DECIPHER integrates disentangled representation learning and prototype-based cell-type deconvolution across molecular modalities
- 1Department of Systems Biomedical Sciences, School of Medicine, Jinan University, Guangzhou 510632, China
Abstract
The growing availability of single-cell resources creates new opportunities to extract cell-type-resolved information from the vast body of existing bulk omics data through cell-type deconvolution. Conventional deconvolution methods often rely on linear mixture models or specific probabilistic assumptions and can be sensitive to batch effects, whereas many deep-learning approaches are modality-specific or lack a unified end-to-end learning framework. Here, we developed DECIPHER, an end-to-end representation-learning framework for cell-type deconvolution that can be applied across multiple molecular modalities. DECIPHER learns a domain-constant representation (Zc) for deconvolution and a domain-specific representation (Zs) to model domain-associated variation. By integrating nonlinear representation learning with differentiable non-negative least-squares optimization, DECIPHER estimates cell-type proportions from Zc. Across simulated datasets, experimentally generated bulk-cell mixtures, real-world datasets, and multiple molecular modalities, DECIPHER showed robust and competitive cell-type deconvolution performance. Beyond cell-type proportion estimation, the learned Zc supported chronological age prediction across independent cohorts and prognostic stratification in lung adenocarcinoma, demonstrating that DECIPHER can transform high-dimensional bulk omics data into low-dimensional, biologically informative representations. DECIPHER thus broadens the utility of existing bulk omics resources for biological discovery and clinical research.
Related
- ImmuneLens: linking transcriptional states and TCR clonotypes through disentangled multimodal learning
- AC-Diff: Anatomy-Contrast Disentangled Diffusion for Multi-Vendor Liver MRI Harmonization
- A blastema in sea star larvae integrates wound signaling to drive regeneration-specific and developmental gene expression patterns.
- The SPL8‐YAB3/TCP19 module integrates transcriptional and protein–protein regulatory mechanisms to orchestrate anther development in Arabidopsis
- Quantitative spatial connectomics of liver reveals a partially exposed sympathetic network that integrates multiple cell types across the portal tract
BibTeX
@misc{lai2026decipher,
title = {{DECIPHER integrates disentangled representation learning and prototype-based cell-type deconvolution across molecular modalities}},
author = {W. Lai and C. Li and Q. Deng and Y. Zhu and C. Liu and Z. Li and O. J. Luo},
howpublished = {bioRxiv},
year = {2026},
month = sep,
doi = {10.64898/2026.09.24.753940},
url = {https://www.biorxiv.org/content/10.64898/2026.09.24.753940v1},
note = {CC BY-NC},
}
RIS
TY - UNPB
TI - DECIPHER integrates disentangled representation learning and prototype-based cell-type deconvolution across molecular modalities
AU - W. Lai
AU - C. Li
AU - Q. Deng
AU - Y. Zhu
AU - C. Liu
AU - Z. Li
AU - O. J. Luo
PY - 2026
DA - 2026/09/24
PB - bioRxiv
DO - 10.64898/2026.09.24.753940
UR - https://www.biorxiv.org/content/10.64898/2026.09.24.753940v1
AB - The growing availability of single-cell resources creates new opportunities to extract cell-type-resolved information from the vast body of existing bulk omics data through cell-type deconvolution. Conventional deconvolution methods often rely on linear mixture models or specific probabilistic assumptions and can be sensitive to batch effects, whereas many deep-learning approaches are modality-specific or lack a unified end-to-end learning framework. Here, we developed DECIPHER, an end-to-end representation-learning framework for cell-type deconvolution that can be applied across multiple molecular modalities. DECIPHER learns a domain-constant representation (Zc) for deconvolution and a domain-specific representation (Zs) to model domain-associated variation. By integrating nonlinear representation learning with differentiable non-negative least-squares optimization, DECIPHER estimates cell-type proportions from Zc. Across simulated datasets, experimentally generated bulk-cell mixtures, real-world datasets, and multiple molecular modalities, DECIPHER showed robust and competitive cell-type deconvolution performance. Beyond cell-type proportion estimation, the learned Zc supported chronological age prediction across independent cohorts and prognostic stratification in lung adenocarcinoma, demonstrating that DECIPHER can transform high-dimensional bulk omics data into low-dimensional, biologically informative representations. DECIPHER thus broadens the utility of existing bulk omics resources for biological discovery and clinical research.
ER -