Not one to shy away from unproven technology, Craig Venter said last week that simulation is still a bit immature for his tastes when it comes to solving biological problems.
During his opening keynote address at the Bio-IT World conference last week, Venter spoke about his plans to create “artificial chromosomes” and eventually synthesize entire cells with properties of interest for environmental research. Artificial organisms with improved carbon dioxide conversion capabilities or hydrogen production properties would fit into the mission of one of his non-profit groups, the Institute for Biological Energy Alternatives, he said.
Later, in an interview with BioInform, when asked what role in silico modeling and simulation technology would play in this work, Venter was uncharacteristically cautious.
“Computer modeling is not sophisticated enough yet,” he said. Although an early proponent of the approach — he co-authored the original E-Cell paper — Venter said the technique is still constrained by the limited amount of biological data available.
“We don’t even know what one-third of the genes do in a minimal cell,” he said of Escherichia coli, “so it’s really hard to model them.”
While the technology shows promise, “you can’t design an organism in a computer yet,” he said.
On the hardware side, however, Venter was happy to prognosticate about what the future will hold for life science computing: Technology will continue to follow Moore’s law, resulting in smaller, faster, and cheaper computers that will make the 700-Alpha processor system his team used to assemble the human genome at Celera look like a Univac. While the Celera system ate up 6,000 square feet to deliver 1.5 teraflops and 120 terabytes, Venter said that by 2025, he expects to see 50-square-foot machines delivering 550 petaflops and 10 exabytes.
He had just returned from a visit to Sandia, he said, and was pleased to see the work the lab is doing to build low-cost, energy efficient computers. A one-teraflop system is now available for $2 million, he said — good news for biology, which is still “limited by compute cycles.”
Forget the far-off cellular modeling stuff: Sufficient compute capacity isn’t even available to verify the finished Celera sequence against the public sequence, he said, which means the field is still not able to establish “basic standards for accuracy” for genomic data.