iDCF: Interpretable deconvolution of cell fractions via biologically-informed deep learning using scRNA-seq data

Precise resolution of cellular heterogeneity within complex tissues is fundamental to deciphering disease etiologies from bulk transcriptomic profiles. While computational deconvolution offers a scalable alternative, current deep learning methods predominantly operate as “black boxes,” neglecting the structural constraints of biological laws. This reliance on purely data-driven feature extraction often yields biologically incoherent predictions and limited mechanistic interpretability. iDCF (Interpretable Deconvolution of Cell Fractions) is a novel framework that enforces biological topology onto deep neural networks. The iDCF architecture employs a dual-stream design, synergizing a standard deep network with a knowledge-based sparse neural network (KSNN) explicitly masked by pathway definitions and protein-protein interaction (PPI) networks. In comprehensive benchmarks, iDCF achieves top-tier performance, consistently ranking among state-of-the-art methods in accuracy and robustness. iDCF integrates the SHapley Additive exPlanations (SHAP) framework, bridging the gap between computational inference and biological intuition. The model’s decision logic is governed by established biological mechanisms rather than spurious statistical correlations, validating its reliability. Validations across clinical contexts, including Alzheimer’s disease, ovarian cancer, and diabetes, demonstrate iDCF’s ability to recover disease-relevant cellular dynamics. iDCF offers a high-performance, interpretable, and biologically grounded tool for deconvolving cell-type proportions, facilitating deeper insights into tissue heterogeneity in health and disease.Author summary: Understanding the cellular composition of tissues is essential for unraveling disease mechanisms, yet bulk RNA sequencing only captures averaged signals from mixed cell populations. Computational deconvolution can estimate cell-type proportions from bulk data, but existing deep learning methods largely function as “black boxes,” ignoring the biological structure that governs gene expression. Here, we introduce iDCF, a deep learning framework that explicitly embeds biological prior knowledge, including signaling pathways and protein-protein interaction networks, into its neural network architecture. By constraining the model’s connectivity to biologically validated interactions, iDCF ensures that its predictions are driven by genuine biological signals rather than spurious statistical correlations. Through integration of the SHAP interpretability framework, iDCF further provides transparent, gene-level explanations for each prediction, enabling researchers to verify whether model decisions align with known biology. Extensive benchmarks across peripheral blood, brain tissue from Alzheimer’s disease patients, ovarian tumors, and pancreatic islets demonstrate that iDCF achieves superior accuracy while maintaining mechanistic transparency. We believe iDCF offers a reliable and interpretable tool for dissecting cellular heterogeneity, facilitating deeper biological insights from the vast archives of bulk transcriptomic data.