Dynamic Structure Investigation and Spectra Prediction of Peptides Using Machine Learning Techniques

DASHH Doctoral Researchers: Amir Kotobi & Carlos Ortiz-Mahecha

Supervisors: Prof. Sadia Bari (DESY), Prof. Robert Meißner (TUHH, Hereon)

Including supervised and unsupervised machine learning (ML) techniques into the simulator’s toolbox is an important approach to facilitate the understanding of peptide complexities as well as to reduce the high computational costs of quantum mechanics (QM) methods for large systems such as proteins. This project focuses on implementing supervised and unsupervised ML to understand structure-property relationship of peptides. In particular, feature engineering, clustering and dimensionality reduction of atomistic datasets are used to explore the molecular structural landscape. Moreover, graph neural networks (GNN), as a powerful and efficient way, is used to explain the interplay between the biomolecules’ conformations and spectroscopic data.

Learn more in this short video.