Aligning protein sequences with predicted secondary structure.

Area of Science:

• Bioinformatics
• Computational Biology
• Structural Bioinformatics

Background:

• Accurate protein sequence alignment is crucial for understanding protein function and evolution.
• Standard alignment methods struggle with distant protein sequences due to insufficient information in amino acid sequences alone.
• Incorporating additional data, such as secondary structure, can enhance alignment accuracy.

Purpose of the Study:

• To develop and evaluate novel models for protein sequence alignment that leverage predicted secondary structure information.
• To introduce efficient algorithms for optimal alignment using these enhanced models.
• To improve parameter learning for alignment models through inverse alignment techniques.

Main Methods:

• Development of new scoring models that incorporate predicted secondary structure and its confidence.
• Implementation of near-quadratic time algorithms for optimal pairwise alignment.
• Application of inverse alignment with a novel criterion to balance score and recovery errors for parameter learning.
• Rigorous experimental evaluation on benchmark alignments using known 3D structures.

Main Results:

• New models significantly improve alignment accuracy for distant protein sequences compared to standard methods.
• Pairwise alignment accuracy improved by up to 15% for sequences with <25% identity.
• Multiple alignment accuracy increased by over 20% for difficult benchmarks where standard tools achieved only ~40% accuracy.

Conclusions:

• Predicted secondary structure is a valuable feature for improving protein sequence alignment, particularly for distantly related sequences.
• The developed models and algorithms offer a substantial advancement in alignment accuracy and efficiency.
• These findings have implications for comparative genomics, protein structure prediction, and functional annotation.