A tissue-aware machine learning framework enhances the mechanistic understanding and genetic diagnosis of Mendelian and rare diseases

• DOI: 10.1101/2021.02.16.430825
• Publication Date: 2021-02-18T02:27:24Z
• Authors: Eyal Simonovsky; Moran Sharon; Maya Ziv; Omry Mauer; Idan Hekselman; Juman Jubran; Ekaterina Vinogradov; Chanan M. Argov; Omer Basha; Lior Kerber; Yuval Yogev; Ayellet V. Segrè; Hae Kyung Im; ; Ohad Birk; Lior Rokach; Esti Yeger-Lotem
• Keywords: Mendelian inheritance; Candidate gene

ABSTRACT Genetic studies of Mendelian and rare diseases face the critical challenges identifying pathogenic gene variants their modes-of-action. Previous efforts rarely utilized tissue-selective manifestation these for elucidation. Here we introduce an interpretable machine learning (ML) platform that utilizes heterogeneous large-scale tissue-aware datasets human genes, rigorously, concurrently quantitatively assesses hundreds candidate mechanisms per disease. The resulting ML is applicable in gene-specific, tissue-specific, or patient-specific modes. Application to selected disease genes pinpointed lead tissue-specific manifestation. When applied jointly manifest same tissue, models revealed common known previously underappreciated factors underlie Lastly, harnessed our toward genetic diagnosis diseases. Patient-specific disease-causing from 50 patients successfully prioritized 86% cases, implying tissue-selectivity aids filtering out unlikely genes. Thus, can boost mechanistic understanding heritable A webserver supporting prioritization available at https://netbio.bgu.ac.il/trace/ .