Technical Walkthrough

In the Trenches at GTC: Swift: A GPU-based Smith-Waterman Sequence Alignment Program

Discuss (0)

By Jike Chong, Parasians (GTC 2012 Guest Blogger)

This week at GTC, Pankaj Gupta, a bioinformatics application developer at St. Jude Children’s Research Hospital, presented his work in the area of sequence alignment. Sequence alignment is an important component of bioinformatics that is crucial for the vision of personalized medicine. Sequence alignment matches new sequences to known sequences, while detecting point mutations, insertion mutations or deletion mutations for important initiatives such as finding better cures.

Current sequencing machines produce reads of gene samples through optical or electrical means. The read signals are converted into sequences of symbols called “bases” through machine-specific signal processing analysis, otherwise known as “primary analysis”. Machine agnostic sequence alignment algorithms are then applied to interpret the read sequences in the context of known sequences.

Pankaj Gupta’s new program Swift helps accelerate the sequence alignment process on the GPU. He chose to implement the Smith-Waterman algorithm targeting alignments of reads of ~100 base pairs and references of ~300 base pairs. Each read and reference is assigned to a thread block and threads within a thread block operate on the dynamic programming dependency wave front.

An initial version of Swift is available on with a GPL license sourceforge (see: The current version accepts a read file in FASTA format and a reference file in FASTA format as inputs, and produces reference name, read name, gapped alignment, alignment score, alignment start and end positions, and alignment length as output.

I see Pankaj’s current work as a promising starting point to making an impact on the field of bioinformatics. I have a close relative who is living with cancer.  Pankaj’s work provides me with hope that a whole generation of young researchers will master both bioinformatics and computer science to substantively advance this field in working towards a cure.

Pankaj can be reached at

Follow this link to watch a streamcast of this presentation: Swift: A GPU-based Smith-Waterman Sequence Alignment Program

About our Guest Blogger

Jike Chong is an adjunct faculty at Carnegie Mellon University, the founder of Parasians, and the organizer of the Silicon Valley GPU Meetup.