Molecular Interactions and Binding Prediction in Drug Discovery

Hello everyone,
We would like to invite you to our AI in Pharma MeetUp focused on Molecular Interactions and Binding Prediction in Drug Discovery

Please save the date: 19 May 2026 for this AI in Biochemistry MeetUp in Munich (Address: Luise-Ullrich-Straße 14, 80636 Munich).

Many thanks to our co-sponsor: Reply Concept

Agenda Overview:
7:00 PM - Door Open, Drinks & Networking

7:30 PM - Welcome Introduction

7:35 PM - Talk 1: Target Preference Mapping: A Connectivity-Free Approach to Ligand and Protein Interaction Prediction
Dr. Sergey Vilov
Machine Learning Engineer ,
Khumbu AI

8:05 PM - Talk 2: A Structured Map of Antibody Development: Opportunities for AI
Dr. Martin Schatte
Head of Protein Science,
Antibody Development,
Lead Discovery Center (LDC) GmbH

8:35 PM - Drinks, Snacks & Networking

Talks Details:

Talk 1
Title: Target Preference Mapping: A Connectivity-Free Approach to Ligand and Protein Interaction Prediction

Abstract:
Structure-based drug discovery has achieved remarkable success in protein structure prediction, yet the translation from protein structure to functional molecular binders remains fundamentally limited. Existing machine learning and docking approaches rely heavily on historical ligand data, chemical connectivity, and affinity measurements, resulting in strong bias, poor generalization, and limited applicability to novel chemistry, protein–protein interactions, and dynamic systems.
Target Preference Mapping (TPM) is a connectivity-free, affinity-independent framework that learns interaction preferences directly from structural microenvironments. TPM decomposes protein binding sites into local spatial voxels and predicts probabilistic preferences for ligand atom electronic states without encoding ligand structure or bonding. This abstraction enables the model to learn non-bonded chemistry principles, directionality, and conformational sensitivity purely from structural context.
TPM accurately reproduces known binding modes, correlates with experimental affinities despite never being trained on them, and generalizes beyond kinase-dominated chemical space. The approach is validated prospectively through blind kinase selectivity screening, biochemical assays, and ex vivo human tissue experiments, where TPM-selected compounds perform comparably to or better than clinical candidates. TPM extends naturally to protein–protein interactions and conformationally dynamic targets.
TPM provides a new paradigm for molecular discovery, enabling bias-free screening, rational selectivity design, interpretation of variants of unknown significance, and a foundation for future generative models in small-molecule and protein engineering.

Speaker:
Dr. Sergey Vilov,
Machine Learning Engineer,
Khumbu AI

Biography:
Dr. Sergey Vilov, earned his PhD in Physics from the University of Grenoble in 2019. He then joined Helmholtz Zentrum Munich as a postdoctoral researcher, where he developed deep learning-based approaches for various genomic tasks. Since summer 2025, he has been working as a machine learning engineer at Khumbu.ai, developing ML-based approaches for early-stage drug discovery.

Talk 2
Title: A Structured Map of Antibody Development: Opportunities for AI

Abstract:
While artificial intelligence and machine learning have generated considerable enthusiasm regarding complete in silico antibody design, current computational approaches remain insufficient to fully replace experimental screening and validation across the therapeutic development pipeline. However, rather than representing a limitation, this reality reveals a more pragmatic and productive opportunity: AI can substantially enhance and accelerate virtually every stage of antibody development when strategically integrated with experimental workflows.
This presentation maps the comprehensive antibody development pipeline—from immunogen design through humanization—and systematically demonstrates where AI technologies can address critical bottlenecks and improve efficiency. We examine each development phase, contrasting traditional methodologies with AI-supported or AI-driven approaches: computational immunogen optimization to enhance immune responses; machine learning-guided library design to increase binder discovery rates; advanced sequence analysis and binding prediction for accelerated screening; developability assessment through predictive modeling to eliminate problematic candidates early; computational affinity maturation to improve binding characteristics; and automated humanization strategies. The presentation concludes with a practical case study detailing our recent applications and evaluating their performance in a real-world setting.

Speaker:
Dr. Martin Schatte
Head of Protein Science,
Antibody Development,
Lead Discovery Center (LDC) GmbH

Biography:
Dr. Schatte earned both his Bachelor’s and Master’s degrees in Biochemistry from Technical University of Munich. He subsequently worked as a scientist at Oregon Health & Science University, focusing on lymphocyte biology, before joining Roche, where he contributed to the engineering and purification development of diagnostic enzymes.
Dr. Who completed his PhD at RWTH Aachen University under the supervision of Prof. Schwaneberg, specializing in the engineering of sortases. He later worked as a scientist at Clariant, developing industrial enzymes and microbial strains. Currently, he serves as Head of Protein Science at the Lead Discovery Center, where he leads efforts in the development of therapeutic antibodies.