MD-LAIs Software: Computing Whole-Sequence and Amino Acid-Level "Embeddings" for Peptides and Proteins

Area of Science:

• Computational biology
• Bioinformatics
• Cheminformatics

Background:

• Existing computational tools for molecular descriptors (MDs) in peptides and proteins have limitations, including poor data curation, output overlap, and limited amino acid (aa)-level descriptors.
• There is a need for flexible software that can compute both global and aa-level MDs with improved data handling.

Purpose of the Study:

• To develop MD-LAIs, a Java-based software tool to compute sequence-based molecular descriptors for peptides and proteins.
• To address limitations of existing tools by providing data curation, feature selection, and flexible computation of aa-level MDs.

Main Methods:

• MD-LAIs applies aggregation operators (AOs), including classical and nonclassical types, to macromolecular vectors of amino acid properties.
• A sliding window approach and fuzzy membership functions are used for aa-level descriptor generalization and weighting.
• The software incorporates data curation and feature selection modules.

Main Results:

• MD-LAIs computes both whole-sequence and aa-level molecular descriptors for peptides and proteins.
• The software offers a data curation module, feature selection, and generates low-dimensional lists of informative MDs.
• It provides flexibility in computing specific MDs and addresses output overlap issues.

Conclusions:

• MD-LAIs is a valuable, freely available tool for encoding protein and peptide sequences.
• The software enhances the analysis of peptides and proteins by providing comprehensive and flexible molecular descriptor calculation.
• It is expected to aid researchers in various fields requiring sequence-based molecular data.