Computational deconvolution of drug sensitivity via single-cell–to–bulk transcriptome mapping
- 1EPFL - Ecole Polytechnique Fédérale de Lausanne, Lausanne, Switzerland
- 2SDSC - Swiss Data Science Center, Lausanne, Switzerland
- 3Oncology department, Centre Hospitalier Universitaire Vaudois (CHUV), Hematology Service and Central laboratory, Lausanne, Switzerland
Abstract
Analyses routinely conducted in both fundamental research and biomedical studies often report aggregate signals derived from heterogeneous mixtures of cells. These mixtures reflect the intrinsic anatomical and functional complexity of the tissues and organs from which they are derived. Such bulk measurements obscure cell-specific signals, thereby hindering accurate interpretation of the underlying biological processes. Here, we introduce CLIMB and CLIFF, a two-step machine-learning framework that deconvolves bulk RNA-seq into cell-type composition and expression, and then resolves bulk drug sensitivity measurements into cell subtype- and single-cell-specific profiles. We first validate CLIMB using in-vitro mixtures of leukemic cell lines with experimentally known cell-type proportions, then extend this validation to large-scale, cross-dataset pseudo-bulk analyses built from independent scRNA-seq references across two disease contexts and five cancer types — mimicking the reference-to-target mismatches encountered in real applications — and show that CLIMB accurately infers the cellular composition of bulk samples, outperforming state-of-the-art bulk deconvolution methods. We emphasize CLIMB’s effectiveness in acute myeloid leukemia, where it offers highly accurate estimation of leukemic stem cell fractions across three test datasets and three patient cohorts. We then validate CLIFF, our algorithm for estimating cell subtype- and single-cell-specific drug sensitivity, first on the same in-vitro cell line mixtures using experimentally measured drug responses, and then through in-depth, realistic-scale pseudo-bulk simulations. We applied our deconvolution strategy to the beatAML cohort dataset, currently the most extensive database of ex-vivo drug screening data. This analysis generated a cell subtype–resolved drug sensitivity landscape for 121 compounds, revealing multiple compounds putatively active against leukemic stem cells, including midostaurin, A-674563, SNS-032, and foretinib. Collectively, our work introduces a general computational framework for resolving bulk functional measurements at cell-type and single-cell resolution, providing an attractive new computational tool for drug development and precision medicine. Author summary: Cells can be taken from patients and used to directly test panels of drugs, offering personalized guidance for therapeutic decisions based on the patient’s own cells. These samples are often composed of mixtures of cell types, so the observed signal reflects only the average response over all cells, making it difficult to determine which populations are sensitive or resistant to treatment. Although new technologies can assess RNA expression in individual cells and reveal distinct populations, linking this information to drug response measurements remains challenging. In this study, we introduce a computational approach that helps bridge this gap. By combining gene expression data with drug sensitivity measurements, our method estimates both the cellular composition of samples and how different cell populations respond to drugs. We demonstrate that this approach accurately identifies cell-type composition across multiple cancers and diseases. Applying our framework to acute myeloid leukemia, we analyzed the largest available dataset of patient drug screening experiments and uncovered novel drug response patterns, including compounds predicted to target leukemic stem cells, a cancer cell type capable of initiating disease relapse. This work provides a way to dissect drug response data at the level of individual cell types, revealing hidden patterns and potentially enabling more precise, targeted treatment strategies.
Related
- AI and computational methods for reconstructing cell-state-specific gene regulatory networks from single-cell transcriptome and chromatin accessibility data in cancer: methods, applications, and perspectives
- Integrative discovery and functional dependency mapping uncover essential chimeric RNAs in the active transcriptome
- A single-cell transcriptomic method for the deconvolution of multi-donor blood mixtures.
- Integrated transcriptome and single-cell RNA sequencing identifies transient receptor potential channel-related biomarkers in ulcerative colitis
- The scPagwas analysis of single-cell sequencing combined with Bayesian deconvolution with SMR Reveals Tumor Heterogeneity and Predicts Immunotherapy Response in Neuroblastoma.
BibTeX
@article{coudray2026computational,
title = {{Computational deconvolution of drug sensitivity via single-cell–to–bulk transcriptome mapping}},
author = {Alexandre Coudray and Romain Forey and Benjamin Bejar Haro and Filipe Martins and Joana Carlevaro-Fita and Shaoline Sheppard and Sandra Eloise Offner and Gioele La Manno and Guillaume Obozinski and Didier Trono},
journal = {PLOS Computational Biology},
year = {2026},
month = oct,
doi = {10.1371/journal.pcbi.1014862},
url = {https://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1014862},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Computational deconvolution of drug sensitivity via single-cell–to–bulk transcriptome mapping
AU - Alexandre Coudray
AU - Romain Forey
AU - Benjamin Bejar Haro
AU - Filipe Martins
AU - Joana Carlevaro-Fita
AU - Shaoline Sheppard
AU - Sandra Eloise Offner
AU - Gioele La Manno
AU - Guillaume Obozinski
AU - Didier Trono
PY - 2026
DA - 2026/10/16
JO - PLOS Computational Biology
DO - 10.1371/journal.pcbi.1014862
UR - https://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1014862
AB - Analyses routinely conducted in both fundamental research and biomedical studies often report aggregate signals derived from heterogeneous mixtures of cells. These mixtures reflect the intrinsic anatomical and functional complexity of the tissues and organs from which they are derived. Such bulk measurements obscure cell-specific signals, thereby hindering accurate interpretation of the underlying biological processes. Here, we introduce CLIMB and CLIFF, a two-step machine-learning framework that deconvolves bulk RNA-seq into cell-type composition and expression, and then resolves bulk drug sensitivity measurements into cell subtype- and single-cell-specific profiles. We first validate CLIMB using in-vitro mixtures of leukemic cell lines with experimentally known cell-type proportions, then extend this validation to large-scale, cross-dataset pseudo-bulk analyses built from independent scRNA-seq references across two disease contexts and five cancer types — mimicking the reference-to-target mismatches encountered in real applications — and show that CLIMB accurately infers the cellular composition of bulk samples, outperforming state-of-the-art bulk deconvolution methods. We emphasize CLIMB’s effectiveness in acute myeloid leukemia, where it offers highly accurate estimation of leukemic stem cell fractions across three test datasets and three patient cohorts. We then validate CLIFF, our algorithm for estimating cell subtype- and single-cell-specific drug sensitivity, first on the same in-vitro cell line mixtures using experimentally measured drug responses, and then through in-depth, realistic-scale pseudo-bulk simulations. We applied our deconvolution strategy to the beatAML cohort dataset, currently the most extensive database of ex-vivo drug screening data. This analysis generated a cell subtype–resolved drug sensitivity landscape for 121 compounds, revealing multiple compounds putatively active against leukemic stem cells, including midostaurin, A-674563, SNS-032, and foretinib. Collectively, our work introduces a general computational framework for resolving bulk functional measurements at cell-type and single-cell resolution, providing an attractive new computational tool for drug development and precision medicine. Author summary: Cells can be taken from patients and used to directly test panels of drugs, offering personalized guidance for therapeutic decisions based on the patient’s own cells. These samples are often composed of mixtures of cell types, so the observed signal reflects only the average response over all cells, making it difficult to determine which populations are sensitive or resistant to treatment. Although new technologies can assess RNA expression in individual cells and reveal distinct populations, linking this information to drug response measurements remains challenging. In this study, we introduce a computational approach that helps bridge this gap. By combining gene expression data with drug sensitivity measurements, our method estimates both the cellular composition of samples and how different cell populations respond to drugs. We demonstrate that this approach accurately identifies cell-type composition across multiple cancers and diseases. Applying our framework to acute myeloid leukemia, we analyzed the largest available dataset of patient drug screening experiments and uncovered novel drug response patterns, including compounds predicted to target leukemic stem cells, a cancer cell type capable of initiating disease relapse. This work provides a way to dissect drug response data at the level of individual cell types, revealing hidden patterns and potentially enabling more precise, targeted treatment strategies.
SN - 1553-7358
SN - 1553-734X
ER -