Teva sport sandals, a hot glue gun, and the new J.K. Rowling story “Harry Potter and the Cursed Child”. These are the items recommended for me when I log into Amazon.com. Netflix thinks I’d like to watch Marvel’s Daredevil, Chelsea, and Orange is the New Black. Not bad suggestions, really. But how did they know? These companies, and many others, use recommender systems to guide their customers towards more purchases or content. These systems are like filters, sifting through giant amounts of information in order to predict a person’s preferences.
This is the goal of personalized medicine. We need a way to sift through giant amounts of biological and laboratory data to predict the best course of treatment for an individual patient. This is especially challenging in lupus. No two patients are the same. Their symptoms can be wildly different, as can their responses to treatment. The lupus drug discovery pipeline is more promising than ever, but we have yet to figure out how to decide which treatment option is best for a given patient. Rheumatologists do the best they can with the information available to them, but we just don’t have a solid recommender system for lupus patient care.
A report1 published last month in the journal Cell might represent one step towards personalized medicine for lupus. In this study, the giant amount of data to be sifted through was comprised of more than 80 million datapoints from patients’ blood samples. The sifter itself was a combination of statistical analysis and a software program they previously developed. Their goal was to find the useful bits of laboratory data that can help explain different clinical subtypes of the disease, and to filter everything else out. In the end, they found that the useful bits of data could separate patients into seven major groups. Within each patient group, different data features were correlated with lupus disease activity. One possible interpretation of this finding is that different factors cause worsening disease activity in the different patient groups. If this is the case, the authors suggest that the best therapeutic target for a given patient might depend on his or her group assignment.
Of course, further studies are needed to determine whether this sifting method, or similar strategies, could be the basis for a recommender system predicting the best treatment strategy in lupus. But I think it’s a step in that direction, and it’s worth staying tuned, which I plan to do. Right after I finish watching Season 2 of Daredevil.
1Banchereau R, Hong S, Cantarel B, et al. 2016. Personalized Immunomonitoring Uncovers Molecular Networks that Stratify Lupus Patients. Cell 165(3):551-65.