Representation of amino acids as five-bit or three-bit patterns for filtering protein databases

Area of Science:

• Bioinformatics
• Computational Biology
• Molecular Biology

Background:

• Developing efficient methods for protein database similarity searches is crucial for bioinformatics.
• Existing filter algorithms for sequence alignment lack nuanced scoring of amino acid substitutions.
• Representing amino acids using bit-patterns offers a novel approach to accelerate similarity searches.

Purpose of the Study:

• To develop and evaluate a novel bit-pattern representation for amino acids.
• To implement a filter algorithm based on dynamic programming optimization for rapid elimination of non-homologous protein sequences.
• To enhance similarity searches by distinguishing between conservative and non-conservative amino acid substitutions.

Main Methods:

• Amino acid similarity was assessed using the BLOSUM45 matrix.
• Simulated annealing was employed to determine optimal five-bit and three-bit patterns for amino acid representation.
• The developed bit-pattern representation was tested using dynamic programming for sequence alignment.

Main Results:

• Optimal five-bit and three-bit patterns were identified, assigning similar bit-patterns to biochemically similar amino acids.
• Alignments of the Escherichia coli PhoE precursor and bacteriophage PA2 LC precursor using bit-patterns closely matched those obtained with the BLOSUM45 matrix.
• The proposed filter algorithm effectively eliminated non-homologous regions, demonstrating improved efficiency.

Conclusions:

• Bit-pattern representation provides an effective method for accelerating protein similarity searches in large databases.
• The developed algorithm enhances sequence alignment by incorporating a more sophisticated scoring function for amino acid substitutions.
• This approach offers a valuable tool for computational biology and bioinformatics research.