ChemX

ChemX, developed by Raven Biosciences, is an AI-driven drug design platform that brings advanced machine learning technology to medicinal chemists, cheminformaticians, and project managers. Designed to make QSAR/QSPR modelling accessible to bench scientists and power users alike, ChemX guides users through the entire process of data curation, model training, explainability analysis, and molecular screening — without requiring deep expertise in machine learning.

ChemX is built around three core principles: ease of use, explainability, and best-of-breed performance. The platform has been extensively benchmarked and demonstrated to be faster and more accurate than other QSAR/QSPR solutions, while remaining intuitive enough for students and non-specialists to use effectively.

Who ChemX Is For

• Bench scientists who generate experimental data and want to model and interpret it themselves with guided assistance
• Cheminformaticians and power users who need to go deeper into underlying processes for highly specialised tasks
• Project leads and managers who need quick access to key insights across one or more projects

Predictive Capabilities

• A wide selection of pre-trained standard models covering cell permeability, blood-brain barrier penetration, CYP inhibition, bioavailability, toxicity, and more
• Custom model training for properties such as inhibitor potency, binding affinity, efficacy, intestinal absorption, micellarity, and any other property for which data is available
• Support for both regression and classification modelling problems

Automated QSAR/QSPR Workflow — 5 Steps

1. Prepare your data using ChemX's intuitive interface for importing, managing, and curating datasets
2. Featurize molecules by automatically applying a range of molecular featurizations that translate molecular structure and chemical properties into modellable bitstrings
3. Select the best model through fully automated model and parameter selection, with rich visualisations for comparing performance metrics
4. Extract knowledge using explainability analyses that present contributing and obstructing chemical properties, SAR/SPR relationships mapped onto molecular scaffolds, and AI-generated plain text summaries powered by the latest LLM technology
5. Screen novelties by uploading a screening library, automatically featurizing molecules on the fly, and receiving predicted performance alongside applied model learnings for the top candidates

Key Features and Capabilities

• Assisted data preparation with rich graphical visualisations for rapid data overview and understanding
• Automatic selection of optimal molecular featurizers tailored to each dataset
• Fully automated model selection and hyperparameter optimisation — no machine learning expertise required
• Explainable AI outputs including interpretable plots, scaffold-mapped structure-activity relationships, and LLM-generated plain text summaries
• Seamless integration of trained models into screening workflows with straightforward upload of molecule libraries
• Access to ready-to-use standard models for common ADMET parameters alongside user-trained custom models

ChemX supports an iterative, brainstorming-friendly workflow that allows users to rapidly test new ideas, evaluate datasets, and propose novel molecules for screening. The platform is suitable for both professional drug discovery settings and academic teaching environments, and is accessible via a web-based interface with supporting documentation available for new users.