Correlation-based gene set tests in scRNAseq data

Usual ways to measure gene coexpression or measure dependence of two genes in sparse single-cell RNA<br />sequencing data would be to compute spearman correlation or other non-parametric tests. While these tests<br />can gauge if genes are linearly or monotonically related, they cannot assess whether genes are non-linearly<br />related. Even if these tests could assess there to be some form of dependence, these methods have trouble<br />explaining why. MaxBET is a novel non-parametric method developed by Professor Kai Zhang that tests for dependence.<br />It is efficent at finding non-linear relationships and most importantly, explaining why these relationships<br />exist. This new method will be applied to scRNAseq data to find new non-linear relationships that exist<br />between genes. We will also see how relationships between pairs of genes, gene coexpression, change in cells<br />that do have HIV, and cells that do not have HIV. Another method, AUGUST, will be applied as a two<br />sample test to see if the distributions of a single gene change in healthy cells compared to cells with HIV. The<br />ultimate goal of using MaxBET and AUGUST is to measure change in relationships of gene expression in<br />cells that do and do not have HIV. Our results find many relationships and relationship changes. Specifically,<br />we see these in genes such as the HLA family, that are heavily related with immune system response.&nbsp;