This directory contains protein-protein interaction (PPI) network data from the STRING database (v12.0) for Homo sapiens.
It also contains a sequestered namespace for raw protein language model assets under
esm3/. The ESM-3 namespace is separate from processed MANTRA embedding matrices and from the existing ESM-2-derived embedding artifacts.
It also contains a sequestered namespace for ESM-C raw model assets under
esmc/.
hf upload Wauplin/my-cool-model ./models .
https://huggingface.co/Wauplin/my-cool-model/tree/main
hf upload pvd232/HopTF ./home/machina/MANTRA/data/raw/protein .
-
9606.protein.links.v12.0.txt.gz:
- Content: The core PPI network. Contains pairs of protein IDs and their interaction scores.
- Format: Gzipped text file. Columns likely include
protein1, protein2, combined_score (and potentially sub-scores).
- Organism: Homo sapiens (Taxonomy ID 9606).
-
9606.protein.aliases.v12.0.txt.gz:
- Content: Mapping file to translate STRING protein IDs (e.g.
9606.ENSP00000xxx) to other identifiers (gene symbols, UniProt IDs, Ensembl gene IDs).
- Use Case: Essential for mapping the PPI network to the transcriptomics data (which uses Ensembl Gene IDs or Symbols).
-
esm3/:
- Content: Quarantined ESM-3 raw model download target and provenance files.
- Status:
EvolutionaryScale/esm3-sm-open-v1 snapshot is downloaded locally after gated access.
- Use Case: Raw source for future ESM-3 protein embedding generation. Not directly consumed as a MANTRA perturbation embedding.
-
esmc/:
- Content: Quarantined ESM-C 300M and 600M December 2024 raw model snapshots and provenance files.
- Status:
EvolutionaryScale/esmc-300m-2024-12 and EvolutionaryScale/esmc-600m-2024-12 are downloaded locally.
- Use Case: Raw source for ESM-C protein embedding generation. Not directly consumed as a MANTRA perturbation embedding.
