Sequence database search using jumping alignments

Area of Science:

• Bioinformatics
• Computational Biology
• Protein Sequence Analysis

Background:

• Established methods like profiles and hidden Markov models primarily focus on vertical information in multiple sequence alignments.
• There is a need for methods that can effectively exploit both vertical and horizontal information for robust protein classification.
• Detecting remote homologues is crucial for understanding protein function and evolution.

Purpose of the Study:

• To introduce a new algorithm for amino acid sequence classification and remote homology detection.
• To evaluate the performance of this new algorithm against established methods, specifically hidden Markov models.

Main Methods:

• Developed a 'jumping alignment' algorithm, extending the Smith-Waterman algorithm to compute local alignments between a single sequence and a multiple sequence alignment.
• The algorithm aligns a candidate sequence to a reference sequence within the multiple alignment, allowing the reference sequence to change with a penalty for jumps.
• Compared the discriminative quality of the jumping alignment algorithm against hidden Markov models using a subset of the SCOP database, assessing performance via false positive counts.

Main Results:

• The jumping alignment algorithm exploits both vertical and horizontal information in multiple sequence alignments, unlike methods focusing solely on vertical data.
• For moderate false positive counts (above five), the new algorithm demonstrated a considerably higher success rate in identifying true positives compared to hidden Markov models.
• The method effectively selects proteins belonging to a specific superfamily from a candidate database.

Conclusions:

• The novel jumping alignment algorithm offers a more balanced exploitation of multiple sequence alignment information.
• This approach shows superior performance in detecting remote protein homologues compared to hidden Markov models, particularly in challenging classification tasks.
• The algorithm represents a significant advancement in bioinformatics tools for protein family classification and evolutionary analysis.