# ProtEx Protein Function Predictions: EC Number, Gene Ontology and Pfam Predictions with BLAST Baselines (Google DeepMind)

## Facts

- Provider: Google DeepMind (Peter Shaw et al.)
- License: CC-BY-4.0 (https://github.com/google-deepmind/protex)
- Format: JSON Lines (in sampled files)
- Stored at: Google Cloud Storage, gs://protex/predictions (us)
- Access: public, readable without an account
- Category: Proteomics
- Source: https://www.biorxiv.org/content/10.1101/2024.05.30.596539v1
- Page: https://scigantic.com/archive/15f56dbe-5f4d-576f-b3dd-493be10d08a6
- Keywords: Proteomics, benchmark, JSON Lines, Google Cloud Storage

## About

Released model predictions from ProtEx, a retrieval-augmented method that predicts protein function by conditioning on similar exemplar sequences found by BLAST. The predictions/ folder holds newline-delimited JSON files for each benchmark, with both ProtEx and BLAST baseline outputs: Enzyme Commission (EC) number prediction on the ProteInfer random and clustered Swiss-Prot splits (proteinfer-random-ec-test, proteinfer-clustered-ec-test), Gene Ontology (GO) term prediction on the same splits (up to about 4 GB per file), Pfam protein family prediction (pfam_protex_dev/test_predictions), the DeepFRI EC test set and the CLEAN enzyme benchmarks NEW and Price (clean-new, clean-price), several with three ProtEx runs. Records carry the sequence accession and predicted labels in the repository's common format (sequence, accession, labels). Used to reproduce the paper's results and compare protein function annotation methods. The input datasets are at gs://brain-genomics-public/research/proteins/. License: CC BY 4.0 (materials other than software). Cite Shaw, Gurram, Belanger et al., ProtEx: A Retrieval-Augmented Approach for Protein Function Prediction, bioRxiv 2024 (doi 10.1101/2024.05.30.596539).

## Load the data

```python
import gcsfs
fs = gcsfs.GCSFileSystem(token='anon')
fs.ls('protex/predictions/', detail=False)[:10]
```

Lists the files with anonymous access. Runs on any machine, with no cloud account.

## Data structure

File types in sampled files: JSON Lines

Example file: `predictions/`

## Open it

- In a hosted notebook with the data mounted: https://scigantic.com/archive/15f56dbe-5f4d-576f-b3dd-493be10d08a6
- Full record as JSON (schema card included): https://api.scigantic.com/api/archives/15f56dbe-5f4d-576f-b3dd-493be10d08a6
- Over MCP: `get_schema_card` with id `15f56dbe-5f4d-576f-b3dd-493be10d08a6` at https://api.scigantic.com/mcp

## Related topics

- [Benchmark datasets](https://scigantic.com/topics/benchmark)
- [Proteomics datasets](https://scigantic.com/topics/proteomics)
- [JSON Lines datasets](https://scigantic.com/topics/json-lines)
