mpialign(1) parallel local alignment of biological sequences


mpialign [-s scores] [-S split] [-H hblk] [-V vblk] file1 file2


mpiAlign is a local sequence aligner, especially intended for use with large biological DNA sequences, with more than 1Mbp (Millions of base pairs). It uses the Smith-Waterman exact algorithm with affine gap cost function to perform this task.

General options:

display this help and exit
-s <scores>
specify a comma-separated list of scores to be used while calculating aligment matrices throughout the program. The list must be in the format "-sMATCH,MISMATCH,GAP_OPEN,GAP_EXTENSION" (without quotes), and is parsed in this PRECISE order; no spaces are allowed between values. If there are any unspecified parameters, these are set to default values and a warning message is issued; exceeding parameters are discarded

Stage 2 options:

-S <split>
(mpialign only) number of parts in which to split the alignment matrix; after this step a cyclic block model is applied, subdividing each part equally between all available nodes
-H <hblk>
(mpialign only) number of horizontal subdivisions made by each node to its alignment submatrix; since this value defines block width, a good choice should allow two full matrix lines to fit the processor's cache pages, improving algorithm performance
-V <vblk>
(mpialign only) number of vertical subdivisions made by each node to its alignment submatrix; this value directly affects the amount of internode communication and is used ONLY if 'split' is set to 1, otherwise it is set to the number of available nodes