Exploring chemical space for "druglike" small molecules in the age of AI

Abstract

The announcement of 2024 Nobel Prize in Chemistry to Alphafold has reiterated the role of AI in biology and mainly in the domain of "drug discovery". Till few years ago, structure-based drug design (SBDD) has been the preferred experimental design in many academic and pharmaceutical R and D divisions for developing novel therapeutics. However, with the advent of AI, the drug design field especially has seen a paradigm shift in its R&D across platforms. If "drug design" is a game, there are two main players, the small molecule drug and its target biomolecule, and the rules governing the game are mainly based on the interactions between these two players. In this brief review, we will be discussing our efforts in improving the state-of-the-art technology with respect to small molecules as well as in understanding the rules of the game. The review is broadly divided into five sections with the first section introducing the field and the challenges faced and the role of AI in this domain. In the second section, we describe some of the existing small molecule libraries developed in our labs and follow-up this section with a more recent knowledge-based resource available for public use. In section four, we describe some of the screening tools developed in our laboratories and are available for public use. Finally, section five delves into how domain knowledge is improving the utilization of AI in drug design. We provide three case studies from our work to illustrate this work. Finally, we conclude with our thoughts on the future scope of AI in drug design.

Keywords: BIMP; artificial intelligence; computer aided drug design (CADD); machine learning (ML); small molecules.

FIGURE 1

An overview of the evolution, current state, and future directions of small molecule drug discovery in the age of AI is presented in this Figure, highlighting the key developments from historical advances to emerging challenges.

FIGURE 2

Workflow and architecture of Sanjeevini.

FIGURE 3

The Sanjeevini pathway for active site directed lead compound design in silico.

FIGURE 4

Workflow schematic of the molecular modeling pipeline. The process begins with molecular graph construction, followed by generation of a bond order matrix. A semi-master node is then incorporated to enhance information flow across the molecular structure. Finally, the Message Passing Neural Network (MPNN) processes this enhanced representation to generate predictions. This stepwise approach enables comprehensive molecular analysis while maintaining computational efficiency.

FIGURE 5

Summary of module L, which incorporates a deep network. The structure, parameters, and loss function correspond to the inputs of the deep network.

FIGURE 6

This figure shows the process of integrating domain knowledge into molecular graphs using a logical inference engine, followed by toxicity prediction with BotGNN.

FIGURE 7

Architectural overview of the Compositional Relational Machine (CRM). The diagram illustrates how the model iteratively constructs complex features from simple features, resulting in a feedforward network structure where each non-input node represents a complex feature that can be used for toxicity prediction explanations.

FIGURE 8

Example of a tree-like explanation structure generated by the CRM. The visualization demonstrates how activations are backtraced through the network to provide interpretable explanations of toxicity predictions, showing the hierarchical relationship between simple and complex features that contributed to the model’s decision.

FIGURE 9

Schematic representation of the Language Models with Logical Feedback (LMLF) framework. The diagram shows the integration of domain knowledge into the model’s loss function, highlighting the feedback loop mechanism that iteratively refines the generation of molecules meeting specified logical constraints for target macromolecule binding.