Fragment-Based Ligand Generation Guided By Geometric Deep Learning On Protein-Ligand Structure

Abstract

Computationally-aided design of novel molecules has the potential to accelerate drug discovery. Several recent generative models aimed to create new molecules for specific protein targets. However, a rate limiting step in drug development is molecule optimization, which can take several years due to the challenge of optimizing multiple molecular properties at once. We developed a method to solve a specific molecular optimization problem in silico: expanding a small, fragment-like starting molecule bound to a protein pocket into a larger molecule that matches that physiochemical properties of known drugs. Using data efficient E(3) equivariant based neural networks and a 3D atomic point cloud representation, our model learns how to attach new molecular fragments to a growing structure by recognizing realistic intermediates generated en route to a final ligand. This approach always generates chemically valid molecules and incorporates all relevant 3D spatial information from the protein pocket. This framework produces promising molecules as assessed by multiple properties that address binding affinity, ease of synthesis, and solubility. Overall, we demonstrate the feasibility of 3D molecular structure expansion conditioned on protein pockets while maintaining desirable drug-like physiochemical properties and developed a tool that could accelerate the work of medicinal chemists.

A High-Quality Data Set of Protein–Ligand Binding Interactions Via Comparative Complex Structure Modeling

“…The BindingNet data set can also be used to develop and benchmark ML-based models for binding affinity and binding pose prediction ,− and molecular generation. − PDBbind is the most widely used training data set for ML methods, , and the binding affinity prediction performances of ML-based scoring functions on PDBbind have been reported with the R p around 0.8. − However, the insufficient size and sparsity of PDBbind have resulted in poor generalization capability on out-of-distribution data sets. ,,− Both Yang et al and Mastropietro et al have found that the ligand memorization often dominates the predictions of ML-based models. , In addition, Zhu et al performed an interpretable analysis of PDBbind-trained models in 2022 and revealed that these models rely on the buried SASA-related features to make predictions . Since the BindingNet data set has significantly increased the number of available complex structures, it would be interesting to explore the performance of ML models trained on BindingNet.…”

Structure‐Based Drug Discovery with Deep Learning**

“…Later works used diffusion models, [126,127] variational autoencoders, [118] and reinforcement learning [119,120,128] to directly generate ligand conformations inside the binding pocket. Recently, equivariant neural networks coupled with point-cloud representations have been used for molecule optimization, via pocket-based fragment expansion, [200] as well as generative adversarial networks that represent proteins at the atomic level. [129] Compared to 3D graphs, molecular strings are usually easier to generate and might match or outperform graph-based models.…”

Semi-equivariant conditional normalizing flows, with applications to target-aware molecule generation

“…By contrast, Fialková et al (2021), Li et al (2021), Thomas et al (2021), Fu et al (2022)use approaches based on reinforcement learning. Other works include fragment-based ligand generation, in which new molecular fragments are sequentially attached to the growing molecule (Powers et al 2022); abstraction of the geometric interaction features of the receptor-ligand complex to a latent space, for generative models such as Bayesian sampling (Wang et al 2022b) and RNNs (Zhang and Chen 2022); and use of experimental electron densities as training data for the conditional generative model (Wang et al 2022a).…”