Combinatorial prediction of marker panels from single‐cell transcriptomic data

Abstract

Single‐cell transcriptomic studies are identifying novel cell populations with exciting functional roles in various in vivo contexts, but identification of succinct gene marker panels for such populations remains a challenge. In this work, we introduce COMET, a computational framework for the identification of candidate marker panels consisting of one or more genes for cell populations of interest identified with single‐cell RNA‐seq data. We show that COMET outperforms other methods for the identification of single‐gene panels and enables, for the first time, prediction of multi‐gene marker panels ranked by relevance. Staining by flow cytometry assay confirmed the accuracy of COMET's predictions in identifying marker panels for cellular subtypes, at both the single‐ and multi‐gene levels, validating COMET's applicability and accuracy in predicting favorable marker panels from transcriptomic input. COMET is a general non‐parametric statistical framework and can be used as‐is on various high‐throughput datasets in addition to single‐cell RNA‐sequencing data. COMET is available for use via a web interface (http://www.cometsc.com/) or a stand‐alone software package (https://github.com/MSingerLab/COMETSC).

The yin and yang of co-inhibitory receptors: toward anti-tumor immunity without autoimmunity

“…Additional studies using novel unbiased computational methods are going to be required in the future to identify and test mediators of T-cell exhaustion in the treatment of autoimmune diseases. 124 Furthermore, a better analysis of the tissue-specific molecular and cellular pathways inducing T-cell exhaustion remains mandatory in order to improve its targeting.…”

Single-cell analyses identify circulating anti-tumor CD8 T cells and markers for their enrichment

“…To determine if better markers for isolating these cells could be identified in humans as we had done in mice, we again used COMET (Delaney et al, 2019) to evaluate surface-expressed genes (Chihara et al, 2018). For each of the six patient blood samples, COMET identified transcripts that significantly enriched for the TM component ( Fig.…”

“…Cells were classified as positive for the PD-1 transcript (Pdcd1 in mice, PDCD1 in humans) if they had any number of reads above zero. To classify mouse cells as positive for the Klrk1 (encoding NKG2D), Entpd1 (encoding CD39), and/or Cx3cr1 (encoding CX3CR1), a more stringent cut off was used for a cell to qualify as positive, determined by COMET (Delaney et al, 2019). The COMET-determined thresholds for all markers in the COMET database are provided for the mouse dataset in Table S4, and the human dataset in Table S10, and can be found in the column titled "cutoff_val".…”

“…COMET (Delaney et al, 2019) runs were conducted with version 0.1.12 the default X parameter (0.15) and with the L parameter set to the minimum of (1) 10*K and (2) 0.35*N, where K is the number of tumor matching cells and N is the total number of cells with at least one alpha and one beta chain annotated, to account for our willingness to allow for greater levels of contamination in the identified tumor-matching samples than allowed by default.…”

“…(a-b) Logistic regression showing classification of cells as TM or non-TM based on (a) all genes and (b) a pre-selected list enriched for surface-marker genes(Chihara et al, 2018).Shown are the first two principal component projections (left), ROC curves (middle), and the Recall-Precision plots (right) with 5-fold cross validation. (c) Top surface markers for identifying TM cells from non-TM cells in the blood based on COMET analysis(Delaney et al, 2019). The20 top scoring genes by q-value are shown.…”