Predicting lymphoma outcomes and risk factors in patients with primary Sjögren’s Syndrome using gradient boosting tree ensembles

Primary Sjogren's Syndrome (pSS) is a chronic autoimmune disease followed by exocrine gland dysfunction, where it has been long stated that 5% of pSS patients are prone to lymphoma development. In this work, we use clinical data from 449 pSS patients to develop a first, rule-based, supervised learning model that can be used to predict lymphoma outcomes, as well as, identify prominent features for lymphoma prediction in pSS patients. Towards this direction, the gradient boosting method combined with regression tree ensembles is used to derive a rule-based, decision model for lymphoma prediction. Our results reveal an average accuracy 87.1% and area under the curve score 88%, highlighting the importance of the C4 value, the rheumatoid factor and the lymphadenopathy factor as prominent lymphoma predictors, among others.