Bioinformatics analyses of NanoString data

Raw reads were assessed for technical quality control flags using the nSolver software (NanoString) and transferred into RStudio (Boston, MA, USA). As the design of our probe set was nonrandom, we utilized a custom approach for data analysis as previously described to account for large differences in the number of observed counts between mock-infected and infected samples (114). Briefly, we used count data within each sample vector as an internal reference, against which we compared genes of interest (GOIs). We identified reference genes by first measuring the coefficient of variation (CV) for all genes in the probe set, which included GOI, housekeeping (HK) genes, and positive and negative controls. We then used the calculated CVs to select the least variable HK genes. To obtain our within-sample reference value, we calculated the geometric mean of the selected HK genes. We then “anchored” all count data from each sample to its within-sample reference in ratio form to derive anchored gene counts (AGCs). To calculate the log2 fold change, we formed a ratio using the AGC for each GOI against its values from mock infection and took the log2 of these values, which were used to generate PCA plots and line graphs.