ChemX

ChemX offers an intuitive user interface designed to streamline the process of curating data and preparing it for training robust machine learning models. It applies a variety of molecular featurizations to help identify the most suitable one for your data. The platform allows for easy comparison of resulting models through rich visualizations, enabling the selection of the best model based on performance metrics for both regression and classification models.

After model training, ChemX extracts insights and presents them as interpretable plots showing contributing and obstructing chemical properties, as well as structure/property relationships mapped onto molecular scaffolds. Additionally, AI-generated plain text summaries provide further clarity.

The platform facilitates seamless integration of the best model into screening workflows. Users can upload a list of molecules, which are automatically prepared for screening. ChemX returns predictions of performance along with applied learnings of the model on the best predicted molecules.

Automated QSAR/QSPR in 5 Easy Steps

1. Prepare your data
2. Featurize molecules
3. Select the best model
4. Extract knowledge
5. Screen novelties

ChemX's user-friendly interface simplifies data import, management, and preparation for machine learning, allowing users to focus on chemistry rather than data handling. Rich graphical visualizations provide a rapid overview and understanding of the data.

The platform automatically applies a range of molecular featurizations, translating molecular structures and chemical properties into modellable bitstrings to identify the best featurizer for your data.

For users not specialized in machine learning, ChemX automatically selects the best models and parameters for your data, whether for regression or classification problems. Powerful visualizations facilitate easy comparison of model performance.

With ChemX, users gain insights into why certain molecules are selected for testing. The platform extracts model learnings and presents them as interpretable plots and AI-generated summaries, using the latest in LLM technology.

Implementation is straightforward: upload a screening library and select models to apply. Molecules are featurized on the fly, with options to use custom models or standard models for predicting common ADMET parameters.

ChemX has been extensively benchmarked, demonstrating superior speed and performance compared to other QSAR/QSPR solutions.