This special edition of the BioExcel webinar series features student speakers who were awarded poster prizes at the BioExcel Summer School 2024.

Marta Devodier

Marta is a physicist with a Master’s degree in Biophysics and Soft Matter Physics obtained from the University of Parma with honors. In November 2023, she joined the research group of Prof. Paolo Carloni at Forschungszentrum Jülich, as an Early Stage Researcher in the “Advanced computing, QuanTum algorIthms and data-driVen Approaches for science, Technology and Engineering” (AQTIVATE) project within the Horizon MSCA Doctoral Networks programme, under the supervision of Prof. Paolo Carloni and that of Prof. Giulia Rossetti. During her master, she had the chance to acquire a solid formation in computational methods, including molecular dynamics, and principal topics in advanced theoretical physics, statistical physics, biochemistry and molecular biology. Marta worked on her Master’s thesis at Forschungszentrum Jülich as an ERASMUS+ student. Here, she had the opportunity to acquire knowledge in the biophysics of membrane protein essential for brain function and to gain expertise in setting up and running QM/MM simulations using the supercomputers of the Jülich Supercomputing Centre. During her PhD, she is working on ligand/protein interaction by means of free energy calculations. Key issues of code scalability for this problem will be addressed using a recently developed method that builds on metadynamics and ML.

Title:  Effective Data-Driven Collective Variables for Free Energy Calculations from Metadynamics of Paths: the case of the human Adenosine Receptor type 2A

 The free energy landscape associated with biomolecular processes contains energetic, kinetic and structural information of great practical importance. For example, studying the (un)binding of ligands (from) to their protein targets can help the design of novel drugs. Enhanced sampling methods such as metadynamics can predict this landscape in terms of few collective variables (CVs). Their success critically depends on the ability of the CVs to capture the relevant slow degrees of freedom (that are generally unknown). The discovery of the CVs is an unmet challenge in biomolecular simulation. Recently, some of us have recently developed an iterative protocol for the construction of efficient CVs which combines machine learning-aided CV discovery with the Metadynamics of Paths (MoP) path sampling technique for the generation of high-quality datasets [1]. This combination allows for the successive exploration of transition pathways between reactant and product states, but so far it has been applied only to simplified model systems and its performance on real biological events is unknown. Scaling up the method to real-life systems is a real challenge. Here, we apply this iterative procedure to protein-ligand unbinding processes. In particular, we focus on a real-life system, the human adenosine receptor type 2A in binding to its high-affinity antagonist ZMA. The choice of this process was driven by both the availability of results from a previous enhanced sampling study obtained by our group and by its high pharmacological relevance, as this protein is a promising drug target for combating Parkinson’s disease.

The work in this project, if successful, will pave the way to a robust and accurate computational protocol for CV discovery for free energy calculations on biological systems of pharmacological relevance, which, in turn, may help elucidate key cell functions and development of novel drugs.

Sachin Shivakumar

 Sachin completed his Master’s in Physics from the National Institute of Technology Hamirpur, India. In his master’s project, he used Density Functional Theory (DFT) to simulate the structural and electronic properties of silicon and iron oxide crystals. Following this, he worked as a Junior Research Fellow at the Indian Institute of Science Education and Research Bhopal (IISER B), where he was part of the software development team for the LITESOPH project. The aim of the project was to build user-friendly applications for excited state simulations, and it was funded by the National Supercomputing Mission, India.

He is currently a PhD student working in the computational biomedicine group at Forschungszentrum Julich GmbH, under the supervision of Dr. Paolo Carloni and Dr. Erik Lindahl. And his PhD is part of the AQTIVATE training program, funded by the European Union. It is an initiative to train students in HPC, developing scalable algorithms and machine learning approaches to solve research problems in physics, engineering, and biology. His research focuses on using high-quality QM/MM simulation data to reparameterize classical biomolecular force fields, as well as training machine learning models in order to achieve accurate biomolecular simulations at a lower computational cost.

Title: Implementation of Force Matching Algorthim within the Extremely Scalable MiMiC Framework for Multiscale QM/MM Simulations

 Multiscale quantum mechanics/molecular mechanics (QM/MM) molecular dynamics (MD) simulations is a state-of-the-art technique for modeling biochemical processes. Computer aided drug discovery and design could greatly benefit from the use of QM/MM MD, particularly for predicting the kinetics and thermodynamics of ligand (un)binding (from) to their therapeutic targets. However, the application of QM/MM simulations in this field remains limited due to its high computational costs1, which limit the accessible time scales. As a consequence, cheaper, but less accurate, general empirical force fields are usually adopted. Nevertheless, data from short QM/MM MD trajectories can still be leveraged to improve the accuracy of the prediction of classical MD by ad-hoc refitting of the force fields parameters for the specific system under study. The force matching algorithm is a particularly successful algorithm for such fitting against QM/MM MD data2. This method not only improves the force fields accuracy2,3 , but it also streamlines their parametrization for any new molecule. Our research group, within a consortium of European universities, has recently developed MiMiC4, an extremely scalable, density functional theory (DFT) QM/MM MD code. The objective of this project is to integrate the force-matching algorithm within the MiMiC framework. This will allow exploiting the scalability of MiMiC to generate high quality DFT QM/MM data to fit accurate biomolecular force fields for small molecules in complex with their target. The method is being implemented and tested, working on the case study of the CAG hairpin complex composed of the ligand DB213 and the CAG RNA. The latter is implicated in cellular apoptosis and has been associated with Huntington’s disease5. Our work will contribute towards developing more accurate and efficient molecular simulation protocols for applications in drug discovery.