Computational Mechanistic Chemistry and Drug Discovery

At our research group, we specialize in creating diverse libraries of molecules, both small and large, to explore the vast universe of chemical space. By leveraging natural products as foundational scaffolds, we enhance our ability to discover novel medicinal compounds. Our innovative methodologies combine exhaustive and combinatorial approaches, allowing us to identify compounds with promising medicinal properties for new therapeutic discoveries.

We have developed key programs, including T_SELEX, which enables the large-scale generation and folding of RNA aptamers using the Base Randomization Algorithm. This tool significantly enhances our capacity to create extensive RNA libraries efficiently. Additionally, our DerivatizeME program generates derivatives of specific compounds, transforming those that may not initially meet medicinal criteria into more viable candidates. By shifting these compounds into more promising areas of chemical space, we unlock novel bioactives and privileged scaffolds, driving advancements in drug discovery and development.

In our research, we also focus on chemical reactions at larger scales. We developed a program called AMADAR, which helps us study reactions in detail and identifies which molecules or substrates can undergo certain reactions. For instance, this program allows us to analyze a wide variety of DA reactions, focusing on their (a)synchronicity and how the reaction proceeds over time. By examining a large dataset of 2,000 potential DA cycloadducts, AMADAR uses methods like reaction force analysis to reveal the mechanisms involved.

Our research group focuses on the development of advanced artificial intelligence and machine learning methodologies for computational chemistry, molecular modeling, and AI-driven drug discovery. Our research emphasizes the design of novel neural network architectures and graph-based learning systems for molecular representation, reaction mechanism prediction, and quantum-inspired molecular learning. We explore hybrid AI frameworks that integrate molecular graphs, density surfaces, spatial mappings, transformers, reinforcement learning, and physics-informed constraints to enhance chemical prediction accuracy, mechanistic understanding, and model interpretability. A major component of our work involves AI-assisted drug discovery, including molecular property prediction, retrosynthesis, reaction pathway modeling, molecular docking workflows, and adaptive optimization systems for chemical design. In addition, we develop reinforcement learning strategies for ensemble optimization and autonomous scientific decision-making, with the broader goal of creating next-generation intelligent systems capable of learning complex chemical behavior from structural, spatial, temporal, and physicochemical information simultaneously.