SPaRTAN, a computational framework for linking cell-surface receptors to transcriptional regulators

Abstract

The identity and functions of specialized cell types are dependent on the complex interplay between signaling and transcriptional networks. Recently single-cell technologies have been developed that enable simultaneous quantitative analysis of cell-surface receptor expression with transcriptional states. To date, these datasets have not been used to systematically develop cell-context-specific maps of the interface between signaling and transcriptional regulators orchestrating cellular identity and function. We present SPaRTAN (Single-cell Proteomic and RNA based Transcription factor Activity Network), a computational method to link cell-surface receptors to transcription factors (TFs) by exploiting cellular indexing of transcriptomes and epitopes by sequencing (CITE-seq) datasets with cis-regulatory information. SPaRTAN is applied to immune cell types in the blood to predict the coupling of signaling receptors with cell context-specific TFs. Selected predictions are validated by prior knowledge and flow cytometry analyses. SPaRTAN is then used to predict the signaling coupled TF states of tumor infiltrating CD8+ T cells in malignant peritoneal and pleural mesotheliomas. SPaRTAN enhances the utility of CITE-seq datasets to uncover TF and cell-surface receptor relationships in diverse cellular states.

Isolated BAP1 Genomic Alteration in Malignant Pleural Mesothelioma Predicts Distinct Immunogenicity with Implications for Immunotherapeutic Response

“…The proportion of different immune, tumor and stromal cells in the tumor microenvironment was estimated from RNA-seq data using CIBERSORTx [ 46 ]. The CIBESORTx algorithm was run with default settings, excluding quantile normalization, for 100 permutations with our signature matrix based MPM CITE-seq data [ 47 ] from one tumor sample (GSE172155) to estimate the abundance of immune, stromal and malignant cells types ( Supplementary Materials Figure S2 ). Then, we evaluated the association of cell types with each TSG genotype using one-way ANOVA.…”

“…The following supporting information can be downloaded at: , Figure S1: Comparison of immune checkpoint gene mRNA expression levels as a function of TSG genotypes in 86 MPM samples from the TCGA cohort; Figure S2: CIBERSORTx analysis of TCGA MPM RNA-seq dataset with MPM scRNA-seq reference of ten major cell types; Table S1: Comparison of pleural mesothelioma histological subtypes as a function of TSG genotypes in 86 MPM samples from the TCGA cohort; Table S2: Comparison of pleural mesothelioma histological subtypes as a function of TSG genotypes in 61 MPM samples from the MSK-IMPACT cohort; Table S3: Multivariate cox regression analysis based on BAP1 , NF2 and CDKN2A/B status for TCGA MPM dataset; Table S4: Multivariate cox regression analysis based on BAP1 , NF2 and CDKN2A/B status for MSK-IMPACT dataset; Table S5: Genes for Pemetrexed response signature; Table S6: Genes for Palbociclib response signature; Table S7: Genes for anti-PD-1 resistance signature; Table S8: The anti-PD-1-resistant mRNA signature was used to predict the subgroups; Table S9: Candidate TF regulators (5% FDR) based on Figure 4 A. References [ 20 , 21 , 22 , 47 ] are cited in the Supplementary Materials.…”

Interpretable deep learning for chromatin-informed inference of transcriptional programs driven by somatic alterations across cancers

“…For statistical evaluation, we computed the mean Spearman correlation between predicted and measured gene expression profiles on held-out samples (see Methods). We obtained significantly better performance than a regularized bilinear regression algorithm called affinity regression (AR) 15–17 that was trained independently for each cancer type and explains gene expression across tumors in terms of SA status and presence of TF binding sites based on pan-cancer ATAC-seq atlas ( Fig. 2A ).…”

COVID-19db linkage maps of cell surface proteins and transcription factors in immune cells

“…We used the trained interaction matrix (W) to predict TF activity from the surface protein expression profile of a cell. We trained sample-specific and cell type-specific SPaRTAN models [18] and predicted cell-specific TF activities for each CITE-seq dataset [6][7][8][9][10]. We also calculated the correlation between surface protein expression and TF activities across cells.…”

“…We recently developed SPaRTAN (Single-cell Proteomic and RNA-based Transcription factor Activity Network) to mine the single-cell proteomic (scADT-seq) and corresponding scRNA-seq datasets obtained by CITE-seq. SPaRTAN links cell-specific expression of surface proteins with inferred transcription factor (TF) activities [5]. Although the cell surface phenotype of immune cells can be readily determined by flow cytometry, signaling pathways downstream of cell surface receptors/co-receptors drive changes in transcription and chromatin states.…”