0
0The paper introduces expiMap, a specialized deep-learning architecture designed to improve the interpretation and integration of single-cell genomics data. By incorporating biologically informed gene programs into its structure, the model allows researchers to map new data into reference atlases using understandable categories like pathways instead of abstract variables. ExpiMap excels at identifying specific biological responses to diseases, such as COVID-19 and diabetes, and can even discover novel gene programs not previously recorded in databases. It effectively balances complex data integration with enhanced interpretability, outperforming traditional methods by maintaining a clear connection between cellular changes and known biological functions. The authors demonstrate that this approach resolves cellular heterogeneity and provides a more robust framework for analyzing how different tissues respond to medical perturbations.
References:
Lotfollahi M, Rybakov S, Hrovatin K, Hediyeh-Zadeh S, Talavera-López C, Misharin AV, Theis FJ. Biologically informed deep learning to query gene programs in single-cell atlases. Nat Cell Biol. 2023 Feb;25(2):337-350. doi: 10.1038/s41556-022-01072-x. Epub 2023 Feb 2. PMID: 36732632; PMCID: PMC9928587.
