Challenges in computational discovery of bioactive peptides in 'omics data

Area of Science:

• Biotechnology
• Genomics
• Bioinformatics

Background:

• Peptides exhibit diverse biological activities with biotechnological potential.
• Omics data offers opportunities for novel molecule discovery, but peptide identification is complex.
• Traditional protein discovery methods fail for peptides due to their small size.

Purpose of the Study:

• To review limitations of traditional methods for peptide discovery.
• To highlight alternative machine learning approaches for identifying novel bioactive peptides.
• To focus on peptide discovery in prokaryotic genomes and metagenomes.

Main Methods:

• Review of traditional sequence similarity and functional annotation limitations.
• Exploration of machine learning-based methods for peptide prediction.
• Analysis of challenges in identifying peptides from short open reading frames (smORFs) and proteolysis.

Main Results:

• Traditional methods yield high false positives for peptide discovery.
• Machine learning offers a more effective alternative for functional annotation of short sequences.
• Prokaryotic genomes and metagenomes are rich sources for novel peptide mining.

Conclusions:

• Novel machine learning approaches are crucial for overcoming limitations in bioactive peptide discovery.
• Effective mining of omics data requires specialized bioinformatics tools for peptides.
• Understanding peptide origins (smORFs vs. proteolysis) aids discovery efforts.