Human-Protein-Language-Model-0.5

Human-Protein-Language-Model-0.5

Overview

This repository contains 356 Pfam-level protein language model checkpoints generated for the study “Does domain-specific unsupervised fine-tuning improve protein language model performance?”.

The checkpoints were derived from the ESM-2 650M model through unsupervised masked language modeling on sequences associated with human-proteome Pfam families. Protein sequences were clustered at a sequence-identity threshold of 0.5 before model fine-tuning.

Repository organization

Model checkpoints are organized by their corresponding Pfam identifiers. Each directory contains the checkpoint associated with that protein family or clan.

Intended use

These checkpoints are provided for research on protein representation learning, protein language model benchmarking, and the evaluation of domain-specific unsupervised fine-tuning.

The models are intended for research use only and have not been validated for clinical or diagnostic applications.

Usage

The checkpoints can be loaded using the fair-esm package. Usage examples and benchmarking code are available in the project repository:

https://github.com/TianBoxue-lab/DS-UFT-Benchmark

Related resources

The complete model collection is available at:

https://huggingface.co/collections/zxcyfr/ds-uft-protein-language-models-for-human-protein-families

License

This repository is distributed under the AFL-3.0 license.