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.

Generation of drug-like molecules with high binding affinity to target proteins remains a difficult and resource-intensive task in drug discovery. Existing approaches primarily employ reinforcement learning, Markov sampling, or deep generative models guided by Gaussian processes, which can be prohibitively slow when generating molecules with high binding affinity calculated by computationally-expensive physics-based methods. We present Latent Inceptionism on Molecules (LIMO), which significantly accelerates molecule generation with an inceptionism-like technique. LIMO employs a variational autoencoder-generated latent space and property prediction by two neural networks in sequence to enable faster gradient-based reverse-optimization of molecular properties. Comprehensive experiments show that LIMO performs competitively on benchmark tasks and markedly outperforms state-of-the-art techniques on the novel task of generating drug-like compounds with high binding affinity, reaching nanomolar range against two protein targets. We corroborate these docking-based results with more accurate molecular dynamics-based calculations of absolute binding free energy and show that one of our generated drug-like compounds has a predicted $K_D$ (a measure of binding affinity) of $6 \cdot 10^{-14}$ M against the human estrogen receptor, well beyond the affinities of typical early-stage drug candidates and most FDA-approved drugs to their respective targets. Code is available at https://github.com/Rose-STL-Lab/LIMO.

Abstract Motivation: The large and growing body of experimental data on biomolecular binding is of enormous value in developing a deeper understanding of molecular biology, in developing new therapeutics, and in various molecular design applications. However, most of these data are found only in the published literature and are therefore difficult to access and use. No existing public database has focused on measured binding affinities and has provided query capabilities that include chemical structure and sequence homology searches. Methods & results: We have created Binding DataBase (BindingDB), a public, web-accessible database of measured binding affinities. BindingDB is based upon a relational data specification for describing binding measurements via Isothermal Titration Calorimetry (ITC) and enzyme inhibition. A corresponding XML Document Type Definition (DTD) is used to create and parse intermediate files during the on-line deposition process and will also be used for data interchange, including collection of data from other sources. The on-line query interface, which is constructed with Java Servlet technology, supports standard SQL queries as well as searches for molecules by chemical structure and sequence homology. The on-line deposition interface uses Java Server Pages and JavaBean objects to generate dynamic HTML and to store intermediate results. The resulting data resource provides a range of functionality with brisk response-times, and lends itself well to continued development and enhancement. Availability: The BindingDB is publicly available at http://www.bindingdb.org. The database schema and DTD file are available at http://www.bindingdb.org/bind/entity˙report.html, and http://www.bindingdb.org/bind/deposition/xml/BindingDB.xml, respectively. Contact: [email protected] * To whom correspondence should be addressed. 2 Present address: NovaScreen, 7170 Standard Drive, Hanover, MD 21076, USA.