Bioinformatics Advances to Accelerate Omics Innovations and Applications in the Postgenomic Era

To the Editor:

We thank Li et al. (2017) for their comments on our study. We are pleased that they found our study useful and appreciate their observations and critical comments. We agree that high-throughput data sets can often be noisy and result in false positives. We also agree that one requires additional filters along with fold change thresholds to identify reliable candidates.

One concern as pointed out by Li et al. (2017) was that only selecting miRNAs with greater than twofold change in expression is not reliable and suitable for high-level microarray analysis. However, we indeed used statistical thresholds in addition to the fold change threshold in our analysis. The data were extracted using the Affymetrix miRNA QC tool using the robust multiarray average background correction (Irizarry et al., 2003) and median polish summarization. For each miRNA, background-corrected summarized data were obtained from the Affymetrix miRNA QC tool along with p values. To control the identification of false positives, a two-sample t-test was carried out for every miRNA and multiplicity correction was followed to control the false discovery rate. Candidate miRNAs that were increased or decreased in expression with fold changes ≥twofold after treatment with cigarette smoke condensate were considered to be differentially regulated.

The other concern raised by Li et al. (2017) is lack of further validation of differentially expressed miRNAs by RT-PCR or other methods. We agree that an orthogonal approach to verify results from microarray analysis is a valid approach. However, the sheer number of differentially expressed miRNAs identified in an unbiased global profiling study makes it impractical to validate all of them. We generally employ that strategy when specific molecule/s is/are to be selected for further functional evaluation. As this study was designed to catalog miRNA expression changes and protein expression changes associated with cigarette smoke exposure and provide it as a resource, we have not carried out further validation studies. However, we used mirTarBase (Chou et al., 2016), a database of experimentally proven targets of miRNA to correlate miRNA–protein pairs from our data sets to evaluate inverse relationship in their relative expression pattern. In addition, we also compared our data set with nonsmall cell lung cancer (NSCLC) miRNA data from The Cancer Genome Atlas (TCGA) with the smoker and nonsmoker patient history (TCGA, https://portal.gdc.cancer.gov/projects/TCGA-LUAD and https://portal.gdc.cancer.gov/projects/TCGA-LUSC). This comparison also showed many miRNAs that displayed altered expression in cigarette smoke-treated cells in our study, and showed similar expression pattern in tumors from smokers.

In summary, we believe that our results in this study using an integrated approach to analyze miRNA and protein expression profiles of NSCLC cells chronically exposed to cigarette smoke provide important insights and a resource of molecular perturbations on a global scale. However, we do acknowledge the suggestions by Li et al. (2017) that candidate molecules should be further verified by orthogonal approaches before planning future experiments to evaluate their functional significance.