Dr. Tembe's solution is specific to genomic sequencing data. In addition to analyzing the frequency of the ACGT letters that make up DNA, G-SQZ also can encode the annotation information, including the data's quality, as well as erroneous entries, such as unidentified bases.
The indexing system used in G-SQZ allows access at regular intervals, such as every millionth data point, so all the information need not be decoded from the start.
"It's not enough to compress the information. The compressed representation should allow quick retrieval and querying," Dr. Tembe said. "To that end, G-SQZ has been designed as an efficient practical approach, rather than a theoretically optimal compression algorithm."
Even faster advancements on the horizon
Dr. Tembe is moving ahead with improving his current design to accommodate what he calls "parallel computing."
Because G-SQZ compression keeps the data ordered and indexed, the squeezed data can be split into smaller "chunks," allowing multiple computer processors to decode and analyze different parts of the same file simultaneously, he said. For example, if a file is indexed at 1,000 places, it can be fed into a supercomputer, allowing 1,000 processors to analyze the data at the same time, speeding up the results. Analysis tools using parallel programming approaches can take advantage of the G-SQZ encoding format.
"While indexed and compressed representation is ready, the parallel computing functionality is undergoing a testing phase," Dr. Tembe said. "But this is where it is headed. Sequencing hundreds of billions of bases per run is now a reality. The real impact of G-SQZ lies in the storage, transfer and processing of genomic sequencing data, where substantial room for improvement still exists."
|Contact: Steve Yozwiak|
The Translational Genomics Research Institute