Deep Generative Models for 3D Linker Design

Abstract

Rational compound design remains a challenging problem for both computational methods and medicinal chemists. Computational generative methods have begun to show promising results for the design problem. However, they have not yet used the power of three-dimensional (3D) structural information. We have developed a novel graph-based deep generative model that combines state-of-the-art machine learning techniques with structural knowledge. Our method (“DeLinker”) takes two fragments or partial structures and designs a molecule incorporating both. The generation process is protein-context-dependent, utilizing the relative distance and orientation between the partial structures. This 3D information is vital to successful compound design, and we demonstrate its impact on the generation process and the limitations of omitting such information. In a large-scale evaluation, DeLinker designed 60% more molecules with high 3D similarity to the original molecule than a database baseline. When considering the more relevant problem of longer linkers with at least five atoms, the outperformance increased to 200%. We demonstrate the effectiveness and applicability of this approach on a diverse range of design problems: fragment linking, scaffold hopping, and proteolysis targeting chimera (PROTAC) design. As far as we are aware, this is the first molecular generative model to incorporate 3D structural information directly in the design process. The code is available at .

Generating Property-Matched Decoy Molecules Using Deep Learning

“…Imrie et al (8) builds on the generative process introduced by Liu et al (14) that constructs molecules "bond-by-bond" in a breadth-first manner. The most substantial differences with Imrie et al (8) are the input data and goal of the generative process.…”

“…In order to generate decoys we use an adapted version of Imrie et al (8), which was designed for linker generation. Imrie et al (8) builds on the generative process introduced by Liu et al (14) that constructs molecules "bond-by-bond" in a breadth-first manner.…”

“…To improve the quality of generated molecules, we adopted a novel loss function that deviates from a standard cross entropy loss for the sequence of actions adopted by Imrie et al (8) and Liu et al (14). Instead of each step in the generative processes having equal importance, we reweighted the probabilities of actions by the frequencies of 2.2 Training set the induced subgraphs across the training set of molecules, leading to the revised cross-entropy loss:…”

“…For a more detailed description of the model, see Imrie et al (8) and the Supplementary Information.…”

PRosettaC: Rosetta based modeling of PROTAC mediated ternary complexes

“…Notably, RosettaDock was used by Nowak et al 27 to inform the design of selective PROTACs. Two relevant computational approaches for linker design were recently introduced 28,29 that can be used to design PROTAC linkers based on the protein-protein binary complex. While these may be suitable as a sub-procedure within a general protocol, they cannot in and of themselves model ternary complexes.…”