scGPT: toward building a foundation model for single-cell multi-omics using generative AI

scGPT: toward building a foundation model for single-cell multi-omics using generative AI

Abstract

• Foundation model for single-cell
• Cell type annotation, batch integration, multi-omic integration, purturbation response prediction, GRN inference.

Main methods

Adamson. The Adamson perturbation dataset contains gene expres- sion data from the K562 leukemia cell line perturbed by Perturb-seq33. This dataset includes 87 unique one-gene perturbations by CRISPR interference, each replicated in around 100 cells.

a, Comparison between scGPT and other perturbation prediction methods. Pearson correlation between predicted and actual gene expression changes
is reported. The metric is computed for all genes and the top differentially expressed (DE) genes, respectively.