Searching DNA databases for similarities to DNA sequences: when is a match significant?

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• DNA sequence searching is fundamental to sequence analysis.
• Limited systematic studies exist on determining the biological significance of DNA sequence matches.
• This gap hinders the effective utilization of DNA searching algorithms.

Purpose of the Study:

• To systematically evaluate the performance of various DNA database searching algorithms.
• To establish guidelines for assessing the biological significance of DNA sequence search results.
• To identify optimal algorithms for DNA sequence analysis based on sensitivity, coverage, and speed.

Main Methods:

• Construction of a test set comprising artificially evolved and real DNA sequences.
• Testing of multiple algorithms, including BLAST, BLAST2, FASTA, and Smith-Waterman, against a subset of the EMBL database.
• Analysis of algorithm sensitivity, coverage, and performance based on sequence length and identity.

Main Results:

• The Smith-Waterman algorithm offered the best coverage but lowest sensitivity.
• The default BLASTN algorithm (word length 11) showed good sensitivity but poor coverage.
• FASTA and BLAST (word length 6) provide a balance of speed, sensitivity, and coverage.
• Algorithm performance degrades for probe sequences shorter than 200 bases.
• Accurate identification of matches in coding regions is possible with >35% protein sequence identity.

Conclusions:

• Guidelines are provided to assess the significance of DNA database search results.
• The choice of algorithm (FASTA or BLAST with word length 6) offers a practical compromise for sequence analysis.
• DNA sequence searching is a valuable tool when appropriate algorithms and criteria are applied.