DrugHIVE: Target-specific spatial drug design and optimization with a hierarchical generative model

Jesse A. Weller,Remo Rohs

DOI: https://doi.org/10.1101/2023.12.22.573155

2024-03-17

Abstract:Rapid advancement in the computational methods of structure-based drug design has led to their widespread adoption as key tools in the early drug development process. Recently, the remarkable growth of available crystal structure data and libraries of commercially available or readily synthesizable molecules have unlocked previously inaccessible regions of chemical space for drug development. Paired with improvements in virtual ligand screening methods, these expanded libraries are having a significant impact on the success of early drug design efforts. However, screening-based methods are limited in their scalability due to computational limits and the sheer scale of drug-like space. An approach within the quickly evolving field of artificial intelligence (AI), deep generative modeling, is extending the reach of molecular design beyond classical methods by learning the fundamental intra- and inter-molecular relationships in drug-target systems from existing data. In this work we introduce DrugHIVE, a deep hierarchical structure-based generative model that enables fine-grained control over molecular generation. Our model outperforms state of the art autoregressive and diffusion-based methods on common benchmarks and in speed of generation. Here, we demonstrate DrugHIVE’s capacity to accelerate a wide range of common drug design tasks such as de novo generation, molecular optimization, scaffold hopping, linker design, and high throughput pattern replacement. Our method is highly scalable and can be applied to high confidence AlphaFold predicted receptors, extending our ability to generate high quality drug-like molecules to a majority of the unsolved human proteome.

Bioinformatics

What problem does this paper attempt to address?

This paper introduces a deep learning model called DrugHIVE for targeted drug design and optimization. Currently, despite the progress in computer-aided drug design methods that have improved the efficiency of early drug development, they are still limited by scalability and the complexity of manually adjusting candidate molecules. DrugHIVE addresses these issues through hierarchical generation models, which allow for fine-grained control over molecule generation and improve the efficiency of drug design tasks, such as de novo generation, molecular optimization, scaffold hopping, linker design, and high-throughput pattern replacement.

The distinguishing feature of DrugHIVE is its hierarchical structure, which encodes molecule structures at different spatial scales, enabling precise spatial control when generating new molecules. The model outperforms existing autoregressive and diffusion models in standard benchmark tests and has fast generation speed. Furthermore, DrugHIVE can be applied to receptor structures predicted based on AlphaFold, expanding the ability to design potential drug molecules within the unresolved human proteome.

The paper demonstrates the application of DrugHIVE in various drug design tasks, including property optimization, multi-objective evolutionary search, linker design, substructure modification (scaffold hopping), fragment growth, and pattern replacement. Experimental results show that DrugHIVE excels in enhancing binding affinity, drug similarity, and selectivity, and can even be used for drug design using protein structures predicted by AlphaFold, overcoming the limitations of relying on crystal structures. These advances provide faster and more accurate methods for the drug discovery process, helping expand the exploration range of drug design in chemical space.