As many readers will know, the 2024 Nobel Prize in Physics was awarded to professors John Hopfield and Geoffrey Hinton. Their groundbreaking work on artificial neural networks laid the foundation for modern machine learning, which over the past 10–15 years has transformed many aspects of society. Today, applications of their discoveries and inventions have become part of everyday life for billions of people around the world. Aided by digital tools, we are processing enormous amounts of data, both for work and personal use at speeds unthinkable just a few decades back.
Uppsala Monitoring Centre (UMC) was one of the early adopters of the technology. In the early 2000s, UMC’s initial work on syndrome detection, which later led to the development of the cluster analysis method vigiGroup, was based on a special type of Hopfield neural network. In this network, the weights of the nodes are information component (IC) values, like those used for disproportionality analysis. Whereas in disproportionality analysis, the IC is a measure of the disproportionality between an observed and expected number of reports for a certain drug-event combination, the syndrome detection neural network used it as a measure of the degree of association between patterns of adverse drug reactions. The abstract of the scientific paper that UMC published in 2005 concludes that:
With its better scaling properties, the neural network is a promising tool for unsupervised pattern recognition in huge databases of incomplete observations.
Niklas Norén, Head of Research at UMC, was part of the team that wrote the research article that led to vigiGroup. At the time, however, he was working for a small consultancy company developing data mining methods on UMC’s behalf.
You were in touch with Professor Hopfield during that work, is that right?
Yes, very briefly. I emailed him about a suspected sign error in an equation, which was causing me grief.
And …?
This was in his second most cited paper, so it took me a couple of days to work up the courage to write that email. But he was quick and courteous in his reply, apologising for the error. And then our calculations worked out.
So, in other words, you’ve corrected a Nobel Prize laureate?
Ha ha, I’d say that I rather ‘discovered a typo in a paper written by a Nobel Prize laureate’. I should probably add that to my resumé!
UMC launched the vigiGroup algorithm almost 10 years ago as a method to identify and group reports with similar adverse event profiles. This approach is different to more classic methods for signal detection, which focus on one adverse event term at a time. By processing the full set of reported terms, vigiGroup offers a complementary, new way of evaluating safety data.
From a pharmacovigilance perspective, the discoveries of Professors Hopfield and Hinton have not only provided a new tool, but opened the doors to an entire toolkit for making medicines safer for everyone, everywhere.
Read more