AlignHUSH: alignment of HMMs using structure and hydrophobicity information

Area of Science:

• Bioinformatics
• Computational Biology
• Structural Bioinformatics

Background:

• Accurate protein remote homology detection and alignment are crucial for function annotation and structural modeling, especially in the 'midnight zone' of sequence similarity.
• Existing methods struggle with distantly related protein families, necessitating improved algorithms for sensitive detection.

Purpose of the Study:

• To develop and evaluate AlignHUSH, a novel algorithm for HMM-HMM alignment designed for enhanced remote homology detection.
• To leverage structural information and amino acid properties for more sensitive protein sequence alignments.

Main Methods:

• AlignHUSH aligns profile hidden Markov models (HMMs) using predicted secondary structure probabilities and amino acid hydrophobicity.
• The algorithm incorporates information from adjoining columns to improve alignment sensitivity.
• Performance was assessed using evolutionary relationships from SCOP and structure-based alignments from BaliBASE.

Main Results:

• AlignHUSH demonstrates superior performance compared to existing HMM-HMM alignment methods, particularly at higher error rates.
• The method achieves sensitive remote homology detection and generates alignments suitable for function annotation and homology modeling.
• Alignment accuracy is comparable to established profile-profile alignment techniques.

Conclusions:

• AlignHUSH offers improved sensitivity for HMM-HMM alignment, outperforming other programs at error rates of 10% and above.
• The developed method can significantly aid in deciphering the functions of proteins with unknown roles.
• A web server implementing AlignHUSH is publicly accessible for broader research application.