Path2Space — trained model weights (154-checkpoint ensemble + CTransPath)
收藏官方服务：
资源简介：
Trained model weights for Path2Space, the AI model that predicts spatial transcriptomics from H&E histopathology (Shulman et al., Cell 2026). This deposit contains: ctranspath.pth — pretrained CTransPath SwinTransformer feature extractor (768-dim per tile). mlp_ensemble.tar.gz — the manuscript 22 × 7 = 154 MLP regression heads (outer x inner cross-validation folds), each predicting ~14k genes from CTransPath features. genes.txt — the gene list (output dimension order) used by the ensemble. MD5SUMS.txt — checksums for integrity verification. The inference code for using these weights is at github.com/eldadshulman/path2space-companion. For non-commercial academic use. Please cite the Cell paper: doi:10.1016/j.cell.2026.04.023
提供机构：
Zenodo创建时间：
2026-05-14
