A benchmark study of sequence alignment methods for protein clustering.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics and Proteomics

Background:

• Protein sequence alignment is vital in bioinformatics.
• Multiple Sequence Alignment (MSA) and Pair-wise Sequence Alignment (PSA) are key methods.
• Previous studies noted MSA drawbacks in nucleotide alignments.

Purpose of the Study:

• To investigate if MSA methods have similar drawbacks in protein sequence alignment.
• To introduce a novel benchmark framework for protein clustering using cluster validity.
• To evaluate alignment quality based on biological goals, not just sequence-level comparisons.

Main Methods:

• Developed a new benchmark framework for protein clustering based on cluster validity.
• Calculated cluster validity scores using sequence distances to avoid clustering method bias.
• Tested performance on BAliBASE benchmark datasets and re-sampled datasets.

Main Results:

• PSA methods demonstrated superior performance compared to MSA methods across most BAliBASE datasets.
• Consistent results were observed in analyses of 80 re-sampled benchmark datasets.
• The proposed framework confirmed the impact of alignment method choice on results.

Conclusions:

• The study validates that MSA methods exhibit drawbacks in protein sequence alignment, mirroring those found in nucleotide alignments.
• These drawbacks negatively affect the accuracy of protein sequence analysis results.
• The novel benchmark framework provides a reliable method for evaluating protein alignment quality based on biological relevance.