Trusted Cheminformatics, Zero Integration Headaches: ML in Python with Chemaxon

Tired of stitching together structure handling, descriptor generation, and your ML framework across tools that have a hard time talking to each other? The new Chemaxon Python APIs put trusted, production-grade cheminformatics directly in your Python stack!

In this session, Certara’s Jan Christopherson will go from a SMILES string to a working activity prediction model, live in a Jupyter notebook. Using a real ChEMBL cholinesterase target, he builds a simple model predicting pChEMBL (pIC50-style) activity, a clear, reproducible workflow you can adapt to your own data, not a black box.

• Build an ML model, start to finish. Read, standardize, and featurize structures, then train a model on a real activity endpoint. Walk away with a pattern you can reuse immediately.
• See what 27 years of cheminformatics buys you. Reliable molecule standardization, accurate physicochemical properties, and fingerprints you can trust as ML descriptors. Clean chemistry in means trustworthy predictions out.
• Watch it slot into your existing pipeline. pandas, your ML libraries, Jupyter. No rebuilding required.

• Computational chemists and cheminformaticians building predictive models
• ML and data scientists in drug discovery working with molecular data
• Scientific programmers scripting analyses in Python and Jupyter
• Chemaxon and Certara users curious about the new Python APIs