Massively Parallel Sequence Comparison on a 16,384 Processor MasPar MP-2

The Laboratory of Mathematical Biology in conjunction with the Frederick Biomedical Supercomputer Center have been exploring the use of massively parallel computation for sequence comparison. Given the large quantities of protein and nucleic acid sequence information produced in laboratories every year, it is apparent that rapid methodologies are needed for comparing sequences against other sequences in these rapidly growing databases. We have been using MPSRCH on our 16384 processor MasPar MP-2 in this endeavor. The program is a parallel implementations of the Smith-Waterman dynamic programming algorithm for sequence comparison. We have seen rates as high as 1.5 billion cell updates per second with searches using MPSRCH against a non-redundant nucleic acid database consisting of over 800 million bases (this includes the reverse complement form of the database). Such rapid comparisons increase researcher productivity and allow room for varied experimentation.

MPsrch consists of a suite of eight programs for searching the protein and nucleic acid databases. These programs are available through a Moasaic server and consist of the following set:

Protein sequence querie(s) versus a protein sequence database.
Protein sequence querie(s) versus a protein sequence database. Uses affine gap penalties.
Nucleic acid sequence queries(s) versus a nucleic acid sequence database.
Nucleic acid sequence queries(s) versus a nucleic acid sequence database. Uses affine gap penalties.
Nucleic acid sequence queries(s) versus a protein sequence database that has been back translated into nucleic acid sequences.
Nucleic acid sequence querie(s) versus a protein sequence database that has been back translated into nucleic acid sequences. Uses affine gap penalities.
Protein sequence querie(s) that have been backtranslated into nucleic acid sequences versus a protein sequence database.
Protein sequence querie(s) that have been backtranslated into nucleic acid sequences versus a protein sequence database. Uses affine gap penalties.

At the current time non-NIH server access requires a password. To activate server CLICK HERE!


Please contact: Dr. Bruce Shapiro
                Laboratory of Mathematical Biology
                National Cancer Institute
                Frederick Cancer Research and Development Center
                Building 469, Room 150
                Frederick, Maryland 21702

                email: bshapiro@ncifcrf.gov
                301-846-5535

                     or

                Gary Smythers
                NCI-FCRDC
                Advanced Scientific Computing Laboratory.
                PO Box B, Bldg. 430
                Frederick Md. 21702-1201

                gws@ncifcrf.gov
                301-846-5779