General-purpose pre-trained large cellular models for single-cell transcriptomics  

作　　者：Haiyang Bian Yixin Chen Erpai Luo Xinze Wu Minsheng Hao Lei Wei Xuegong Zhang 

机构地区：[1]MOE Key Laboratory of Bioinformatics and Bioinformatics Division of BNRIST,Department of Automation,Tsinghua University,China [2]Center for Synthetic and Systems Biology,School of Life Sciences and School of Medicine,Tsinghua University,China

出　　处：《National Science Review》2024年第11期14-20,共7页国家科学评论（英文版）

基　　金：supported in part by the National Natural Science Foundation of China(62250005);the National Key R&D Program of China(2021YFF1200900)and research funding of BNRIST,Tsinghua University.

摘　　要：The great capability of AI large lan-guage models(LLMs)pre-trained on massive natural language data has in-spired scientists to develop a few large-scale AI foundation models for single-cell transcriptomics,or large cellular models(LCMs).LCMs are first pre-trained on massive single-cell RNA-seq data in a self-supervised manner without specific de-sign for downstream tasks.Then,through transfer learning and model fine-tuning,they have demonstrated superior perfor-mance across a wide spectrum of tasks such as cell type annotation,data integra-tion,and drug-sensitivity or perturbation response prediction.The success opened a promising new route toward develop-ing AI models to grasp underlying bio-logical knowledge from massive data to a scale that cannot be handled by human analysis.This review introduces the basic principles,major technical variations,and typical applications of current LCMs,and shares our perspective on open questions and future directions of this exciting field.

关 键 词：ICS MASSIVE HANDLE 

分 类 号：R329.2[医药卫生—人体解剖和组织胚胎学]