AI Health

Friday Roundup

The AI Health Friday Roundup highlights the week’s news and publications related to artificial intelligence, data science, public health, and clinical research.

July 29, 2022

In today’s Duke AI Health Friday Roundup: the evolution of lactose tolerance; philosophical NLP AI hard to tell from the real thing; possible data fraud rocks Alzheimer research; free library of AlphaFold protein structures released; humans may be less resilient to extreme heat than thought; nursing homes pursue aggressive legal tactics over unpaid bills; study homes in on COVID outbreak epicenter; White House pivots toward harm reduction in drug policy; why data breaches keep happening; much more:


A glass storefront window with an illuminated neon sign that says “Do not trust robots.” Image credit: Nick Fewings/Unsplash
Image credit: Nick Fewings/Unsplash
  • “…the Dennett quiz revealed how, as natural language processing systems become more sophisticated and common, we’ll need to grapple with the implications of how easy it can be to be deceived by them. …The Dennett quiz prompts discussions around the ethics of replicating someone’s words or likeness, and how we might better educate people about the limitations of such systems…” Motherboard’s Shayla Love reports on an experiment that showed many participants had difficulty distinguishing between the words of an actual philosopher and an NLP AI (based on the popular GPT-3 model) trained on his printed works.
  • “…we introduce a deep learning framework to learn from high-dimensional dynamical data while maintaining stable, ecologically valid interpretations. Results successfully demonstrate that the proposed framework enables learning the dynamics of resting-state fMRI directly from small data and capturing compact, stable interpretations of features predictive of function and dysfunction.” An article by Rahman and colleagues, recently published in Scientific Reports, debuts a deep learning model for functional MRI data.
  • “…we must examine the recurring elements that allow data breaches to happen and try to learn from them. Common plotlines include human error, unnecessary data collection, consolidated storage and careless mistakes. Countless stories involve organizations that spent a ton of money on security and still ended up breached. Only when we learn from these recurring stories can we make headway in stopping the cycle.” An article in Scientific American by Daniel J. Solove and Woodrow Hartzog asks why we are not collectively doing better in terms of preventing data breaches.
  • “Despite a statistically significant association between PRS and 30-year risk of CHD, the C statistic improved only marginally with the addition of PRS to the traditional risk factor model among young adults and did not improve among midlife adults. PRS, an immutable factor that cannot be directly intervened on, has minimal clinical utility for long-term CHD prediction when added to a traditional risk factor model.” A research paper by a group of Duke researchers, published in Circulation, casts doubt on the clinical usefulness of polygenic risk scores for long-term prediction of coronary heart disease.
  • “Last year, DeepMind released the source code of AlphaFold and made the structures of 1 million proteins, including nearly every protein in the human body, available in its AlphaFold Protein Structure Database. The database was built together with the European Molecular Biology Laboratory, an international public research institute that already hosts a large database of protein information.” MIT Technology Review’s Melissa Heikkilä reports that after a remarkable run of success in predicting protein structures, DeepMind’s AlphaFold is releasing its enormous library of hundreds of millions of proteins, free to all comers.


Closeup photo of a curious black and white cow, staring straight into the camera lens. Image credit: Jan Huber/Unsplash
Image credit: Jan Huber/Unsplash
  • “We propose that lactase non-persistent individuals consumed milk when it became available but, under conditions of famine and/or increased pathogen exposure, this was disadvantageous, driving LP selection in prehistoric Europe. Comparison of model likelihoods indicates that population fluctuations, settlement density and wild animal exploitation—proxies for these drivers—provide better explanations of LP selection than the extent of milk exploitation. A paper published in Nature by Evershed and colleagues sheds light on the genetic and evolutionary forces affecting the ability in humans to digest lactose after early childhood (H/T @AdamRutherford).
  • “A leading independent image analyst and several top Alzheimer’s researchers—including George Perry of the University of Texas, San Antonio, and John Forsayeth of the University of California, San Francisco (UCSF)—reviewed most of Schrag’s findings at Science’s request. They concurred with his overall conclusions, which cast doubt on hundreds of images, including more than 70 in Lesné’s papers. Some look like “shockingly blatant” examples of image tampering, says Donna Wilcock, an Alzheimer’s expert at the University of Kentucky.” A stunning report by Science’s Charles Piller documents the work of data sleuths who may have uncovered a massively consequential episode of academic fraud at the foundations of 15 years’ worth of research into beta amyloid as a therapeutic target in Alzheimer disease. In tangentially related news, Reuters reports that Cassava Sciences is facing a criminal investigation into allegations of scientific fraud related to one of its Alzheimer candidate drugs.
  • “Brutal heat waves are quickly becoming the hallmark of the summer of 2022….And even as climate change continues to crank up the temperature, scientists are working fast to understand the limits of humans’ resilience to heat extremes. Recent research suggests that heat stress tolerance in people may be lower than previously thought. If true, millions more people could be at risk of succumbing to dangerous temperatures sooner than expected.” At Science News, Carolyn Gramling examines recent studies that suggest humans may be at higher risk from extreme heat than had been assumed.
  • “We found little to no evidence for a causal connection between game play and well-being. However, results suggested that motivations play a role in players’ well-being. For good or ill, the average effects of time spent playing video games on players’ well-being are probably very small, and further industry data are required to determine potential risks and supportive factors to health.” Good news (maybe?) for gamers: a recent study by Vuorre and colleagues, published in Royal Society Open Science, finds negligible evidence for a causal relationship between the amount of time spent playing video games and mental well-being.
  • “We show the earliest known COVID-19 cases from December 2019, including those without reported direct links, were geographically centered on [the Huanan Seafood Wholesale Market in Wuhan, China] …While there is insufficient evidence to define upstream events, and exact circumstances remain obscure, our analyses indicate that the emergence of SARS-CoV-2 occurred via the live wildlife trade in China, and show that the Huanan market was the epicenter of the COVID-19 pandemic.” A paper published this Tuesday in Science by Worobey and colleagues helps narrow down the possible jumping-off point for the COVID pandemic.

COMMUNICATION, Health Equity & Policy

Woman assisting an elderly woman in a wheel chair in a sunlit meadow with a tree in the foreground. Image credit: Domenik Lange/Unsplash
Image credit: Domenik Lange/Unsplash
  • “The nursing home industry has quietly developed what consumer attorneys and patient advocates say is a pernicious strategy of pursuing family and friends of patients despite federal law that was enacted to protect them from debt collection. ‘The level of aggression that nursing homes are using to collect unpaid debt is severely increasing,’ said Lisa Neeley, a Massachusetts elder law attorney.” NPR’s Noam Levey reports on a joint NPR-Kaiser Health News investigation that reveals aggressive legal tactics by nursing homes aimed at recovering on unpaid bills – even if that means pursuing family and friends who have no actual legal liability for the debt.
  • “…an increasing number of adjectives and adverbs were used and the readability of scientific texts have decreased in the examined years. More importantly, the use of emotion adjectives and adverbs also demonstrated an upward trend while that of nonemotion adjectives and adverbs did not increase.” A paper by Ju Wen and Lei Lei published in Scientometrics evaluates the use of adjectives and adverbs in scientific writing and its putative impact on readability over three decades.
  • “Vehicle data hubs ingest vehicle and movement data from several different sources: from OEMs [original equipment manufacturers], from other connected vehicle data providers, directly from vehicles using aftermarket hardware (such as an onboard diagnostic [OBD] dongle), or from smartphone apps. The companies normalize the data and offer it to customers in the form of a dashboard or insights derived from analysis or other data products.” At the Markup, Jon Keegan and Alfred Ng describe the torrents of data that modern autos are transmitting to an array of listeners – many or all of them unsuspected by the cars’ owners.
  • “The strategy largely rests on the concept of harm reduction, focused not on helping drug users achieve abstinence but on lowering their risk of dying or acquiring infectious diseases. A central piece is providing sterile needles to use in injecting drugs, tools to check drugs for fentanyl and other lethal substances, and naloxone, a medication that can revive people who have overdosed. Mr. Biden is the first president to support the approach.” The New York Times’ Noah Weiland reports on the Biden Administration’s pivot – helmed by drug czar Rahul Gupta – toward harm reduction approaches as it develops drug control policies.