scFoundation: A Large-Scale Model for Single-Cell Transcriptomics

Researchers have developed scFoundation, a large-scale pretrained model designed to advance single-cell transcriptomics. With 100 million parameters covering approximately 20,000 genes, scFoundation was trained on over 50 million human single-cell transcriptomic profiles. Its asymmetric transformer-like architecture effectively captures complex gene interactions across diverse cell types and states. The model excels in various single-cell analysis tasks, including gene expression enhancement, tissue drug response prediction, single-cell drug response classification, perturbation prediction, cell type annotation, and gene module inference, demonstrating its potential as a foundational tool in biomedical research.