Many newborn babies admitted to intensive care have genetic disorders. The symptoms can be obvious — uncontrollable seizures, spontaneously peeling skin, abnormal heartbeats — but the cause often remains unknown. That uncertainty has painful ripples: physicians have little knowledge about how to guide treatment and parents are left unsure whether to have further children.

Genome sequencing can help. Using the fastest available sequencing instrument and software designed to guide clinicians through analysis, a team at the Children's Mercy Hospital in Kansas City, Missouri, reported in Science Translational Medicine last week that they had used sequencing in newborns to sift for rare genetic mutations that might cause disease (C. J. Saunders et al. Sci. Transl. Med. 4, 154ra135; 2012). The results were impressive. For three of the four infants, probable culprits were identified.

To reach these conclusions, the team considered not just genetic regions in the babies, but also those in their relatives and in the scientific literature. However, for sequencing to reach its medical potential, researchers must be able to access even more genomes. Each person has millions of genetic variants — or sequences that differ from the human reference genome — making it hard to find those that might affect health. The key is to locate variants that recur in people with similar illnesses.

With so many sequencing projects under way, clinicians are always eager to know whether a variant has been observed in patients at other institutions. Analysis tools are available to help (see page 157). Yet there is currently no quick, reliable or convenient way to spread this information.

There is currently no quick, reliable or convenient way to spread genomic information.

Data sharing through scientific publication has fuelled an impressive collection of databases that reveal frequencies of common variants. When variants or genes have been associated with disease, those results are also deposited in databases. ClinVar, a database from the US National Institutes of Health (NIH), for instance, gathers health-related genetic variations from the literature. And the NIH has set money aside to create a separate resource for clinically relevant genetic variants: essentially a curated database of variants for which some sort of clinical action is advised.

These are valuable efforts, but are inherently limited. Publication is too slow, and data collected about many variants will never be published. Researchers need to be able to query not just variants in the literature, but also those that have been found in other patients but not reported.

This January, an advisory group to the UK Department of Health said that the country should create a centralized facility to store genomic data to improve treatments and diagnoses. However, in the United States, where many sequencing projects are based, regulations about sharing patient data will make setting up a centralized repository more difficult.

One option would be to give patients their own sequenced genome data, letting them deposit it where they choose. Already, 23andMe, a consumer-genetics company in Mountain View, California, has used data and DNA supplied by its customers to discover (and, controversially, to patent) disease-associated variants.

Another option is for medical-research institutions to agree on ways to share information with each other. Rather than transferring a full medical record, for example, a researcher at one institute could learn whether a variant had been observed in other people and, if so, what diseases they had. For this to work, new technology and shared platforms would have to be developed.

There are other problems. Sequencing data are imperfect. High-throughput sequencing technology sometimes overlooks variants or makes other errors. Downstream issues are rife: are the benefits of sequencing worth the costs? Can the information be protected? How accurate are conclusions? Should information not related to the immediate medical question be shared with patients, even if a diagnosis is uncertain or no treatment is available?

All of these questions would be easier to answer if genomics data could provide more certainty. Yet to achieve this, researchers must look at the genomes and health information of more people. At this stage, the best path forward remains unclear. But for genomics to advance, the community must communicate. Institutions must consider not just what is best for their particular situation, but also what is best for the broader community. A good place to start is for staff at genomics centres and hospitals to meet in person, to share experiences and best practices.