PyroForge training data: ESM-2 embeddings, stability labels, and Meltome Tm
收藏官方服务:
资源简介:
Public training and test data for PyroForge (iGEM 2026, Team HK-Joint-School). esm2_train_public.pt: 544,365 rows, 499 parent proteins. ESM-2 650M mean-pooled embeddings plus dg, ddg, tm, trypsin_ml, and chymotrypsin_ml. Sequences are not included. esm2_test_public.pt: 21,325 rows, 54 parent proteins, split by parent so no protein is in both files. meltome_tm.npz: absolute melting temperatures from the Meltome Atlas via the FLIP mixed split. This is the only file used to train the Tm model. Column definitions / what was verified / and what could not be attributed are in DATA.md in the PyroForge source repository. The tm column inside the esm2 files is a different assay from meltome_tm.npz.
提供机构:
Zenodo创建时间:
2026-10-18



