Machine learning approaches to optimize small-molecule inhibitors for RNA targeting

Abstract

In the era of data science, data-driven algorithms have emerged as powerful platforms that can consolidate bioisosteric rules for preferential modifications on small molecules with a common molecular scaffold. Here we present complementary data-driven algorithms to minimize the search in chemical space for phenylthiazole-containing molecules that bind the RNA hairpin within the ribosomal peptidyl transferase center (PTC) of Mycobacterium tuberculosis. Our results indicate visual, geometrical, and chemical features that enhance the binding to the targeted RNA. Functional validation was conducted after synthesizing 10 small molecules pinpointed computationally. Four of the 10 were found to be potent inhibitors that target hairpin 91 in the ribosomal PTC of M. tuberculosis and, as a result, stop translation. Graphical Abstract

The Experimentalist’s Guide to Machine Learning for Small Molecule Design

“…However, the amount of computation required scales quickly with the number of algorithms combined; therefore, it is not a widely utilized modality compared to the use of singular algorithms. On the other hand, its utility has been demonstrated in the 2022 study by Grimberg et al, in which they combined lasso regression, a decision tree, and a convolutional neural network to identify small molecules capable of targeting the RNA hairpin in the ribosomal peptidyl transferase region of M. tuberculosis . The activity of a number of these predicted molecules was confirmed experimentally, and 4 out of 10 of those synthesized resulted in the inhibition of protein translation in the bacterium.…”

Quantitative Structure–Activity Relationship (QSAR) Study Predicts Small-Molecule Binding to RNA Structure

“…Despite its success in protein-based ligand design, however, a few QSAR studies have been conducted for identifying RNA-targeted small molecules. 31 − 34 While significant work has been done to explore key descriptors involved in RNA recognition, 35 − 37 these existing data cannot be used as input for a QSAR approach targeting a specific RNA structure, as these data are derived from disparate methods and RNA targets.…”

“…Recently, machine learning-aided mechanistic studies and ligand predictions have shown success in multiple complex tasks, including the design of enantioselective catalysts in organic synthesis and bioactive ligands for kinase inhibition. − Among multiple computational tools, quantitative structure–activity relationship (QSAR) studies can pinpoint guiding principles for a specific target by correlating the experimentally observed binding properties with the molecular descriptors of the ligands. − A robust and predictive QSAR model has been proven to be an efficient tool to predict the activities of small-molecule candidates and to drive hit optimization. Despite its success in protein-based ligand design, however, a few QSAR studies have been conducted for identifying RNA-targeted small molecules. − While significant work has been done to explore key descriptors involved in RNA recognition, − these existing data cannot be used as input for a QSAR approach targeting a specific RNA structure, as these data are derived from disparate methods and RNA targets.…”

Amplifying gene expression with RNA-targeted therapeutics

“…Given the unique nature of such interactions, rSMs are typically found via extensive screening. Recent advances in computational modelling of RNA have facilitated in silico identification of rSMs that can directly bind RNA molecules to modify the function of a given RNA structural moiety, inactivate the RNA transcript by generating covalent bonds or even trigger degradation of the RNA transcript 81 .…”