Next-Generation Sequencing Reveals HIV-1-Mediated Suppression of T Cell Activation and RNA Processing and Regulation of Noncoding RNA Expression in a CD4(+) T Cell Line

Chang, Stewart T.; Sova, Pavel; Peng, Xinxia; Weiss, Jeffrey; Law, G. Lynn; Palermo, Robert E.; Katze, Michael G.

Differential expression analysis. To compare transcript expression across different conditions, we first normalized transcript abundances by the following methodology. Transcript abundance was quantified as FPKM (fragments per kilobase of exon per million mapped fragments) values estimated by Cufflinks. We chose one sample arbitrarily as a reference. Distributions of log 2 -transformed FPKM values between the reference and remaining samples were compared by quantile-quantile plots. We determined the scaling factor for each sample as the median difference of the corresponding quantile values of the sample and reference. Only genes/transcripts with raw FPKM values of Õ†1 in all samples were considered in the estimation of scaling factors. We retained those genes/transcripts with nonzero FPKM values in 100% of the samples of at least one biological condition (our detection criterion). An offset of 1 was added to all normalized values to facilitate the comparisons involving one or more FPKM values of zero and to reduce the variability of the log ratios for low expression values. Transcripts were mapped to RefSeq gene loci, resulting in 9,992 loci with detectable reads. The data are available at http://www .viromics.washington.edu or upon request. The normalized expression data were analyzed for differential expression by using linear model methods as implemented in the R package limma (46) . P values were derived from linear model-based t tests between infected and time-matched mock-infected samples. Unless otherwise noted, we defined differential expression by Benjamini-Hochberg-adjusted P values of less than 0.05 based on the assumption that a false discovery rate of 5% provided an acceptable balance of false-positive control and statistical power. Fold changes (FC) were derived from comparing the means of these groups, and multiple groupings of fold changes were used (1.0 to 1.5 FC, 1.5 to 2.0 FC, and 2.0Ï© FC) based on previous observed fold change ranges observed in high-throughput and in particular NGS data in virus-infected systems (8) .