The DeCypher® and CodeQuest ™ biocomputing systems deliver the
sensitivity and high-throughput searching required for microarray probe
design, SNP mapping, and RNA interference. These systems process searches
on the DeCypher Engine™ accelerator card to deliver performance of
hundreds to thousands of CPUs within a single computer. Our accelerated
Tera-BLAST algorithm (1) meets or exceeds the sensitivity and performance
of both NCBI BLAST (2) and Smith-Waterman for oligonucleotide searches.
Profilesearch is also available on these systems and has been used to
increase the speed of Osprey probe design applications by nearly 100-fold.
Introduction
Comparing short sequences to a genome database is essential for
minimizing non-specific hybridization in microarray probe design. Software
Smith-Waterman (ssearch, 3) is computationally expensive and slow to
process. BLAST software, while faster, is less sensitive than
Smith-Waterman and may miss valid hits.
Probe design: Smith-Waterman sensitivity
at the speed of BLAST
Figure 1. Comparing 10,000
oligonucleotide sequences (30-mers) to Unigene Mouse genome A
DeCypher system with 4 cards and 4 CPUs compared 10,000 oligos to
the Unigene mouse genome with Tera-BLAST in 1.5 hours. A similar
search requires 6 hours on a Paracel GeneMatcher2 system with 9
boards and 9 CPUs.
|
Probe candidates are compared to a genomic database to assess binding
frequency. Oligo sequences that "hit" frequently throughout a
genome are removed to minimize potential cross-hybridization. TimeLogic
has optimized performance and sensitivity of Tera-BLAST for
oligonucleotide searches by adding a query-global alignment method. Using
query global alignments and parameter settings that require each 9-base
region of the target to be analyzed for potential hits, better sensitivity
and specificity than other methods are demonstrated. A set of 10,000
oligonucleotide sequences was compared to the mouse genome. The top 10,000
scores and alignments from Tera-BLAST matched 99.999% of the hits reported
by Smith-Waterman on a Paracel GeneMatcher2. Because of its improved gap
handling at the ends of each alignment, it identified an additional 9,000
unique hits missed by GeneMatcher (Fig. 1).
Increasing probe design speed using
Osprey and Profilesearch
Osprey (4), an application for selecting optimal microarray probes,
employs Profilesearch to improve the specificity and sensitivity of
secondary binding searches. Utilizing the current DeCypher Engine G4
cards, a 91-fold performance improvement is observed using Profilesearch
on DeCypher (equivalent to processing Osprey with 728 CPUs).
