Deep Learning, Genomics Spur Advances in Understanding and Hope for Treatment. But journalists should also focus on regulation of the AI and electronic medical records that underlie biotech gains.
By Chris Adams, National Press Foundation

“Deep learning” is generating excitement for rare disease diagnosis and treatment. “Artificial intelligence” is the overarching label for computers that learn and reason like humans (or better). “Machine learning” and “deep learning” are subsets of AI. Ben Solomon of the U.S. National Institutes of Health explains deep learning here. “We can use artificial intelligence to sort all the different possible compounds and molecules that might be able to address a genetic condition or a rare condition,” said Ben Solomon of the U.S. National Institutes of Health. An exciting development comes from the ability to run through myriad sequences of how proteins fold, Solomon said, citing new research from DeepMind. “People are talking about it like it’s going to be as impactful as CRISPR gene editing. I’m not sure about that, but it’s a giant, giant, exciting, big deal.”

Deep learning is about sensing patterns, learning from those patterns and then sensing new ones. Early facial recognition programs recognized light and darkness, then facial features like eyes and noses, then identified individual people. Now the programs may be able to spot facial features of patients who may have specific diseases, such as Noonan syndrome or Williams syndrome, Solomon said. The technology is already often good enough to outperform a human geneticist, but doctors are not going to lose their jobs anytime soon.

Reporting on the regulation of the AI and electronic health records that drive genomic research will be critical. This applies to rare and common diseases alike, Solomon said. Stat news recently reported on a health information management conference in Las Vegas where one of the main takeaways was that “machine learning needs a watchdog.” The AI and Human Rights Project at the Electronic Privacy Information Center (EPIC) is a source for keeping abreast of legislation and other regulatory developments.

Not all the research in rare diseases gets published. A little over half of deep learning studies of genetic conditions focus on diagnosis, followed by studies assessing disease progression and identifying phenotypic features of diseases. U.S. researchers have taken the lead, although China and Italy are playing an important role. And while the number of published studies has skyrocketed in recent years, much of it remains hidden, Solomon said. “There’s tons of this going on that’s not published – lots of biotech companies because they might not care as much about publications. They want to get a therapeutic out there or get a new test out there. … Even though we see hundreds or thousands of articles, even recently on this, this is just the tip of the iceberg.”

Speaker: Ben Solomon, Clinical Director, National Human Genome Research Institute, U.S. National Institutes of Health
This program was funded by Fondation Ipsen. NPF is solely responsible for the content.

More about Rare Diseases: [ Ссылка ]
AI Driving Breakthroughs on Rare Diseases: [ Ссылка ]
NPF website: [ Ссылка ]

свернуть