Despite recent progress in high-throughput protein characterization, protein databases remain largely unannotated, presenting ample opportunities for discovering promising biocatalysts. One of their major desired properties is high protein stability [1,2]. However, the complexity of the mechanisms governing protein stability makes its prediction still an unresolved problem. Due to the growing amount of experimental data, the recent progress in tackling this problem is based on machine learning methods [3].

This project aims to extract the available data from the literature, replicate the protocol in [3], and build upon it by using more advanced feature-generating tools and a more comprehensive range of machine learning methods. The practical part of the thesis will be based on using the Python environment and scikit-learn library.