Researchers from Stanford Developed ADMET-AI: A Machine Learning Platform that Provides Fast and Accurate ADMET Predictions both as a Website and as a Python Package

The combination of enlarging chemical repositories and incorporating generative AI into drug discovery procedures has produced many promising drug candidates. Nevertheless, the real challenge lies in effectively identifying compounds with ideal druglike characteristics, specifically those related to absorption, distribution, metabolism, extraction, and toxicity (ADMET). Conventional screening methods can be tedious and may not provide the desired level of precision. To tackle this issue, a team of esteemed researchers from Stanford University and Greenstone Biosciences have introduced ADMET-AI, an advanced machine-learning platform designed to forecast ADMET properties for extensive chemical libraries rapidly and accurately.

In drug discovery, high-throughput docking and generative AI have greatly increased the number of potential candidates for new drugs. However, these methods often produce molecules that may not have the best properties for use as drugs. This means that there is a need for a screening tool that is both fast and accurate. The proposed solution to this problem is ADMET-AI, which uses a graph neural network called Chemprop-RDKit. This network has been trained on 41 datasets from the Therapeutics Data Commons, allowing it to outperform other prediction tools in speed and accuracy. ADMET-AI also has unique features, such as making predictions on batches of molecules and providing contextualized predictions based on a set of approved drugs.

The architecture of ADMET-AI, specifically the Chemprop-RDKit integration, combines a graph neural network with 200 physicochemical molecular features that RDKit computes. This unique combination allows the model to accurately predict a wide range of ADMET properties, which has resulted in its outstanding performance and highest average rank on the TDC ADMET Benchmark Group leaderboard. The platform has demonstrated its effectiveness across 41 TDC ADMET datasets, excelling in regression and classification tasks. A particularly impressive feature is the web server’s exceptional speed, 45% faster than the next fastest ADMET web server. Additionally, the local version of ADMET-AI enhances its practicality by providing high-throughput prediction capabilities, which can process one million molecules in just 3.1 hours.

In conclusion, ADMET-AI is a unique force that is revolutionizing the field of drug discovery by providing a fast, precise, and adaptable platform for analyzing massive chemical libraries. ADMET-AI is an indispensable tool for researchers and practitioners due to its accuracy in predicting ADMET features and its special ability to provide contextualized predictions against a reference set of licensed medications. Due to its speed, accuracy, and user-friendly interfaces, the platform represents a substantial leap in identifying drug candidates with optimum ADMET profiles for further development. It is available as a web-based service or a local tool. The capabilities of ADMET-AI meet the pressing demand for an effective screening tool in light of the growing complexity of drug discovery campaigns and the expansion of chemical spaces. The pace and accuracy of drug discovery efforts are increasing as they expand.

Check out the Paper, Project, and Github. All credit for this research goes to the researchers of this project. Also, don’t forget to follow us on Twitter. Join our 36k+ ML SubReddit, 41k+ Facebook Community, Discord Channel, and LinkedIn Group.