Reproduce “SeqProFT: Sequence-only Protein Property Prediction with LoRA Finetuning” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Shuo Zhang, Jian K. Liu
SeqProFT adapts pretrained protein language models for sequence-based property prediction using low-rank updates and optional model-derived contact maps. It compares ESM-2-150M with LoRA and ESM-2-650M without LoRA on enzyme classification, reporting scores of 0.854 and 0.838. Contact-enhanced prediction leads on Fold-superfamily but not across tasks. For ESM-2-150M, LoRA trains 9.49% of parameters but takes 2.7 times as long.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.