PREDICTION OF SIDEROPHORES PARTITION COEFFICIENT USING ARTIFICIAL NEURAL NETWORKS

Abstract

The octanol-water partition coefficient (logP) is a crucial indicator in the study of lipophilicity and cell permeability, making it a recurring molecular descriptor in empirical rules for evaluating a molecule's pharmacokinetics. Siderophores are pharmacologically relevant molecules due to their potential Trojan horse effect; however, two major challenges arise: their molecular weight often exceeds 500 Daltons, and there is a lack of databases containing atomic coordinates of their three-dimensional structures and molecular descriptors. In this work, we have created a database containing the SMILES codes of siderophores, their names, associated microorganisms, molecular descriptors, among other information, which is available in our repository at https://github.com/inefable12/siderophores_database. We have also developed a web page to visualize the 2D and 3D structures (https://sideroforos.streamlit.app). Additionally, we demonstrate a quick and efficient way to estimate the logP for siderophores using artificial neural networks in R. The information provided in this article aims to facilitate the structural study of siderophores, the design of potential metallodrugs, the generation of their three-dimensional structures for docking and molecular dynamics simulations, as well as the development of new predictive models for properties using artificial intelligence.