MD-LAIs Software: Computing Whole-Sequence and Amino\nAcid-Level “Embeddings” for Peptides and Proteins

Several computational tools have\nbeen developed to calculate sequence-based\nmolecular descriptors (MDs) for peptides and proteins. However, these\ntools have certain limitations: 1) They generally lack capabilities\nfor curating input data. 2) Their outputs often exhibit significant\noverlap. 3) There is limited availability of MDs at the amino acid\n(<i>aa</i>) level. 4) They lack flexibility in computing\nspecific MDs. To address these issues, we developed <b>MD-LAIs</b> (<b>M</b>olecular <b>D</b>escriptors from <b>L</b>ocal <b>A</b>mino acid <b>I</b>nvariant<b>s</b>), Java-based software designed to compute both whole-sequence and <i>aa</i>-level MDs for peptides and proteins. These MDs are generated\nby applying aggregation operators (<b>AOs</b>) to macromolecular\nvectors containing the chemical-physical and structural properties\nof <i>aas</i>. The set of <b>AOs</b> includes both\nnonclassical (e.g., Minkowski norms) and classical <b>AOs</b> (e.g., Radial Distribution Function). Classical <b>AOs</b> capture neighborhood structural information at different <i>k</i> levels, while nonclassical <b>AOs</b> are applied\nusing a sliding window to generalize the <i>aa</i>-level\noutput. A weighting system based on fuzzy membership functions is\nalso included to account for the contributions of individual <i>aas</i>. <b>MD-LAIs</b> features: 1) a module for data\ncuration tasks, 2) a feature selection module, 3) projects of highly\nrelevant MDs, and 4) low-dimensional lists of informative global and <i>aa</i>-level MDs. Overall, we expect that <b>MD-LAIs</b> will be a valuable tool for encoding protein or peptide sequences.\nThe software is freely available as a stand-alone system on GitHub\n(https://github.com/Grupo-Medicina-Molecular-y-Traslacional/MD_LAIS).