Machine learning, or AI, used in the field of medical research has helped advance several frontiers. Predictive analytics for disease prevention, diagnostics and medical imaging, along with drug discovery are among the many promising areas of application for these new AI tools. 

Some people may take it as a comfort that this technology is being used for good, while others may find it disconcerting. A 57% majority of Americans say AI is bad for society. Without federal regulation on AI it continues to develop, and is being integrated into most other technologies, regardless of the public’s opinion toward it. 

One of the most famous cases of machine learning breaking into a medical frontier was the announcement of AlphaFold2 by Nobel Prize winner John Jumpers, at the 2020 Protein Folding Dynamics Gordon Research Conference. For over three decades scientists have been gathering together to solve the famously impossible Protein Folding Problem: what is the simplest way to predict the three-dimensional shape of a protein based only on its one-dimensional molecular code? AlphaFold2 was demonstrated to predictively model the folding of protein structures with over a 90% accuracy, a five time increase from its closest competitor. This came as a shock to many in the field, for some had dedicated their entire careers to discovering incrementally better prediction methods. 

In the field of predictive analytics for disease prevention, machine learning can help identify health risks before symptoms emerge. According to the National Institute of Health (NIH), “AI can evaluate cardiovascular risk using data from routine blood tests and lifestyle surveys, providing clinicians with alerts that support early interventions, such as prescribing statins or recommending dietary modifications.” This could ultimately help the healthcare system shift further toward pre-emptive care, over reactive treatment. 

In the field of diagnostic and medical imaging, machine learning can detect abnormalities with more sensitivity than humans. “Evidence indicates that AI algorithms generally achieve higher sensitivity ranging from 56.4% to 95.7% compared to radiologists, whose sensitivity ranges from 23.2% to 76% while maintaining comparable specificity,” according to the NIH. Perhaps with the widespread integration of these models, more lives can be saved through even earlier detection. 

In the field of drug discovery, machine learning has accelerated multiple of the lengthy steps in pharmaceutical research. According to the NIH, “By integrating genomics, proteomics, and molecular structure data, AI can identify disease-associated processes and promising therapeutic targets. Once identified, AI supports high-throughput virtual screening of large chemical libraries to predict binding affinities and prioritize compounds for testing.” Time consuming data analysis can be done much faster and then be reviewed by medical scientists to confirm viability for testing. 

One of several concerns with widespread integration is the bias observed in healthcare AI. Research on humans still mostly relies on heavily biased data with most trial participants fitting the western, educated, industrialized, rich, and democratic (WEIRD) profile. When models are trained on this data set they tend to reinforce those patterns of bias and do not achieve a more globalized approach. 

The NIH acknowledges these ethical issues but states, “to properly evaluate such influences, future studies must incorporate standardized metadata on factors that may introduce bias.” In other words, they’re working on it.  

With AI taking society by storm, there are many considerations affecting how people feel about its weight in the world. This includes the environmental impact of large data centers; changes to the job market for entire industries; and at the individual level, the blow to cognitive critical thinking skills. It may be comforting to know that some of the brightest scientific minds have tasked themselves with harnessing this revolutionary technology for good.