A benchmarking framework for comparative evaluation of low-complexity region detection tools in the human proteome

Area of Science:

• Proteomics
• Bioinformatics
• Computational Biology

Background:

• Low-complexity regions (LCRs) are crucial protein segments involved in molecular recognition and phase separation.
• Accurate computational detection of LCRs is hindered by methodological variability.

Purpose of the Study:

• To comprehensively benchmark eight common LCR detection methods across the human proteome.
• To develop a standardized framework for evaluating LCR detection tool performance.

Main Methods:

• Systematic comparison of LCR detection tools using a modular computational framework.
• Residue-centric and protein-centric analyses of detected LCRs, including length, composition, and entropy.
• Consensus and similarity analyses (Jaccard index) to evaluate algorithm agreement.

Main Results:

• Consensus LCRs were generally longer, more repetitive, and compositionally purer, indicating higher functional relevance.
• Distinct algorithm clusters emerged based on shared detection principles, visualized via Jaccard similarity.
• Significant differences in captured sequence complexity were observed among tools, particularly in entropy and purity metrics.

Conclusions:

• The study provides a unified and reproducible framework for LCR detection evaluation.
• Findings offer practical guidelines for accurate LCR annotation in large-scale proteomic studies.
• Highlights the importance of consensus approaches for identifying functionally significant LCRs.