Protein mutational effect prediction with machine learning and computational function validation.
Machine learning-based protein variant effect prediction is a powerful tool for improving functional proteins such as enzymes and flourescent proteins. However, a large amount of experimental data is required to train machine learning models, which makes predictions difficult when experimental data is limited.
To solve it, we augment the training data by incorporating computational values obtained through molecular simulation and protein language model with experimental data aiming to improve prediction accuracy of ML model.