This repository contains data and scripts to reproduce results from the following preprint:
Thea K. Schulze, Lasse M. Blaabjerg, Matteo Cagiada, Kresten Lindorff-Larsen (2025) Supervised learning of protein variant effects across large-scale mutagenesis datasets. bioRxiv 2025.04.02.646878; doi: https://doi.org/10.1101/2025.04.02.646878
We provide example scripts and data to train and validate supervised models against VAMP-seq abundance scores:
/output/models.zip: A selection of pretrained models corresponding to those reported in the preprint./output/feature_set_1: Examples of scripts to run training and validation pipeline for different versions of our model architecture./data/df_vamp.csv: All training data (VAMP-seq scores for six different proteins) and model input features (structure-based features, including ESM-IF scores).