AutoGeneS: Automatic gene selection using multi-objective optimization for RNA-seq deconvolution

Abstract

Tissues are complex systems of interacting cell types. Knowing cell-type proportions in a tissue is very important to identify which cells or cell types are targeted by a disease or perturbation. When measuring such responses using RNA-seq, bulk RNA-seq masks cellular heterogeneity. Hence, several computational methods have been proposed to infer cell-type proportions from bulk RNA samples. Their performance with noisy reference profiles highly depends on the set of genes undergoing deconvolution. These genes are often selected based on prior knowledge or a single-criterion test that might not be useful to dissect closely correlated cell types. In this work, we introduce AutoGeneS, a tool that automatically extracts informative genes and reveals the cellular heterogeneity of bulk RNA samples. AutoGeneS requires no prior knowledge about marker genes and selects genes by simultaneously optimizing multiple criteria: minimizing the correlation and maximizing the distance between cell types. It can be applied to reference profiles from various sources like single-cell experiments or sorted cell populations. Results from human samples of peripheral blood illustrate that AutoGeneS outperforms other methods. Our results also highlight the impact of our approach on analyzing bulk RNA samples with noisy single-cell reference profiles and closely correlated cell types. Ground truth cell proportions analyzed by flow cytometry confirmed the accuracy of the predictions of AutoGeneS in identifying cell-type proportions. AutoGeneS is available for use via a standalone Python package (https://github.com/theislab/AutoGeneS).

Gene signatures from scRNA‐seq accurately quantify mast cells in biopsies in asthma

“…The data that support the findings of this study are openly available Jian Jiang 1,2 Alen Faiz 1,3,4 Marijn Berg 1,2 Orestes A. Carpaij 1,3 Corneel J. Vermeulen 1,3 Sharon Brouwer 1,2 Laura Hesse 1,2 Sarah A. Teichmann 5,6,7 Nick H. T. ten Hacken 1,3 Wim Timens 1,2 Maarten van den Berge 1,3 Martijin C. Nawijn 3,5…”

Circuits between infected macrophages and T cells in SARS-CoV-2 pneumonia

“…Deconvolution of bulk RNA-seq alveolar macrophage signatures was performed using AutoGeneS v1.0.3 68 and signatures derived from the integrated single-cell RNA-seq object. We used an integrated single-cell RNA-seq object containing the first six subjects included into analysis (patients 1-6) to train the AutoGeneS model.…”

Single-cell analysis of two severe COVID-19 patients reveals a monocyte-associated and tocilizumab-responding cytokine storm

“…To gather additional empirical support from COVID-19 patients, we downloaded the bulk RNA-seq data of PBMCs from three severe COVID-19 patients and three healthy controls 31 and applied AutoGeneS 32 to deconvolute the composition of the cell clusters based on the signature genes identified in our single-cell analysis. Our results indicated that there were significantly more severe stage-specific monocytes (cluster 9), plasma B cells (cluster 11), and proliferating CD8 + T cells (cluster 12) in severe COVID-19 patients than in healthy controls ( Supplementary Fig.…”

“…Deconvolution of cell clusters from bulk RNA-seq data. We applied Auto-GeneS 32 to deconvolute the composition of the cell clusters based on the signature genes identified in our single-cell analysis. Specifically, we first obtained a gene-bycluster expression matrix from our normalized single-cell profile, in which the matrix elements represented the average expression of each gene in each cell cluster.…”