Mchaourab Lab: Towards Enabling AI for Spectroscopy and Protein Folding
Interested in biomedicine, protein folding, protein dynamics, biological function, or spectroscopy? Want to learn more about the intersection of these topics with artificial intelligence (AI), deep learning (DL), computation, and data science? This may be the opportunity for you!
The Mchaourab Lab investigates mechanisms of protein folding, and currently welcomes skilled students at all levels and from all disciplines. Positions are currently open for Summer 2022, and new students are always welcome to join in the Fall. Learn more about the projects below!
Dynamic Structures of Protein Folding
First, we are interested in applying the deep learning algorithm AlphaFold2 for predicting the structures of dynamic proteins. This means finding ways of “hacking” the algorithm as it was designed to predict static structures. We founds several avenues for manipulating the inputs of the program, for example by reducing the amount of information provided to AlphaFold2, or by artificially modifying the sequence of interest. We have published a paper on this topic and have another one in review.
Leveraging Transformers for Spectroscopy
Second, we are developing deep learning approaches to analyze spectroscopy data. The backbone of our method is the Transformer, an architecture developed for natural language processing tasks that has recently been extended to other domains such as computer vision. The primary advantage of this approach is its flexibility, namely its ability to analyze sequences of experimental data that can vary in duration, noise level, and composition. Our research indicates that this approach has performance that is competitive with state-of-the-art published methods. The proof of concept on this project has been established and there could be a relatively easy path to a first publication.
Students from all disciplines and academic levels are invited to reach out; students with experience programming in Python, knowledge of DL/transformers, and AI/ML are preferred. For more information, please email Dr. Hassane Mchaourab, Louise B. McGavock Chair of the VU Department of Molecular Physiology. To learn more about the work of his lab, visit the Mchaourab Lab webpage!