Chemical Computing Group and Chemaxon will each teach a free workshop on Sunday 1 June before the conference starts. Participation in the workshops is free, but limited to conference participants. Registration to the workshops is mandatory and can be done by means of the registration tool. There are 40 seats available per workshop.
Generative design of novel active molecules based on multiple objective criteria
Sunday 1 June, 15:00-17:00
Barbara Sander and Guido Kirsten, Chemical Computing Group
New lead compounds discovered computationally – or identified by virtual, experimental and/or fragment screening – invariably need to be optimised for activity, ease of synthesis and other pharmacokinetic properties. This workshop explores how computational applications in the Molecular Operating Environment (MOE) software system are used to facilitate both the discovery and optimization processes. These include;
- Using binding pocket information to guide de novo design
- Scaffold Replacement, Fragment Growing and Reaction-based Transformation of starting molecules
- Bioisosteric Replacement
- Multi-objective compound scoring and filtering (descriptors, models, pharmacophores, docking scores)
- Intelligent enumeration of reactions for library design
- Cheminformatics analysis of SAR data sets
Attendees will therefore gain an awareness of the arsenal of techniques available to address topics in computational drug design.
Additionally, we will demonstrate the exploration and exploitation of chemical space using the PNN based REINVENT approach. The QSAR, pharmacophore, fingerprint and docking scoring functions in MOE can be used in an interface to filter the output of the generative model efficiently. The trained PNN will be used to sample a manageable number of interesting compounds.
Chemaxon Workshop: Machine Learning Assisted Molecular Design with Chemaxon’s Python Toolkit and Design Hub
Sunday 1 June, 15:00-17:00
Mark Szabo, Chemaxon
Design Hub is a compound design and tracking platform for drug discovery teams and their external collaborators that connects scientific hypotheses, candidate compound selection, and computational capabilities. With the rise of machine learning (ML) in drug discovery, platforms like Design Hub, combined with ML-driven methods, offer increased value for rational compound design. By leveraging Chemaxon’s freshly released native Python libraries, researchers can directly access molecular fingerprints, descriptors, and physico-chemical property calculations, making the design process more data-driven and efficient than ever before.
During this workshop, the participants will have the opportunity to build a machine learning workflow using open-source libraries such as scikit-learn while leveraging Chemaxon’s Python libraries for feature generation. A lightweight predictive model will be trained and wrapped as a Python-based plugin, which will be callable from Design Hub. The second part of the workshop will demonstrate how this custom plugin can be integrated into Design Hub. We will build a workflow where structures designed within the platform will be sent to the ML plugin, and the returned predictions — e.g., activity, synthetic feasibility, or other calculated properties — will guide compound prioritization within an interactive design generation workflow.
By the end of the session, participants will have seen a concrete example of how open-source ML tools, powered by Chemaxon descriptors, can be effectively combined with Design Hub’s production-grade cheminformatics infrastructure to support real-world molecular design workflows
A hands-on training is possible for participants using laptops and arriving 30 minutes before the workshop starts for installation and setup. Participants without coding experience are welcome too! For more information, please contact Mark Szabo: mszabo21@chemaxon.com