ProFormer: generalizable classification of single-cell and plasma proteomes using deep learning.

ProFormer achieves over 14-fold improved accuracy in classifying single-cell and plasma proteomes using deep learning on mass spectrometry data.

• Why it matters: Accurate, high-throughput proteomic classification is essential for biomarker discovery and patient stratification but is limited by current data interpretation methods.
• What they did: The team developed ProFormer, a transformer-based deep-learning pipeline that directly analyzes MS1-level features from peptide ions, outperforming traditional models across 14 architectures.
• The result: ProFormer enables rapid, accurate classification of cell types, disease states, and differentiation stages, facilitating advances in personalized medicine, early diagnosis, and single-cell proteomics.