Exploring the Limitations of Peripheral Blood Transcriptional Biomarkers in Predicting Influenza Vaccine Responsiveness

Figure 1

Graphical representation of the analysis pipeline implemented to compute biomarkers (see Materials and Methods). The analysis started with the dataset related to the 2008/09 TIV vaccination campaign (set #2). After a preliminary preprocessing of the dataset for removing subjects with intermediate answer to the vaccine (“NA subjects,” maximum across strains equal to 4) and for computing gene fold changes (day 3/day 0 and day 7/day 0), we applied the classification algorithm SCUDO to obtain the biomarker. The classification process on the training dataset exhibited 100% accuracy with a 10-fold cross-validation scheme. Moreover, its statistical significance was assessed by a permutation test providing a value < 0.001. Finally, the identified biomarker was validated using an independent dataset (2007/08 TIV vaccination campaign, set #1). The classification on the validation dataset exhibited 78% accuracy.