3D-Scaffold: A Deep Learning Framework to Generate 3D Coordinates of Drug-like Molecules with Desired Scaffolds

Abstract

The prerequisite of therapeutic drug design and discovery is to identify novel molecules and developing lead candidates with desired biophysical and biochemical properties. Deep generative models have demonstrated their ability to find such molecules by exploring a huge chemical space efficiently. An effective way to generate new molecules with desired target properties is by constraining the critical fucntional groups or the core scaffolds in the generation process. To this end, we developed a domain aware generative framework called 3D-Scaffold that takes 3D coordinates of the desired scaffold as an input and generates 3D coordinates of novel therapeutic candidates as an output while always preserving the desired scaffolds in generated structures. We demonstrated that our framework generates predominantly valid, unique, novel, and experimentally synthesizable molecules that have drug-like properties similar to the molecules in the training set. Using domain specific data sets, we generate covalent and noncovalent antiviral inhibitors targeting viral proteins. To measure the success of our framework in generating therapeutic candidates, generated structures were subjected to high throughput virtual screening via docking simulations, which shows favorable interaction against SARS-CoV-2 main protease (Mpro) and nonstructural protein endoribonuclease (NSP15) targets. Most importantly, our deep learning model performs well with relatively small 3D structural training data and quickly learns to generalize to new scaffolds, highlighting its potential application to other domains for generating target specific candidates.

Artificial Intelligence for Autonomous Molecular Design: A Perspective

“…In their models, information is propagated back and forth in the molecules in the form of waves, making it possible to pass the information locally while simultaneously traveling the entire molecule in a single pass. With the unprecedented success of learned molecular representations for predictive modeling, they are also adopted with success for generative models [57,69].…”

“…They then use the 3D coordinates of the molecules to learn the representation to map it to a space, which is then used to generate 3D coordinates of the novel molecules. Building on this for a drug discovery application, we recently proposed a model [69] to generate 3D coordinates of molecules while always preserving the desired scaffolds, as depicted in Figure 5. This approach has generated synthesizable drug-like molecules that show a high docking score against the target protein.…”

“…Figure5. Generative model such as 3D-scaffold[69] can be used to inverse design novel candidates with desired target properties starting from core scaffold or functional group.…”

Inverse design of 3d molecular structures with conditional generative neural networks

“…In contrast, other models aim at sampling directly from distributions of 3d molecules with arbitrary composition 47 – 56 , making them suitable for general inverse design settings. These models need to be biased towards structures with properties of interest, e.g., using reinforcement learning 51 , 52 , 56 , fine-tuning on a biased dataset 48 , or other heuristics 54 .…”

“…Some of us have previously proposed G-SchNet 48 , an autoregressive deep neural network that generates diverse, small organic molecules by placing atom after atom in Euclidean space. It has been applied in the 3D-Scaffold framework to build molecules around a functional group associated with properties of interest in order to discover novel drug candidates 54 . Such an approach requires prior knowledge about the relationship between functional groups and target properties and might prevent the model from unfolding its potential by limiting sampling to very specific molecules.…”