Generative pretraining from large-scale transcriptomes for single-cell deciphering

Exponential accumulation of single-cell transcriptomes poses great challenge for efficient assimilation. Here, we present an approach entitled generative pretrain-ing from transcriptomes (tGPT) for learning feature representation of transcrip-tomes. tGPT is conceptually simple in that it autoregressive models the ranking of a gene in the context of its preceding neighbors. We developed tGPT with 22.3 million single-cell transcriptomes and used four single-cell datasets to eval-utate its performance on single-cell analysis tasks. In addition, we examine its ap-plications on bulk tissues. The single-cell clusters and cell lineage trajectories derived from tGPT are highly aligned with known cell labels and states. The feature patterns of tumor bulk tissues learned by tGPT are associated with a wide range of genomic alteration events, prognosis, and treatment outcome of immunotherapy. tGPT represents a new analytical paradigm for integrating and deciphering massive amounts of transcriptome data and it will facilitate the inter-pretation and clinical translation of single-cell transcriptomes.

Automated summary

tGPT is a method for learning feature representation of transcriptomes, which efficiently handles the exponential growth of single-cell transcriptome data. It shows good performance on tasks such as single-cell analysis and tumor tissue analysis.