Automated detection and localization of genome inversions using principal component analysis

Abstract

Inversions occur when sections of a chromosome (DNA molecule) are completely reversed end-to-end. Large inversions (multiple megabases in length) can be detected, localized, and genotyped using principal component analysis (PCA) of single nucleotide polymorphisms (SNPs). However, detection and localization tasks are performed and interpreted manually. We propose a novel pipeline for the detection and localization tasks in an automated manner. We compare our results with manual analysis for localization and show that our algorithm can achieve a similarity score of 0.95 on average. For the classification task, we achieve an accuracy of 0.88 as compared to manual classification. Our results suggest that our proposed methods are fast and accurate for these tasks and can be used as tools for detection and localization.