Fig 1.
Sketch of a potential improved screening system for DNA synthesis orders.
Iterative hashing would enable companies to send out sequences to be screened externally while protecting trade secrets. Order fragments of approximately 40 bp could be screened for exact matches against a hashed database of hazardous sequences by a cooperative international network of servers verifiable by third parties. Orders and database could be kept private using uniquely salted local hashes plus a multiparty ball-and-chain, or possibly homomorphic encryption. Exact sequence comparison and the size of the sequence space relative to order volume could effectively eliminate false positives unless database salting is desired for improved security. Hazardous sequences could be filtered from crowdsourced suggestions by an international team of experts from synthesis companies, universities, and other institutions. Ideally, individual members could add new sequences privately to minimize information hazards.
Fig 2.
Paths towards the adoption of universal screening of commercial synthetic DNA orders for hazardous sequences.
Covering the up-front cost of screening system development and initial adoption would eliminate barriers that might prevent companies from participating. Publishers could incentivize participation by declining to publish submitted manuscripts that rely on unscreened DNA, whereas universities, societies, companies, and funders could boycott groups that decline to adopt screening. Once in place, this system could effectively require hardware-level locks on future hardware permitting distributed synthesis. International governments could subsidize all screened oligonucleotides to impose an effective market-based requirement for approximately 1/15 of the current US annual biodefense budget.
Fig 3.
Greater openness could accelerate progress and inoculate science against hazardous mistakes.
Current incentives encourage scientists to keep research plans to themselves until publication (top), which prevents others from suggesting improvements. Fields such as gene drive have moral reasons to shift towards a fully open model (right) in which anyone can share advice, but this may not be practical for all fields due to commercial incentives and the risk of disclosing research plans that would themselves be information hazards. An intermediate model (left) might adapt current grant evaluation systems or national boards to ensure that proposed projects are confidentially preregistered and peer-reviewed by experts from diverse fields who lack conflicts of interest, enabling them to suggest ways of mitigating potential hazards in advance of experiments. This approach might be usefully pioneered by the field of synthetic mammalian virology. In both open models, early advice from peers would likely accelerate discovery relative to the current closeted approach.