Reply to Toker and Pavlidis: Blood biomarkers for Parkinson's disease.

We are grateful for the opportunity to respond to the comments of Toker and Pavlidis (1) regarding our recent publication, in which hepatocyte nuclear factor 4 alpha (HNF4A) and polypyrimidine tract binding protein 1 (PTBP1) mRNAs are identified as potential blood biomarkers for Parkinson’s disease (PD) (2). Early diagnosis of PD remains challenging. Contrary to what the authors suggest, we are cognizant that several markers will be required to accurately identify earlystage PD patients. Our previous studies identified several markers, most of which were replicated in two independent clinical trials (3, 4). It is our view that a combination of RNA and biochemical markers in blood and cerebrospinal fluid, for example, will be required for accurate diagnosis. We propose that network analyses provide us with a highly beneficial tool for identifying biologically relevant markers (5). We expect that these markers will complement, not replace, a neurological examination. Contrary to the claim by Toker and Pavlidis (1), HNF4A and PTBP1 were identified as candidate biomarkers by both transcriptomic metaanalysis in Integrative MetaAnalysis of Expression Data (INMEX) and network analysis using NetworkAnalyst, as described (2). HNF4A and PTBP1 were identified according to the metaanalysis and network topology measures. Toker and Pavlidis also claim that the dataset GSE22491 is an outlier. To address this concern, each individual dataset was visualized in box plots to ensure identical distribution among the samples and identify potential outliers as described in our report (2). Nonetheless, we repeated the metaanalysis excluding the GSE22491 dataset. Contrary to the findings of Toker and Pavlidis (1), differential expression of HNF4A (P = 0.01) and PTBP1 (P = 0.02) remained significant. Of note, Toker and Pavlidis (1) used R instead of INMEX, different settings for quality control and datasets from Gemma instead of those provided by GEO, as described in our study (2), likely accounting for the discrepancies identified in their analysis. Microarrays and network analyses are both susceptible to false-positives and technical issues may arise when datasets used for the analysis have not been “preprocessed” properly. This is why researchers use highly sensitive methods to quantify gene expression to verify the findings. In this case, both markers were confirmed in blood samples from two independent clinical studies (n = 192) using quantitative PCR (2). Therefore, the findings presented in this study are indisputable. We believe that this study is a major step toward the translation of HNF4A and PTBP1 into the clinic. As stated previously, evaluation of these markers in a larger prospective study will be important to assess their clinical utility as a diagnostic tool for PD (2).