Social Sciences; multidisciplinary; Phenotypes; Science; Science (multidisciplinary)
• ISSN: 2045-2322; 2045-2322
• DOI: 10.1038/s41598-024-53461-x

Phenotype-guided gene prioritizers have proved a highly efficient approach to identifying causal genes for Mendelian diseases. In our previous study, we preliminarily evaluated the performance of ten prioritizers. However, all the selected software was run based on default settings and singleton mode. With a large-scale family dataset from Deciphering Developmental Disorders (DDD) project (N = 305) and an in-house trio cohort (N = 152), the four optimal performers in our prior study including Exomiser, PhenIX, AMELIE, and LIRCIAL were further assessed through parameter optimization and/or the utilization of trio mode. The in-depth assessment revealed high diagnostic yields of the four prioritizers with refined preferences, each alone or together: (1) 83.3–91.8% of the causal genes were presented among the first ten candidates in the final ranking lists of the four tools; (2) Over 97.7% of the causal genes were successfully captured within the top 50 by either of the four software. Exomiser did best in directly hitting the target (ranking the causal gene at the very top) while LIRICAL displayed a predominant overall detection capability. Besides, cases affected by low-penetrance and high-frequency pathogenic variants were found misjudged during the automated prioritization process. The discovery of the limitations shed light on the specific directions of future enhancement for causal-gene ranking tools.