Google may get access to genomic patient data — why we should worry

Artificial intelligence is already being put to use in the NHS, with Google’s AI firm DeepMind providing technology to help monitor patients. Now I have discovered that Google has met with Genomic England — a company set up by the Department of Health to deliver the 100,000 Genomes Project — to discuss whether DeepMind could get involved.

If this were to happen, it could help bring down costs and speed up genetic sequencing — potentially helping the science to flourish. But what are the risks of letting a private company have access to sensitive genetic data?

Genomic sequencing has huge potential — it could hold the key to improving our understanding of a range of diseases, including cancer, and eventually help find treatments for them. The 100,000 Genomes Project was set up by the government to sequence genomes of 100,000 people. And it won’t stop there. A new report from the UK’s chief medical officer, Sally Davies, is calling for an expansion of the project.

However, a statement by the Department of Health in response to a freedom of information (FoI) request I made in February reveals this decision has already been made. The department said in this response that the project will be integrated into a single national genomic database. The purpose of this will be to support “care and research, and the acceleration of industrial usage”. Though it will “inevitably exceed the original 100,000 genomes, we do not anticipate that there will be a set target for how many genomes it should contain”, the statement reads.

The costs of sequencing the genome on a national scale are prohibitive. The first human genome was sequenced at a cost of US$3 billion. However, almost two decades later, Illumina, which is responsible for the sequencing side of the 100,000 Genomes Project, produced the first “$1,000 genome” — a staggering reduction in cost. Applying machine learning to genomics — that is, general artificial intelligence — has the potential to significantly reduce the costs further. By building a neural network, these algorithms can interpret huge amounts of genetic, health, and environmental data to predict a persons health status, such as their level of risk of heart attack.