Improvements to FASTA

• The distribution of similarity scores with ln n are used to scale similarity scores and derive accurate expectation values. Expectation values for the highest scoring unrelated sequences typically range from 0.5 - 2.0.

• Optimized scores are calculated for about half of the library sequences. Searches with ktup=1 use a band-width of 32; ktup=2 uses a width of 16. DNA sequences use a width of 16. This calculation improves the performance of FASTA significantly but slows the program about 2-fold.

• The BLOSUM50 matrix is used for protein sequence comparison by default. Gap penalties are -12, -2 for FASTA and SSEARCH, -14, -4 for LALIGN, -16, -4 for TFASTA. The DNA scoring matrix is the same as that used by BLASTN (+5, -4); -16, -4 is the default gap penalty. (Gap penalties can be changed on the command line.)

• Final alignments are produced with the Smith-Waterman algorithm with no limits on gaps.