Deep Learning for Molecular Dynamics Simulations (AI for Science)

Institut

Professur für Multiscale Modeling of Fluid Materials

Typ

Semesterarbeit / Masterarbeit /

Inhalt

theoretisch /  

Beschreibung

AlphaFold, which recently won the Nobel Prize, successfully solved a long-standing problem in computational biology: predicting a protein’s three-dimensional structure from its amino acid sequence. However, obtaining static protein structures is only part of the story. Researchers are also interested in understanding how protein structures change over time (dynamical properties) and properties are typically studied using molecular dynamics (MD) simulations.

Traditional first-principles simulations based on quantum dynamics are highly accurate but computationally expensive, limiting their scalability. Deep learning provides a promising alternative by balancing accuracy and computational efficiency. It enables the integration of scientific domain knowledge (e.g., physical laws) with advanced neural network architectures such as Graph Neural Networks (GNNs) and Transformers.

This thesis aims to integrate state-of-the-art neural networks to efficiently perform molecular dynamics simulations. The project is under Professurship of Multiscale Modeling of Fluid Materials and offers potential collaboration opportunities with external research partners.

Objectives

• Understand and develop state-of-the-art GNNs and Transformers.
• Gain familiarity with geometric deep learning (GDL) and understand how geometric priors affect computational efficiency.
• Design and benchmark Transformer-based neural network potentials.
• Apply advanced techniques such as LoRA (Low-Rank Adaptation), RoPE (Rotary Positional Embedding), and MoE (Mixture of Experts).

Voraussetzungen

We are looking for technically strong Master’s students with a keen interest in AI/ML methods and their scientific applications.

Preferred skills and interests

• Familiarity with probability theory and statistics.
• Interest in dynamical systems, physics, chemistry, or applied mathematics.
• Strong time management and a proactive mindset.

 

Möglicher Beginn

sofort

Kontakt

Weilong Chen
weilong.chentum.de

Ausschreibung