Contrastive latent variable modeling with application to case-control sequencing experiments

High-throughput RNA-sequencing (RNA-seq) technologies are powerful tools for understanding cellular state. Often, it is of interest to quantify and summarize changes in cell state that occur between experimental or biological conditions. Differential expression typically assessed using univariate tests measure genewise shifts expression. However, these methods largely ignore transcriptional correlation. Furthermore, there a need identify the low-dimensional structure gene shift collections genes change Here, we propose contrastive latent variable models designed count data create richer portrait differential sequencing data. These disentangle sources variation different conditions context an explicit model at baseline. Moreover, develop model-based hypothesis testing framework can test global subset-specific We evaluate our through extensive simulations analyses with count-based from perturbation observational experiments. find effectively complex case-control