[Tsinghua University releases ProtCell, a domestic computing card that reproduces "virtual cell" pre-training] On November 13, Jiang Rui's research group in the Department of Automation of Tsinghua University released ProtCell, a basic model for single-cell proteomics, and completed the paper "ProtCell: a foundation model for single-cell proteomics". This model brings together approximately 380 million cells and 98 public data sets for pre-training and evaluation. It can accurately map cell status and infer unmeasured protein expression, providing key base support for drug research and development and the construction of cutting-edge "virtual cells". The paper also disclosed that the team successfully completed the full-process pre-training reproduction of ProtCell on Moore Thread's flagship AI training and promotion all-in-one smart computing card MTT S5000. The difference between the main quantitative evaluation indicators and the international mainstream GPU platform is only about 0.2%, verifying the high carrying capacity of Moore Thread's full-featured GPU (Universal GPU) and MUSA's full-stack software ecosystem for cutting-edge life science basic models.