Co-varying neighborhood analysis identifies cell populations associated with phenotypes of interest from single-cell transcriptomics

Abstract

As single-cell datasets grow in sample size, there is a critical need to characterize cell states that vary across samples and associate with sample attributes like clinical phenotypes. Current statistical approaches typically map cells to clusters then assess differences in cluster abundance. We present covarying neighborhood analysis (CNA), an unbiased method to identify associated cell populations with greater flexibility than cluster-based approaches. CNA characterizes dominant axes of variation across samples by identifying groups of small regions in transcriptional space—termed neighborhoods—that covary in abundance across samples, suggesting shared function or regulation. CNA performs statistical testing for associations between any sample-level attribute and the abundances of these covarying neighborhood groups. Simulations show that CNA enables more sensitive and accurate identification of disease-associated cell states than a cluster-based approach. When applied to published datasets, CNA captures a Notch activation signature in rheumatoid arthritis, identifies monocyte populations expanded in sepsis, and identifies a novel T-cell population associated with progression to active tuberculosis.

Cellular deconstruction of inflamed synovium defines diverse inflammatory phenotypes in rheumatoid arthritis

“…We next set out to quantify how the composition of fine-grained cell states differed between CTAPs. To accurately identify cell-states associated with individual CTAPs within each given cell type, we used co-varying neighborhood analysis (CNA) 77 . CNA tests highly granular “neighborhoods”—small groups of phenotypically similar cells—rather than larger clusters and accounts for age, sex, and cell count per sample.…”

“…For each major cell type, we used CNA 77 to associate sample-level attributes to the abundances of cell states within that cell type. CNA defines many small cell neighborhoods in the batch-corrected low-dimensional space and stores that fractional abundance of cells from each sample in each neighborhood in a neighborhood abundance matrix (NAM).…”

SOTIP: a Unified Framework for Microenvironment Modelling with Spatial Omics Data

“…Algorithms for identification of changes in response to biological insult (e.g. disease, ontogeny, or experimental perturbations) have been developed based on single-cell RNA sequencing (scRNA-seq) data [40][41][42][43] . The advances of spatial omics technologies have revealed the phenomenon that tissues at different disease state exhibit similar cell type composition but distinct spatial organizations 1,29,34 , thus urging for developing algorithms to identify differential microenvironment among different conditions.…”

A Graph Coarsening Algorithm for Compressing Representations of Single-Cell Data with Clinical or Experimental Attributes

“…MELD 7 accomplishes this by modeling the external attributes as a signal on the graph and computing a score for each cell reflecting its probability of association with each condition. To exemplify another approach, Milo 8 and CNA 9 seek to identify critical cellular neighborhoods , or groups of phenotypically-similar cells enriched across attributes.…”

“…5 Despite effective identification and characterization of immune cell-types, a current challenge is to accurately link these immune cells to external attributes of interest, such as clinical labels or experimental perturbations. [6][7][8][9] For example, it is common in translational applications to profile blood samples from patients across clinical phenotypes or disease states in order to identify the driving, stratifying cell-types. 6,10 Blood samples are also often perturbed through stimulation, 11 and cellular correlates are identified by observing functional responses to the stimulation.…”