scAB detects multiresolution cell states with clinical significance by integrating single-cell genomics and bulk sequencing data

Abstract

Although single-cell sequencing has provided a powerful tool to deconvolute cellular heterogeneity of diseases like cancer, extrapolating clinical significance or identifying clinically-relevant cells remains challenging. Here, we propose a novel computational method scAB, which integrates single-cell genomics data with clinically annotated bulk sequencing data via a knowledge- and graph-guided matrix factorization model. Once combined, scAB provides a coarse- and fine-grain multiresolution perspective of phenotype-associated cell states and prognostic signatures previously not visible by single-cell genomics. We use scAB to enhance live cancer single-cell RNA-seq data, identifying clinically-relevant previously unrecognized cancer and stromal cell subsets whose signatures show a stronger poor-survival association. The identified fine-grain cell subsets are associated with distinct cancer hallmarks and prognosis power. Furthermore, scAB demonstrates its utility as a biomarker identification tool, with the ability to predict immunotherapy, drug responses and survival when applied to melanoma single-cell RNA-seq datasets and glioma single-cell ATAC-seq datasets. Across multiple single-cell and bulk datasets from different cancer types, we also demonstrate the superior performance of scAB in generating prognosis signatures and survival predictions over existing models. Overall, scAB provides an efficient tool for prioritizing clinically-relevant cell subsets and predictive signatures, utilizing large publicly available databases to improve prognosis and treatments.

Detecting phenotype-specific tumor microenvironment by merging bulk and single cell expression data to spatial transcriptomics

“…Next, in quantitative evaluation, the differentially expressed genes (DEGs) were characterized as genes exhibiting significantly different expressions between phenotype-specific cells and others in accordance to previous assessment work from [21]. Based on these DEGs, the Gene Set Variation Analysis (GSVA) scores for each sample within bulk datasets were calculated to distinguishing two sample groups with distinct phenotypes.…”

“…These steps involve identifying genes that are differentially expressed between malignant and non-malignant cells and then investigating the biological pathways associated with these genes. Such methodologies have been extensively covered in literature (e.g., [20], [21]). While acknowledging the significance of these established analysis, our study concentrates on the novel downstream analysis we performed on spatial spots.…”

“…Bridging phenotypic bulk data, it should open a new avenues to predict cell / spot phenotypes for high resolution single cells and spatial spots. Recently, scRNA-seq based methods such as Scissor ( [20]) and scAB ( [21]) have advanced our understanding of cell subpopulations in relation to specific phenotypes measured on bulk samples. These methods are built dependent on the expression correlation between scRNA-seq and RNA-seq data.…”

The molecular subtypes of autoimmune diseases

“…However, the limited availability of single-cell RNA sequencing samples presents a significant challenge to advancing subtyping studies. Although several tools for integrating bulk and single-cell transcriptomes have been developed [154] , [155] , [156] , there is a shortage of matched bulk and single-cell sequencing data in ADs research, making it challenging to integrate these datasets. This scarcity of data poses challenges in transferring subtype information during the analysis process, potentially leading to information loss.…”

Identification of type 2 diabetes- and obesity-associated human β-cells using deep transfer learning

“…Now, there are multiple advanced computational approaches to identify subsets of cells related to disease status, survival, drug response, and other disease metrics. These tools can be broadly categorized as 'cell prioritization algorithms' and include: DEGAS 31 , scAB 41 , Scissor 42 , and scDEAL 43 . Scissor and scAB can assign survival or clinical information from bulk expression data to disparate scRNA-seq datasets using regression and matrix factorization, respectively.…”