MIXALIME: MIXture models for ALlelic IMbalance Estimation in high-throughput sequencing data

TL;DR

MIXALIME is a statistical framework designed to accurately detect allelic imbalance in high-throughput sequencing data, accounting for various biological and technical biases to improve analysis sensitivity and specificity.

Contribution

It introduces a novel mixture model approach for allelic imbalance estimation that considers copy-number variations, mapping bias, and overdispersion in diverse omics datasets.

Findings

Effective in detecting allelic imbalance across multiple sequencing assays.

Accounts for biological variations like copy-number changes and aneuploidy.

Provides flexible scoring models balancing sensitivity and specificity.

Abstract

Modern high-throughput sequencing assays efficiently capture not only gene expression and different levels of gene regulation but also a multitude of genome variants. Focused analysis of alternative alleles of variable sites at homologous chromosomes of the human genome reveals allele-specific gene expression and allele-specific gene regulation by assessing allelic imbalance of read counts at individual sites. Here we formally describe an advanced statistical framework for detecting the allelic imbalance in allelic read counts at single-nucleotide variants detected in diverse omics studies (ChIP-Seq, ATAC-Seq, DNase-Seq, CAGE-Seq, and others). MIXALIME accounts for copy-number variants and aneuploidy, reference read mapping bias, and provides several scoring models to balance between sensitivity and specificity when scoring data with varying levels of experimental noise-caused overdispersion.