Gregory Percy Knowles

DASH: localising dynamic programming for order of magnitude faster, accurate sequence alignment

In this paper we present our genomic and proteomic sequence alignment algorithm, DASH, which results in order of magnitude speed improvement when compared to NCBI-BLAST 2.2.6, with superior sensitivity. Dynamic programming (DP) is the predominant contributor to search time for algorithms such as BLAST and FastA/P. Improving the efficiency of DP provides an opportunity to increase sensitivity, or significantly reduce search times and help offset the effects of the continuing exponential growth in database sizes. Specifically, for nucleotide searching we have demonstrated an order of magnitude speed improvement with significantly improved sensitivity, or alternatively moderate speed up with further sensitivity gains, depending on the parameters selected. Smith-Waterman complete DP is used as the sensitivity benchmark. Similar speed and sensitivity results are presented for protein searching. Since our algorithm is highly parallel, we have developed dedicated hardware which we will present in a companion paper, and a distributed version of our software (DDASH), which we expect to provide linear speedup on a cluster.

Novel architecture for the JPEG2000 block coder

JPEG2000 is the newest international standard to be produced by the Joint Photographic Experts Group (JPEG) and it has recently been published as an International Standards Organization/IEC standard and as an ITU-T recommendation. It complements most of the currently used standards including the original JPEG, usually with superior compression for the same image quality, or alternatively with superior image quality for the same bit rate. The algorithm uses the wavelet transform followed by an embedded block coder incorporating the MQ arithmetic coder. In this article we give a novel hardware architecture for the block coder and MQ coder as used in the JPEG2000 standard. Together with appropriate software, this design enables a high performance system-on-chip solution for JPEG2000 image compression. We have included sufficient introductory material to interest readers unfamiliar with either JPEG2000 or hardware design.

A novel architecture for genomic sequence searching and alignment

Blast[2], FASTA[4] and related algorithms are popular tools for searching genomic data. Accelerating these tools is an increasingly important goal as the growth of the databases outstrips Moore’s Law[6]. Many of the existing hardware designs are little more than direct translations of existing algorithms from the software to hardware domain[3, 7]. In this paper we summarise a novel approach which we have implemented in VHDL and synthesised, with the aim of validating an efficient algorithm and virtual hardware search system which produces near perfect results compared to NCBI Blast 2.2.3. Speed and quality comparisons of the VHDL simulations with respect to NCBI Blast 2.2.3 are included.

A Complete Classification of Short Symmetric-antisymmetric Multiwavelets

In this note we give a complete characterisation of the class of symmetric-antisymmetric multiwavelets with three and four coefficients. That is, we find all three tap multiwavelets satisfying the following conditions: (a) orthogonality (b) one symmetric component and the other antisymmetric (c) approximation order two and (d) compact support in the interval [0,2]. We also find all four tap multiwavelets satisfying the conditions (a), (b), (c) and (d) compact support in the interval [0, 3].