Related Experiment Video
Updated: Apr 30, 2026
09:21
Author Spotlight: Generating Neuronal Phenotypic Profiles - A Protocol to Culture and Image Human Midbrain Dopaminergic Neurons
Published on: July 7, 2023
2.1K
Single-cell phenotype-associated subpopulation identification via transfer foundation model and statistical ensemble
Yuming Zhao1, Xiaonan Pan2, Zeyu Luo2
1College of Computer and Control Engineering, Northeast Forestry University, Harbin, 150040, China. zym@nefu.edu.cn.
BMC Biology
|April 29, 2026
Summary
We developed scPASI, a novel method integrating single-cell and bulk data to link cell subpopulations with complex phenotypes. This tool identifies disease-relevant cell groups and genes, aiding in understanding tumor biology and potential therapeutic strategies.
Area of Science:
- Genomics
- Computational Biology
- Systems Biology
Background:
- Single-cell RNA sequencing (scRNA-seq) reveals cellular heterogeneity but struggles to link cell states to phenotypes.
- Establishing direct cell-phenotype relationships is crucial for understanding complex traits and disease outcomes.
Purpose of the Study:
- To introduce scPASI, a computational framework for integrating single-cell and bulk transcriptomic data.
- To uncover phenotype-associated cell subpopulations and their relevant genes.
- To bridge the gap between cellular heterogeneity and observable phenotypes.
Main Methods:
- scPASI integrates single-cell and bulk data using a pre-trained foundation model (PFM) and a residual variational autoencoder (Res-VAE).
- Cell clustering is performed using the Leiden algorithm.
- Phenotype associations are inferred via LASSO and sparse group LASSO (SGL) regression, stratifying cells into strongly positive (SP), weakly positive (WP), strongly negative (SN), and weakly negative (WN) groups.
Main Results:
- scPASI successfully stratifies cells into four distinct phenotype-associated subpopulations.
- The method identifies phenotype-relevant genes within these subpopulations, offering insights into cellular heterogeneity and bulk phenotypes.
- Evaluations show scPASI outperforms existing methods across diverse datasets and phenotype settings (e.g., tumor status, genetic mutations, clinical prognosis).
- Signature genes from identified subpopulations effectively distinguish tumor cells, genetic alterations, and survival outcomes.
Conclusions:
- scPASI effectively links single-cell transcriptomics with phenotype information to reveal biologically meaningful cell-phenotype associations.
- The framework aids in identifying disease-relevant cell subpopulations, particularly in tumor biology.
- scPASI provides a foundation for potential therapeutic targeting strategies based on cellular heterogeneity.
