Speaking at FutureMed, Andreas Sundquist, CEO and co-founder of DNAnexus, laid out a future in which genomics "really takes off."
In the past 40 years, there has been roughly a million-fold improvement in semiconductor technology. "We have witnessed essentially the same improvement in DNA sequencing in just ten years," explained Dr. Andreas Sundquist, CEO and co-founder of DNAnexus, in his talk at FutureMed. There has been a 100,000-fold improvement in DNA sequencing throughput in only eight years, and nearly a million-fold reduction in the price of sequencing in ten years. "Today, it costs more to manage and analyze the data than it does to produce the data," he said.
Because the exponentials involved in DNA sequencing are so high, sweeping adoption of DNA sequencing is therefore imminent. "By 2020, every single one of you will have your DNA sequenced," Sundquist said to those in attendance.
- A Sign Language Interpreter Glove for Your Smartphone
- Millipore Scepter Handheld Automated Cell Counter
- Phrazer, a Universal Communicator for Clinicians
Sundquist became a computational genomicist by way of computer science, a field in which he holds a Ph.D. from Stanford. "I came into Stanford thinking I was going to do computational architecture," he said. "But I very quickly saw what was happening in [genomics] and noticed that there are some amazing opportunities here."
Nevertheless, there are several challenges, including the massive scale of the data, infrastructural hurdles, and systemic challenges, Sundquist said.
DATA SCALE: By 2014, we are likely to have sequenced one million human genomes. That amounts to around one exabyte (quintillion bytes) of data that will need to be stored somewhere. "This is really equivalent to what some of the biggest data power houses in Silicon Valley are working with today," Sundquist said.
INFRASTRUCTURAL ISSUES: Today, sequencing companies deliver genomic data to customers by shipping them a hard drive. "When you are talking about terabytes, petabytes, exabytes of data, this is a really big problem: getting that data where you need it to analyze it and work with it is actually a logistical nightmare," Sundquist said. "What if instead of moving the data to where you need it, you actually do the reverse? You took the computation, the task you are going to perform, and move it to the data instead." That is technically possible to do, but it is very challenging because of silo-ization. It is hard to accommodate various computer cluster environments and data storage schemes.
SYSTEMIC HURDLES: DNA sequencing is an expert-dominated field. "It is sort of like the computer industry in the '70s, which was a homebrew world," Sundquist said. "You had to actually put together your own computer and then write your own software to run on that computer that you built." Similarly, professionals in the DNA sequencing world must put together their own platforms and write their own software. This is a problem because physicians, who will help advance genomics in clinical practice, by no means have the time to develop custom genomic platforms. Another systemic problem is that the tools developed for data sequencing are not built to work together; interoperability at present is lacking as data formats continue to evolve.