AI in Genomics and Epigenomics

Abstract

Genetics is an important factor determining the predisposition to the development of many diseases and influencing the overall life expectancy of a person. We have already discussed the genetics bases of polygenic diseases; in this chapter we will focus on monogenic diseases. There are currently two main challenges in the genetics diagnosis of monogenic diseases. The first challenge is the detection of an individual’s genomic variants using high-throughput sequencing data. The second challenge is interpretation of the detected genetic variants, i.e. understanding its functional and clinical significance. Both tasks require analysis of large datasets, and significant advances in this area were recently achieved using ML-methods. We will start this chapter from brief introduction of current techniques allowing sequencing of individual genomes. In this part, we will highlight the limitations of current methods, and discuss AI-based approaches that allow to increase specificity and sensitivity of the sequencing data analysis. We will separately discuss challenges associated with the detection of different types of genomic variants: single nucleotide variants, large copy-number variations affecting thousands or millions of base pairs, and balanced chromosomal rearrangements, such as inversions and translocations. Following the review of current tools for genomic variants detection, we will describe state-of-the art methods of data interpretation. We will first focus on mechanisms underlying pathogenic effects of protein-coding variants and AI-based predictive tools allowing to score pathogenic effects of amino acid substitutions. Next, we will briefly explain the complexity of epigenetic mechanisms underlying epigenetic regulation and show how modern AI-based approaches allow the interpretation of genetic variants in non-coding regions of the human genome. Finally, we will highlight new trends focused on the integration of genomic data with clinical description of patient phenotype. In the last paragraph, we describe genetic and epigenetic changes associated with aging. We will show how scoring these changes using AI-based tools allow development of epigenetic clocks of aging and determine individual age-associated risks.