Transcriptomic module fingerprint reveals heterogeneity of whole blood transcriptome in type 1 diabetic patients
BackgroundType 1 diabetes (T1D) is a complex autoimmune disease resulting from {beta}-cell destruction in the pancreas. Yet, no gene signature derived from whole blood samples has been established to differentiate T1D patients from healthy donors, likely because of the high heterogeneity of the underlying pathophysiological mechanisms. MethodsWe analysed whole blood transcriptomic profiles from 39 patients and 43 healthy donors, collected as part of our observational clinical trial, using a multi-step multivariate statistical analysis. This approach combined classical differential analysis, random forest and support vector machine classifications, gene set functional enrichment analysis to construct the most stable and reliable gene signature. ResultsClassical differential analysis did not separate clearly samples into healthy and T1D clusters, but rather spread samples into three clusters. On the contrary, we show that our approach, which combined molecular signatures independently constructed, adds robustness to the analysis without compromising the specificity. This efficiency was demonstrated by the clear separation of the samples according to their diagnosis group. Also, the functional annotation of the gene modules that we obtained was more associated with T1D-related pathways compared to the classical statistical analysis. ImpactsThese results emphasize that single differential analyses are not able to capture immune continuums involved in such a complex pathophysiological process. We hypothesize that T1D patients can have different molecular pathways involved in their pathology or/and can display unsynchronized -omics profiles. These findings call for further investigations to identify the molecular pathways involved in T1D patients, and for revising the nosology of autoimmune diseases.