Computational Prediction of Clinical Phenotypes and Causal Variants in Neurodevelopmental Disorders: An Analysis of Genetic Variants Data

Advancements in genomic sequencing technologies and bioinformatics tools have allowed the understanding of the genetic variations in organisms and their potential effects on phenotypes (traits). This paper presents an approach, employing machine learning models for the automated prediction of patient phenotypes and the identification of genetic variants, specifically focusing on causative, likely causative and contributing variants in neurodevelopmental disorders. It also observes the connection between patient phenotypes and variants, categorizing individuals based on neurodevelopmental manifestations such as intellectual disability, autism, epilepsy, microcephaly, macrocephaly, hypotonia, and ataxia. The study utilized genetic variant data from 565 patients, building upon & combining the previous works in the field. Unlike manual variant filtering and classification, as commonly used in the field for similar purposes, the aim was to contribute to the development of an automated tool. This tool streamlines the variant classification process and enhances disease classification accuracy with a systematic and data-driven approach to variant interpretation. To validate the approach, the results were shared and compared with those of previous groups that participated in the CAGI Challenge (Critical Assessment of Genome Interpretation) in 2018 and 2021, addressing the same task. This analysis provides insights into the performance of the tool in comparison to manual approaches.