MetaDomain: a profile HMM-based protein domain classification tool for short sequences

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Protein homology search is crucial for functional profiling in metagenomic annotation.
• Profile Hidden Markov Model (HMM)-based methods offer high sensitivity for remote homology detection but struggle with short reads.
• Short reads from next-generation sequencing (NGS) present a challenge for accurate protein domain classification.

Purpose of the Study:

• To introduce MetaDomain, a novel computational tool for protein domain classification specifically designed for short sequencing reads.
• To enhance the sensitivity and accuracy of identifying protein domains within short DNA sequences.

Main Methods:

• MetaDomain employs relaxed position-specific score thresholds for aligning reads to profile HMMs.
• It utilizes the distribution of alignment positions as an additional constraint to minimize false positive matches.
• The tool was applied to bacterial genome transcriptomic data and a soil metagenomic dataset.

Main Results:

• MetaDomain demonstrated superior sensitivity compared to state-of-the-art profile HMM alignment tools for short sequences.
• The tool effectively identified encoded protein domains from short reads in both tested datasets.
• Experimental results validated MetaDomain's improved performance in metagenomic annotation.

Conclusions:

• MetaDomain offers a significant advancement in protein domain classification for short reads.
• The tool enhances functional profiling capabilities in metagenomics, particularly for NGS data.
• MetaDomain provides a valuable solution for analyzing complex biological datasets with short sequence reads.