Machine Learning Approaches for Classifying Genetic Variants

I examined the feasibility of using probabilistic soft logic (PSL), a machine learning programming language, to classify variants as either pathogenic i.e. disease-causing, or benign. I implemented a proof-of-concept model of the relationship between variant frequency and pathogenicity using PSL. These results suggest that PSL could be used in the classification workflow to increase the efficiency of expert human interpreters and facilitate the automatic construction of gene or disease-specific classifiers.