Unraveling the encrypted lipophilicity of disulfide bridges in the molecular architecture of proteins and peptides: A machine learning approach in structural bioinformatics

Abstract

Disulfide bridges can be considered the concertmaster in a molecular orchestra that directs folding or inhibits unfolding in the molecular architecture of proteins and peptides, so standing for key structural element in protein stability. Its formation requires that two side chain Sγ atoms of cysteine residues are spatially proximal forming the new cystine residues (CRs). Although cystine-containing biomolecules such as antimicrobial peptides (defensines and lantibiotics), toxins (snake venoms and conopeptides), hyperstable peptides (cyclotides and lasso peptides) are well represented in databases and a wide diversity of residue-based lipophilicity scales are reported in the literature, there are no experimental or theoretical studies that describe this fundamental property in CRs. Here, with the motivation to complement previous studies focused on the lipophilicity of amino acids, we have studied the conformational richness of CRs in the biomolecules mentioned above characterizing structural details such as chiralities of the disulfide bridge but also, computing the QM-based lipophilicity (logP), unraveling for the first time the encrypted lipophilicities of different conformations of CRs. Interestingly, we have found a very diverse lipophilicity profiles of CRs in the peptides and proteins studied, thus indicating a hydrophobic fingerprint that could explain molecular recognition events and the unique properties of these biomolecules. In addition, we have developed a machine learning model capable of recognizing whether a cystine residue is hydrophilic or hydrophobic using chiral descriptors of the disulfide bridge. Our results have the potential to contribute to a better understanding of the implications of CRs on the properties and functions of cystine-containing biomolecules.