Using Artificial Intelligence to Revolutionise Precision Medicine
If you go to any meetings on precision medicine, you’ll have noticed the increasing amount of interest in artificial intelligence (AI). This is largely down to the sharp increase in genomic data being produced, and the growing volume of digitised phenotypic data that is being made available. As a result, it’s made the field of precision medicine, an exciting nexus of technologies.
One of the companies quickest to recognise this potential was FDNA. “We are on the cusp of a new era for healthcare – the era of genomics. The convergence of technologies, such as cloud computing, AI, mobile devices as biometric sensors and genome sequencing, is creating the ‘perfect storm’ conditions for precision medicine”, said CEO, Dekel Gelbman
“What we have seen over the past few years of being active in the field of clinical genetics is that phenotyping is critical for understanding and interpreting genetic test results. Powered by AI, we are now able to take a variety of biometric data never used before on a scalable basis, such as facial photos, and hone in on specific genetic variants that cause diseases.”
FDNA aren’t alone in applying AI to improve healthcare. Computing giants, IBM, have put their Watson AI to work helping users make quicker decisions in a variety of settings. Similarly, Verily (formerly Google Life Sciences), are developing a series of tools to collect health data, and producing tools to help squeeze out every last insight from it. A little closer to home for genomics are Craig Venter-led Human Longevity, who are busy building the world’s most comprehensive database on human genotypes and phenotypes for a variety of uses.
The benefits of AI in the clinic are already apparent. The Swiss company, SOPHiA GENETICS, were only founded in 2011 but have already rolled out their AI into several hundred hospitals around the world. Their platform is helping clinicians interpret genomic data to increase accuracy of diagnoses and efficiency of treatment. At the heart of their success is collaboration and knowledge sharing.
CEO and Co-founder, Dr Jurgi Camblong, explained, “You want to carry out AI through a platform where many people are able to benefit from your technology, and learn from the information that has been produced. Sharing knowledge is very important.”
As a result, SOPHiA AI has successfully been adopted in 370 hospitals which are all connected to one another by a software service programme, supporting the diagnosis of approximately 8,000 patients per month and having analysed over 150,000 patients in a total of 55 countries, thus has created the world’s largest clinical genomics community.
“The beauty of AI in an industry is when you start digitalising information, and this enables you to spread knowledge and eventually every patient will be treated equally,” added Dr Camblong.
AI is at its best when it has a large amount of data to work with. So the networks and data sharing workflows built up by companies like SOPHiA GENETICS are critical to their success.
“We have built a network of clinicians, labs and researchers in human genetics in over 2,000 institutions across 130 countries,” said Gelbman, speaking on FDNA’s Face2Gene phenotyping application. It works with biometric data to help physicians identify causative genetic variants for a suite of diseases. This has contributed to a global database that drives the development of a ubiquitous phenotyping technology – next-generation phenotyping.”
“The technology is made available free of charge to clinicians, so the benefit to the entire community is immediate. As more biometric data is sourced from our network, the technology becomes better.”
However, like anything, the sharing of data brings with it its own challenges. “One of the biggest challenges of sharing data is patient privacy,” commented Gelbman. We have overcome that challenge by extracting only de-identified data from patient records and aggregating that data into algorithm.”
“AI is at its best when it devours huge amounts of data and is able to detect very distinct patterns. We were able to develop translational learning methods that allow us to learn from large cohorts of patients processed through our system and apply these techniques to smaller cohorts.”
The question that hangs over everyone’s head as AI continues to develop is whether or not it will replace humans. While that may be the case in some areas, as far as clinical support goes, it couldn’t be further from the truth. If anything, it’ll free up doctors to do more of what we need them to do. “AI should free up doctors from the burdensome task of sifting through mountains of data to pick up medical clues that lead to a diagnosis and leave them more time to spend with patients,” explained Gelbman.
As with any emerging technology, there are challenges to overcome. For AI, these are largely complex data analysis, data privacy, and defining AI applications. The early success of pioneers such as SOPHiA GENETICS and FDNA, have demonstrated that applying AI to a defined task can make a significant positive impact on how we look after patients. With the influx of computational minds applying themselves to genomics, the coming years should see some very interesting applications revolutionise precision medicine.