As the price of genome sequencing plummets, the output of genomics data is skyrocketing. According to the U.S. National Human Genome Research Institute, over the next few years, the number of human genomes sequenced is expected to explode — from around 30,000 in 2011 to more than a million before 2014. Forget, for a moment, the processing horsepower necessary to transform everyone’s genome into useful medical knowledge; just the storage capacity required to hold all this data is staggering.
As pointed out in a recent post at Technology Review (TR), if every person on the planet had their genome sequenced, it would take up as much digital storage as was available world-wide in 2010, estimated to be just over 721 PB. That’s assuming 100 GB per human genome.
But that 100 GB represents a pretty brute-force storage model. Theoretically, a person’s 3.2 billion base pairs should only take 800 MB (each of the four bases can be packed into 2 bits). The problem, according the TR post, is that a lot of other data is collected about the bases, and the genes are sequenced multiple times for the sake of accuracy.
One solution, at least according to Harvard geneticist George Church, is to only store the differences between a particular genome and some reference genome. According to Church that would reduce the data capacity needed to a mere 4 MB per person. Using this approach, it would take just 28 PB of storage to hold all human genomes.
And if that seems like a lot, keep in mind that the Blue Waters supercomputer will have a storage capacity of over 25 PB when it comes online later this year. By the middle of this decade, when petascale supercomputers are apt to be much more commonplace, that 28 PB is probably going to reflect an average-sized storage capacity for hundreds of systems around the world.