Learn about a groundbreaking research presentation that introduces PepTune, an innovative multi-objective discrete diffusion model designed for therapeutic peptide generation. Explore how this advanced framework addresses the complex challenges of peptide therapeutics development, particularly in optimizing multiple properties simultaneously like target binding affinity, solubility, and membrane permeability. Discover the technical implementation of the Masked Discrete Language Model (MDLM) framework and how it ensures valid peptide structures through state-dependent masking schedules and penalty-based objectives. Understand the novel Monte Carlo Tree Search (MCTS)-based strategy that guides the diffusion process, balancing exploration and exploitation for optimal sequence refinement. Examine real-world applications in generating diverse, chemically-modified peptides optimized for various therapeutic properties across different disease-relevant targets, including treatments for diabetes and cancer. Gain insights into how this powerful and modular approach revolutionizes multi-objective sequence design in discrete state spaces, potentially transforming the future of drug discovery and development.
PepTune - De Novo Generation of Therapeutic Peptides with Multi-Objective-Guided Discrete Diffusion
Valence Labs via YouTube
Overview
Syllabus
PepTune: De Novo Generation of Therapeutic Peptides with Multi-Objective-Guided Discrete Diffusion
Taught by
Valence Labs