️ 28: Cell-Type-Resolved TWAS — scPrediXcan Integrates Deep Learning and Single-Cell Data for Transcriptome-Wide Association Studies

️ Episode 28: Cell-Type-Resolved TWAS — scPrediXcan Integrates Deep Learning and Single-Cell Data for Transcriptome-Wide Association Studies In this episode of Base por Base, we explore the groundbreaking framework scPrediXcan introduced by Zhou et al. (2025) in Cell Genomics. This approach combines a deep learning–based model, ctPred, with single-cell RNA-seq data to predict cell-type-specific gene expression, and then linearizes those predictions into a SNP-based model, ℓ-ctPred, for use in S-PrediXcan. By leveraging only GWAS summary statistics and in silico expression references, scPrediXcan enables transcriptome-wide association studies at unprecedented cellular resolution . Highlights of the study:ctPred leverages a pretrained sequence-to-epigenomics model (Enformer) to predict cell-type-specific gene expression with high accuracy across diverse scRNA-seq datasets.The ℓ-ctPred linear model robustly approximates ctPred outputs, allowing efficient TWAS using GWAS summary statistics without individual-level data.In type 2 diabetes, scPrediXcan identifies 222 candidate genes across 108 independent LD blocks, compared with just 12 and 111 genes recovered by pseudobulk and bulk TWAS approaches.Applied to systemic lupus erythematosus, scPrediXcan nominates 129 genes across 24 LD blocks, uncovering cell-type-specific drivers such as PYCARD and ITGAM that bulk analyses miss.By integrating deep learning–based prediction with cell-type resolution, the framework reveals nuanced disease mechanisms and markedly improves gene prioritization for complex traits. Conclusion:scPrediXcan represents a major advance in genetic epidemiology by enabling large-scale, cell-type-specific TWAS using only summary statistics and single-cell-informed prediction models. It dramatically expands the set of candidate causal genes, refines our understanding of cellular mechanisms in disease, and lays the groundwork for more targeted experimental follow-up and therapeutic discovery. Reference:Zhou, Y., Adeluwa, T., Zhu, L., et al. (2025). scPrediXcan integrates deep learning methods and single-cell data into a cell-type-specific transcriptome-wide association study framework. Cell Genomics, 5, 100875. https://doi.org/10.1016/j.xgen.2025.100875 License:This episode is based on an open-access article published under the Creative Commons Attribution 4.0 International (CC BY 4.0) license – http://creativecommons.org/licenses/by/4.0/

There are no comments yet.

Please log in to write the first comment.