Lausanne-based Sophia Genetics was co-founded in 2011 by molecular biologist and entrepreneur Jurgi Camblong to turn the raw data produced by gene sequencing machines into useful diagnostic information that allows patients to get quicker, more targeted treatment. The company now serves some 160 hospitals in 25 countries, which in turn provide the mass of raw data that allows Sophia Genetics’ artificial intelligence software to better spot the telltale signs of hereditary and other illnesses. The firm’s Chief Information Officer, Adam Molyneaux, describes the technical and ethical challenges.
Technologist: How is your technology used for diagnosis?
Adam Molyneaux: A doctor sends patients’ blood samples to a local hospital lab, which extracts DNA and puts it through a sequencing machine. The resulting genetic sequences are sent to us over a secure data link; we then stitch the sequences together and compare the genomes that emerge with a reference genome from a healthy person. Any differences between the two might indicate disease-causing mutations.
T. Why can’t hospitals do the analyses themselves?
A.M. Each step in the sequencing process adds errors, which means that if you don’t clean up samples what appear to be mutations could in fact just be artefacts. We use neural networks to analyse gene sequences to sort the wheat from the chaff. This saves clinicians time, but it is always they who have the final say on the significance of a particular mutation.
T. Why is it good to connect lots of hospitals?
A.M. The more data we feed our neural network, the more accurate its predictions will be. Every time a clinician decides whether a feature we flag is in fact pathogenic we feed that decision back into the network, so it learns. This means that each decision benefits all of our clients even though they don’t see the underlying data.
T. How do you keep those data private?
A.M. We do whatever hospitals tell us to do with the data – whether to save them or destroy them, for example. We can’t publish them and can’t sell them. What’s more, hospitals keep the data anonymous by replacing each name with a number.
T. Can you explain the problem of incidental findings?
A.M. A patient being tested for colon cancer may, for example, turn out to be susceptible to Alzheimer’s disease. That raises questions: should the clinician see those extra data, and, if so, should they report the results to the patient? It’s also possible for genetic data to identify a child’s real biological parents.
T. Are these problems technological or ethical?
A.M. We’re currently working with a professor of cryptography to mask certain sections of the genome to prevent incidental findings from “leaking out”. But technological fixes can only get you so far. The case of identifying parents, for example, has
to be handled ethically. In Europe, legislation protects the confidentiality of patients’ data.