Mechanistic Interpretability of Fine-Tuned Protein Language Models for Nanobody Thermostability Prediction

# Data and Code for: [Mechanistic Interpretability of Fine-Tuned Protein Language Models for Nanobody Thermostability Prediction] This Zenodo repository contains the **source code, datasets, and pre-trained model weights** associated with the following research paper: > **Title:** [Mechanistic Interpretability of Fine-Tuned Protein Language Models for Nanobody Thermostability Prediction] > **Authors:** [Taihei Murakami, Yuki Hashidate and Yasuhiro Matsunaga] The study analyzes the **thermal stability of VHH antibodies** using **ESM-2 (Protein Language Model)** and **Sparse Autoencoders (SAEs)**.