Fragment Optimization─Simple SMILES String Manipulation Strategies to Optimize a Hit or Fragment

Abstract

AI/ML-based tools for generative and predictive modeling have become important contributors to small molecule drug discovery, enabling virtual compound design and ADME profiling prior to synthesis. Alongside these methods, simpler nonlearned strategies based on direct SMILES manipulation remain useful, particularly for low data targets. Herein is described a method for fragment optimization in which characters of a SMILES string are systematically replaced with alternative SMILES patterns to generate atom-type swaps, ring expansion/contraction, cyclization, ring opening, and fragment grafts. The method was initially implemented as a KNIME workflow and demonstrated on fragment-to-lead optimization of fragments from several programs. A fully open-source counterpart written in Python utilizing AutoDock Vina is also provided. Together, these tools are used in parallel with alternative open-source and commercial solutions to accelerate fragment-to-lead optimization through iterative in silico design and profiling, and require no AI/ML or proprietary licensing for the open source python version.