Particularly in structural and molecular biology and chemistry, available scientific data are rarely shared in a format that supports cooperation and reproducibility of scientific results. In many valuable existing datasets, data documentation, and re-use are not possible due to incomplete metadata and annotations [https://doi.org/10.7554/eLife.90061.1]. The goal of this thesis is to explore the experimental data present in the field of molecular dynamics and to develop an automated workflow that annotates the data. Main valuable information, which is not present in files being shared are identification of biomolecules. The basic building block of this workflow will be the mechanism behind AlphaFind (https://alphafind.fi.muni.cz/), an AI-based protein similarity search engine. As an input, the workflow will use the information on individual atoms of the molecule obtained directly from the simulated trajectory. The output will return the protein identifier and the degree of similarity along with metadata from available databases, serving as basic annotations for such molecular simulation data files.

The student will first become familiar with molecular dynamics data, their digital representations, and metadata, as well as the principles of open science. Then, the student will design, develop, and test a functional prototype of the workflow. The workflow will process simulation trajectory and/or run files and interface with the AlphaFind system (https://alphafind.fi.muni.cz/). The results will include the identification of biomolecules and links to relevant databases such as the Protein Data Bank (PDB). Additionally, on top of that, the student will get familiar with the search techniques based on metadata ontology or data itself - using trajectory embedding as base for more advanced search.

The outcomes of this workflow will be shared as open-source code, data, and potentially a draft of a scientific publication published on a pre-print server.