IT之家 reports Tsinghua team’s ProtCell single-cell proteomics foundation model and its MTT S5000 reproduction

According to IT之家, a research team from Tsinghua University’s Department of Automation has unveiled ProtCell, a foundation model for single-cell proteomics, alongside the paper “ProtCell: a foundation model for single-cell proteomics.” The model was trained and evaluated on about 380 million cells from 98 public datasets. The paper says ProtCell can map cell states and infer unmeasured protein expression. It also reports a full pretraining reproduction on Moore Threads’ MTT S5000 AI card, with key metrics differing by about 0.2% from mainstream GPU platforms.