The computational approach to variant interpretation

Abstract Computational tools are used to obtain fast estimates of the pathogenicity of sequence variants, valuable for processing large-scale sequencing datasets in biomedical applications. However, the number of these methods is growing so fast that it is difficult to know how they work, to apply them more effectively. In this chapter, we describe the principles underlying in silico tools for pathogenicity prediction, focusing on protein sequence variants and variants that affect splicing. For the first group, we present the two main approaches in the field: those based on protein biophysics and those using machine learning principles. For variants affecting splicing, we review the evolution of computational approaches designed to predict splicing affectation and highlight challenges and future of their development. Using examples from the last 5 years, we describe the principles and performance of these tools, highlighting those issues that might be useful to interested users.