AI has the power to search through millions of genetic variants at high speed and identify likely ... [+]
400 million people globally suffer from a rare disease. This is greater than the population of the United States, yet the ominous figures don't end there. According to the Global Genes organization, eight out of ten rare diseases are caused by a faulty gene, yet it takes an average of 4.8 years to arrive at an accurate diagnosis. This is part of the reason why 30% of children with a rare disease won't live to see their fifth birthday.
Neither is this situation helped by the fact that 95% of rare diseases lack an FDA-approved treatment. However, while the rarity of rare diseases means they're often neglected by the medical establishment, artificial intelligence and machine learning have been emerging in recent years as new, promising tools in the fight against uncommon pathology. Several companies are developing platforms that harness AI as a means to identify genetic variants at the roots of rare diseases, while medical researchers and practitioners are using these platforms or developing their own.
One of the most interesting of these is Emedgene. Based in Israel, the company has built a platform which can not only scan the DNA data file of a patient with a rare disease, but can also use natural language processing (NLP) to read the up-to-date medical literature. By doing this, Emedgene's platform can find documented connections between a patient's genetic variants and their condition, helping speed up the diagnosis. And as co-founder and CEO Einat Metzer explains, this would have been a laborious and time-consuming process in the past.
"Every person's DNA contains millions of genetic variants that are harmless, and geneticists in labs have to identify the single causative variant for the diseases out of all of those variants," she says. "When labs don't use an algorithmic approach, this typically involves filtering through hundreds of genetic variants manually. Additionally, they have to search through current literature to make sure they aren't missing any newly discovered connection or other evidence."
As Metzer adds, Emedgene performs all of these steps with machine learning algorithms, automatically identifying the causative variant, as well as supporting evidence from the literature and databases. All geneticists need to do is review the produced results, rather than conducting an analysis from scratch.
On top of this, Emedgene launched a new algorithm last month, Pathorolo, which calculates just how likely it is that a genetic case will be solved. It can also be used to reevaluate unsolved past cases, which account for as many as 60% of all rare genetic cases.
"Labs will periodically reanalyze past open cases with the hope that new information or new bioinformatics tools will help solve the case," she says. "This typically increases the percentage of solved cases by 10%. However, this is painstaking manual work and as the backlog of past unsolved cases continues to grow, unsustainable without machine learning's help. Our Pathorolo algorithm can identify which of those past cases is likely to be solved today, allowing labs to spend time only on those cases, and enabling periodic reanalysis."
Other examples of AI in the diagnosis of rare genetic diseases can be found elsewhere. Some are relatively small, such as the Germany-based startup Nostos Genomics, which like Emedgene uses machine learning to process a patient's genetic variants and identify likely causes, and which featured at this year's Entrepreneur First Demo Day in London. More interestingly, the Boston-based FDNA published a paper in January in Nature Medicine detailing how it had built a smartphone app called Face2Gene that had been trained on masses of images to recognize people with rare genetic disorders.
Others are already quite large, such as Fabric Genomics. Based in San Francisco, it harnesses AI to perform a variety of genetic analyses, having licensed its algorithms for use by more than 1,000 clinical laboratories and academic institutions around the world. In particular, it can be used to screen for genetic variants likely to have caused a rare disease, something which it claims has around a 50% success rate, as opposed to a 25%-30% industry average.
This April, Fabric Genomics' algorithms were used by researchers at Rady Children's Institute for Genomic Medicine in San Diego, where machine-learning processes and clinical natural language processing (CNLP) were able to diagnose rare genetic diseases in "record time." In this case, the overall system was used to help diagnose babies and infants in intensive care, providing genomic sequencing and also interpretation within 20 hours, and operating with 99 percent precision in 95 children with 97 genetic diseases.
"Some people call this artificial intelligence, we call it augmented intelligence," said Stephen Kingsmore, MD, DSc, President and CEO of Rady Children's Institute for Genomic Medicine. "Patient care will always begin and end with the doctor. By harnessing the power of technology, we can quickly and accurately determine the root cause of genetic diseases. We rapidly provide this critical information to intensive care physicians so they can focus on personalizing care for babies who are struggling to survive."
Such successful trials show that AI has much to offer medicine in diagnosing rare diseases, let alone the 400 million people who suffer from a rare disease. And at a time when these diseases are all-too often being neglected by hospitals, who don't always have the resources to deal with them, this is certainly one area where most of us will welcome the encroachment of AI and machine learning into our lives.
