Inferring cellular and molecular processes in single-cell data with non-negative matrix factorization using Python, R, and GenePattern Notebook implementations of CoGAPS

Abstract

Non-negative matrix factorization (NMF) is an unsupervised learning method well suited to high-throughput biology. Still, inferring biological processes requires additional post hoc statistics and annotation for interpretation of features learned from software packages developed for NMF implementation. Here, we aim to introduce a suite of computational tools that implement NMF and provide methods for accurate, clear biological interpretation and analysis. A generalized discussion of NMF covering its benefits, limitations, and open questions in the field is followed by three vignettes for the Bayesian NMF algorithm CoGAPS (Coordinated Gene Activity across Pattern Subsets). Each vignette will demonstrate NMF analysis to quantify cell state transitions in public domain single-cell RNA-sequencing (scRNA-seq) data of malignant epithelial cells in 25 pancreatic ductal adenocarcinoma (PDAC) tumors and 11 control samples. The first uses PyCoGAPS, our new Python interface for CoGAPS that we developed to enhance runtime of Bayesian NMF for large datasets. The second vignette steps through the same analysis using our R CoGAPS interface, and the third introduces two new cloud-based, plug-and-play options for running CoGAPS using GenePattern Notebook and Docker. By providing Python support, cloud-based computing options, and relevant example workflows, we facilitate user-friendly interpretation and implementation of NMF for single-cell analyses.

Intrinsic Gata4 expression sensitizes the aortic root to dilation in a Loeys-Dietz syndrome mouse model

“…CoGAPS 60 , 61 (v3.22), an R package that utilizes non-negative matrix factorization to uncover latent patterns of coordinated gene expression representative of shared biological functions, was used to identify transcriptional patterns associated with VSMC subpopulations, with the npatterns parameter set to 8, in scRNAseq analysis of murine aortas. ProjectR 62 (v1.2), an R package that enables integration and analysis of multiple scRNAseq data sets by identifying transcriptional patterns shared among datasets, was used to project these patterns into scRNAseq analysis of the human aortic root and ascending aorta.…”

“…Gata4-expressing VSMC2 are intrinsically "poised" towards a less-differentiated, maladaptive proinflammatory transcriptional signature. To examine the biological features of VSMC1 and VSMC2, and whether they were recapitulated in both murine and patient-derived LDS VSMCs, we used the Coordinated Gene Activity in Pattern Sets (CoGAPS) algorithm to identify latent patterns of coordinated gene expression in the Tgfbr1 M318R/+ VSMC mouse dataset 60,61 . Two patterns, transcriptional patterns 4 and 5, were found to be enriched in the VSMC2 and VSMC1 subclusters, respectively, in the Tgfbr1 M318R/+ VSMC mouse dataset (Fig.…”

Digitize your Biology! Modeling multicellular systems through interpretable cell behavior

“…Spots with greater than 70% purity of ductal cells were further annotated to assign agent types for the associated tumor and normal cells in each spot. For this annotation, we used our transfer learning method ProjectR 123 version 1.8.0 to distinguish proliferative signaling (modeled as an epithelial phenotype) from co-occurrence of EMT and inflammatory signaling (modeled as the mesenchymal phenotype) as defined in CoGAPS non-negative matrix factorization analysis of our reference scRNA-seq atlas of PDAC tumors using methods described previously 67,87,124 . To locate fibroblasts, Seurat version 4.1.0 was used to compute module scores from a pan-CAF gene signature as described previously 67 .…”

Inflammatory Signaling in Pancreatic Cancer Transfers Between a Single-cell RNA Sequencing Atlas and Co-Culture

“…Single-cell data provide opportunities to isolate subpopulations for inferring intercellular interactions and identifying patterns of cell signaling associated with individual cell types or cell stressors. To investigate the inflammatory signaling responsible for major histocompatibility class (MHC) II expression in tumor cells demonstrated in prior work 17–19 , we focused on inflammatory signaling in epithelial tumor cells learned from applying our Bayesian non-negative matrix factorization method CoGAPS 20 for discovery of gene expression patterns in this comprehensive atlas. Even with a larger cohort for analysis, in silico analyses of these data are limited in their ability to evaluate mechanism and implications of intercellular interactions.…”