Purpose
About 30–40% of patients with diffuse large B-cell lymphoma (DLBCL) relapse or fail to respond to first-line treatment. The molecular heterogeneity is considered to be the main factor affecting the therapeutic response of DLBCL. The existing classification methods can not fully explain these heterogeneity, so we try to explain DLBCL heterogeneity by defining DLBCL subtypes from the perspective of metabolism.
Methods
In this study, we integrated five DLBCL data sets (GSE10846, GSE11318, GSE53786, GSE87371 and GSE23501) (n = 742) from geo database, screened 106 metabolic related genes (MAD > 0.5, Cox P < 0.001), and identified dlbcl2 subclasses (nmftype1, nmftype2) by non-negative matrix factorization clustering (NMF).
Results
nmftype1 showed low metabolic activity ,while nmftype2 showed high metabolic activity. Compared with the two subtypes of immune infiltration, it was found that nmftype1 was mainly infiltrated by B cells, and nmftype2 was mainly infiltrated by T cells and macrophages, and the high expression of nmftype2 was more in immune checkpoint. The difference of metabolic subtype OS was statistically significant, and the overall survival (OS) of nfmtype1 was worse than that of nmftype2. The combination of metabolic subtypes and ABCGCB subtypes can predict the prognosis of DLBCL patients better than the existing ABCGCB subtypes. Finally, 34 gene classifiers were identified. The consistency results were verified by GSE31312 (n = 470), and a new classification of DLBCL based on metabolic gene expression profile was established.
Conclusions
We have obtained a new DLBCL typing method, which has prognostic significance. It has a certain correlation with immune escape and can guide individualization application of immunotherapy and metabolic therapy.