Favored by nerdy bettors, “in-game win probability” uses a continuous stream of various data to predict, in near real time, the likely outcomes of contests. Maximillian Diehn and Ash Alizadeh and their Stanford colleagues Mohammad Shahrokh Esfahani and David Kurtz used a version of the method to create a computer algorithm called Continuous Individualized Risk Index (CIRI) to assess how well a cancer patient is doing at a given point during treatment. The method, which incorporates information on such things as treatment responses and circulating tumor DNA, could help identify patients in need of more aggressive therapy. The researchers trained their algorithm to detect patterns associated with whether a patient lived for at least 24 months after treatment without a relapse using data on more than 2,500 people treated for diffuse large B-cell lymphoma (DLBCL), a common blood cancer. They also included data on 132 patients whose circulating tumor DNA had been monitored. If the value of 1 constitutes a perfect predictive score, CIRI’s dynamic risk profiling yielded a score of 0.8 vs. 0.6 for existing methods. Their paper, published in July in Cell, also showed how CIRI might be used to identify new biomarkers for risk profiling.
This article appeared in the November 2019 issue of Ludwig Link. Click here to download a PDF (1 MB).