ProtPCV: A Fixed Dimensional Numerical Representation of Protein Sequence to Significantly Reduce Sequence Search...

Area of Science:

• Bioinformatics
• Computational Biology
• Protein Science

Background:

• Current protein homologue search methods often overlook the ordered physicochemical properties of protein sequences.
• Existing techniques lack robust benchmarking, questioning their reliability in search engines.

Purpose of the Study:

• To develop a novel numerical representation for protein sequences to improve homologue identification.
• To establish a more comprehensive benchmarking approach for protein similarity search methods.

Main Methods:

• A fixed-dimensional numerical representation of protein sequences was created, extending the periodicity count concept.
• Euclidean distance was employed as a direct similarity measure between protein representations.
• The new method was benchmarked against BLAST, PSI-BLAST, Needleman-Wunsch, and Smith-Waterman using novel correlation-based metrics.

Main Results:

• The novel numerical representation significantly reduces computational complexity in protein sequence searches to O(log(n)).
• The proposed benchmarking methods provide a stronger validation of search technique performance.
• The approach demonstrates potential for various similarity-based operations like clustering and phylogenetic analysis.

Conclusions:

• This new numerical representation offers an efficient and validated method for protein sequence analysis.
• The findings suggest improved accuracy and speed for protein homologue detection and related applications.
• The method facilitates broader applications in protein classification and evolutionary studies.