Exploring DNA methylation data for diagnostic classification of Diffuse large B-cell lymphoma in Dogs

Abstract

Abstract Diffuse large B-cell lymphoma (DLBCL) is a common B-lymphocyte tumor in dogs, making up 60-70% of cases. We assessed the utility of DNA methylation data for the diagnostic classification of DLBCL in dogs. We also assessed the utility of the classification features identified in cDLBCL for diagnostic classification of DLBCL in humans. The GSE94913 cDLBCL DNA methylation dataset from the Gene Expression Omnibus (GEO) was used for analysis. Differential methylation analysis was performed between the 37 cDLBCL and seven control lymph node samples in the dataset. 1701 differentially methylated probes were identified between the cDLBCL and control lymph nodes groups. Applying recursive feature elimination on the 1701 significant probes, 20 probes were selected for machine learning classification tasks. The methylation values of these 20 probes were used to build an SVM model and create the training and testing set. 100% of the test samples were accurately classified by the SVM model. The diagnostic classification utility of the identified differentially methylated CpGs/CDS was also assessed in humans using the GSE28094 human DLBCL dataset. 95% of 98 DLBCL and leukocyte samples obtained from this dataset was correctly classified using clustering techniques on 11 CpG sites of 5 genes (ERBB4, IGF2, PGF, PITX2, TJP1). The utility of DNA methylation data for the diagnostic classification of DLBCL in dogs is demonstrated. Further exploration of this data type for potential biomarker discovery in cDLBCL is necessary.