Supplementary Material for the paper entitled "Scalable nonparametric clustering with unified marker gene selection for single-cell RNA-seq data" (Q8731)

This repo contain supplementary tables from the manuscript entitled: "Scalable nonparametric clustering with unified marker gene selection for single-cell RNA-seq data". Clustering is a common way to identify cell types in single-cell RNA-sequencing (scRNA-seq) data. Unfortunately, current methods (i) require users to make human-in-the-loop decisions, which adds significant runtime to bioinformatic analyses, and (ii) reuse the same data twice when testing for differentially expressed genes, which can lead to an increased number of false discoveries. In this work, we overcome these limitations with NCLUSION: a Bayesian nonparametric method that simultaneously clusters cells and selects marker genes. NCLUSION operates without user-defined heuristics to set model parameters and leverages variational expectation-maximization (EM) for posterior inference which allows it to scale well up to 1 million cells. By analyzing publicly available datasets, we illustrate that NCLUSION matches the state-of-the-art clustering performance of competing approaches, achieves improved computational efficiency, and directly enables identification of biologically relevant gene sets driving cluster definitions.