alignment heuristics

Alignment heuristic parameters control fast but approximate methods for finding the alignment with the highest score. These are distinct from alignment scoring parameters, which determine the score of an alignment. Ideally, changing heuristic parameters would not change the reported alignment (because the best alignment would always be found). By contrast, changing alignment scoring parameters will tend to change the alignment, e.g. increasing gap penalties will reduce the number of gaps.

Full dynamic programming
Heuristics can be disabled entirely by using the -fulldp option, which specifies that full dynamic programming algorithms should be used. These algorithms (Needleman-Wunsch for global alignments and Smith-Waterman for local alignments) are much slower than the heuristic algorithms that USEARCH uses by default, but guarantee to find alignments with the maximum possible score.

Local alignment heuristics

Option	Default	Description
-xdrop_u	16.0	X-drop for ungapped extension.
-xdrop_g	32.0	X-drop for gapped extension.
-hspw	5 nt, 3 aa	Word length for seeding HSPs (ignored by ublast).

Global alignment heuristics

Option	Default	Description
-xdrop_nw	16.0	X-drop for HSP-finding.
-minhsp	16	Minimum HSP length. This sets an upper limit on the minimum [sic]; USEARCH will automatically reduce this limit for short sequences as needed. The default is 32 in versions through 7.0.1090.
-band	16	Minimum band radius for filling gaps between HSPs. The width of the band is 2 x radius + 1, so default width is 33. Smaller values improve speed. Longer values produce better alignments when there are many non-terminal gaps, especially long gaps, in one sequence relative to the other.
-hspw	5 nt, 3 aa	Word length for seeding HSPs.