Protein molecular dynamics trajectories contain vast amounts of data and insights into mechanisms behind protein function and interactions with the environment. Interpretation of this data is typically addressed through visual inspection or manual projection to simplified subspaces, risking oversight of critical patterns due to data complexity and human bias. Machine learning (ML) and explainable artificial intelligence (XAI) offer robust solutions to analyze large and intricate datasets, enabling the identification of patterns, relationships, and predictions, thereby aiding in overcoming these challenges. However, XAI potential for extracting biological insights from molecular dynamics has barely been explored in the field. This thesis will aim to use ML and different XAI techniques to understand which parts of a protein are critical for its dynamics.

The student will solve the following specific tasks:
1. Study the literature on molecular dynamics, ML, XAI;
2. Train ML-based predictors on the existing protein molecular dynamics data;
3. Apply and compare XAI techniques on the trained model;
4. Analyze and interpret the results.
To solve the practical part of the diploma thesis, the student will use Python, TensorFlow or PyTorch, scikit-learn, and the iNNvestigate library.

Recommended literature:
1. Montavon, G., Binder, A., Lapuschkin, S., Samek, W., Müller, K.-R. Layer-Wise Relevance Propagation An Overview. In Explainable AI: Interpreting, Explaining and Visualizing Deep Learning; Samek, W., Montavon, G., Vedaldi, A., Hansen, L. K., Müller, K.-R., Eds.; Lecture Notes in Computer Science; Springer International Publishing: Cham (2019); pp 193–209. https://doi.org/10.1007/978-3-030-28954-6_10.
2. Schenkmayerova, A., Pinto, G.P., Toul, M. et al. Engineering the protein dynamics of an ancestral luciferase. Nat Commun 12, 3616 (2021). https://doi.org/10.1038/s41467-021-23450-z
3. INNvestigate library: https://innvestigate.readthedocs.io/en/latest/, https://github.com/albermax/innvestigate