MF-LAL: Drug Compound Generation Using Multi-Fidelity Latent Space Active Learning

Current generative models for drug discovery primarily use molecular docking as an oracle to guide the generation of active compounds. However, such models are often not useful in practice because even compounds with high docking scores do not consistently show real-world experimental activity. More accurate methods for activity prediction exist, such as molecular dynamics based binding free energy calculations, but they are too computationally expensive to use in a generative model. To address this challenge, we propose Multi-Fidelity Latent space Active Learning (MF-LAL), a generative modeling framework that integrates a set of oracles with varying cost-accuracy tradeoffs. Using active learning, we train a surrogate model for each oracle and use these surrogates to guide generation of compounds with high predicted activity. Unlike previous approaches that separately learn the surrogate model and generative model, MF-LAL combines the generative and multi-fidelity surrogate models into a single framework, allowing for more accurate activity prediction and higher quality samples. Our experiments on two disease-relevant proteins show that MF-LAL produces compounds with significantly better binding free energy scores than other single and multi-fidelity approaches (~50% improvement in mean binding free energy score). The code is available at https://github.com/Rose-STL-Lab/MF-LAL.

Designing new drugs often involves using software to predict which molecules might work well and treat a disease of interest, but these tools can be unreliable. Many current methods rely on a fast but inaccurate method called molecular docking, which often fails to reproduce real-world results. More accurate methods exist, like simulations that predict how strongly a drug binds to its target, but these are too slow and expensive to use often. Our new method, called MF-LAL, integrates the fast and slow methods, and learns when and how to use each tool to balance speed and accuracy. By training a model that learns from all these tools together, MF-LAL does a better job at finding promising drug candidates. In tests on two disease-related proteins, our method found compounds that scored about 50% better using the highest-accuracy method than those from previous approaches.