Subtype classification and prognosis of diffuse large B-cell lymphoma based on variable importance analysis

Abstract

Abstract Gene expression profiling (GEP) is considered as gold standard for cell-of-origin (COO) classification of diffuse large B-cell lymphoma (DLBCL). However, the high dimensionality of GEP limits its application in clinical practice. In this study, we aim to develop a parsimonious model based on GEP to accurately predict COO subtype of DLBCL for clinical applications. We first proposed a variable important measure to solve the instability of penalized regression methods in high-dimensional settings. Then this strategy was applied to six penalized methods to identify a small gene subset for DLBCL classification. Using a training dataset of 350 DLBCL patients, we developed a six-gene model for DLBCL classification. For training and validation datasets, model evaluations showed the six-gene model performed well in terms of discrimination, calibration and clinical usefulness. Subgroups of patients characterized by these six genes showed significantly different prognosis. Furthermore, model comparisons demonstrated that the six-gene model outperformed models constructed by typical penalized regression methods. In conclusion, the six genes had considerable clinical usefulness in DLBCL classification and prognosis. The genes order based on their importance provided a priority for further functional and targeted drug research.