Machine learning based classification of protein states from lipid fingerprints

This project is dedicated to unveiling how proteins within cell membranes adapt to their surroundings, particularly the lipid environment. Employing computer simulations and machine learning (ML), we focus on RAS signaling proteins and Mga2 transcription factors. RAS proteins, crucial for cell growth and division, are anchored to the cellular membrane and often exhibit mutations in cancer patients. Conversely, the Mga2 dimer, a transmembrane protein, plays a role in regulating the synthesis of unsaturated fatty acids. Through simulations of one of the proteins, RAS-RAF complex or Mga2, in diverse lipid environments, we collect data to train ML models, aiming to predict distinct protein states in varied lipid mixtures. Additionally, we explore if different protein states trigger shifts in the lipid environment. At the Centre for Molecular Simulations at the University of Calgary, our expertise lies in leveraging computer simulations to investigate lipid-protein interactions. Integrating ML with our studies opens new avenues for exploring lipid redistribution in cellular membranes and its impact on protein dynamics. Simultaneously, the Lawrence Livermore National Laboratory will benefit from an extended protocol for more complex systems, expanding its application to other proteins and using other lipid parameters.

Faculty Supervisor:

Peter Tieleman

Student:

Partner:

University of Utah

Discipline:

Life Sciences

Sector:

Artificial Intelligence; Life Sciences (not health); Pharmaceuticals

University:

University of Calgary

Program: