Indeed, the study authors pointed out that applying their algorithm to a cell-of-origin or molecular genetic factors will be the next step in improving the model’s prognostic strength. “It is our hope that in the future, prognostic modeling for DLBCL, and other diseases, will be based on a combination of clinical, pathologic, and genetic information, and will make use of modern predictive modeling approaches rather than use discrete risk scores on the basis of a limited number dichotomized clinical variables,” they wrote.

Incorporating genomic and other molecular data into a single predictive algorithm could present a significant technical challenge, said Lee Cooper, PhD, assistant professor of biomedical informatics at Winship Cancer Institute at Emory University, who was also not involved in the study. “The issue with genomic data is that [they] contain so many variables,” Dr Cooper wrote via email. “The more variables you have, the more difficult it is to build accurate models. You need a very large number of samples, and conventional statistical approaches often do not work.”

Continue Reading

Additionally, different ethnic groups may require different risk calculations. For instance, Dr Flowers noted, African American patients in the US are frequently diagnosed with DLBCL at a younger age and more commonly have advanced-stage disease, among other factors. Thus, predictive models would need to be trained on cohorts that are representative of each population of interest.

Related Articles

Still, Dr Flowers added, “this model is a step in the right direction, with the right type of clean, rich data for training.”

Although the tool is available online at, the website bears the warning that the information presented therein should be used for research purposes only.


  1. Biccler JL, Eloranta S, de Nully Brown P, et al. Optimizing outcome prediction in diffuse large B-cell lymphoma by use of machine learning and nationwide lymphoma registries: a Nordic lymphoma group study [published online October 11, 2018]. JCO Clin Cancer Inform. doi: 10.1200/CCI.18.00025