Comparative Study of Machine Learning Models to Classify Gene Variants of ClinVar

摘要

Single-nucleotide polymorphisms (SNPs) are variants that occur in every person's genome. SNP is basically a difference in nucleotide replaced with another nucleotide SNP that is acting as biomarkers in disease prediction by locating genes that are associated. ClinVar is a public archive of reports with human variants, which are SNPs with supporting evidence. Variants reported by laboratories with different classifications like pathogenic, benign, and uncertain significance for the same variant in a gene may cause confusion when clinicians or researchers try to interpret the impact on the disease of a given patient. It is important to classify which SNPs reported have conflicting classifications and which are not. This paper presents the comparative performance analysis of machine learning algorithms that are applied to classify gene variants on a dataset of 65,188 records taken from Kaggle. We found and concluded that the neural networks model shows some bias in classifying the gene variants. In this scenario, the random forest classifier performs well in classifying gene variants by satisfying all the accuracy parameters.