T-Rex-293 cells stably expressing a plasmid encoding GFP were induced with 2.5ug/mL tetracycline for 72 hours for expression of either an empty vector, an shRNA specific to GFP, or an shRNA specific to Tat-SF1.
Growth protocol
T-Rex-293 cells stably expressing a plasmid encoding GFP were grown in DMEM (high glucose) with 10% tetracycline-tested FBS, plus 1% L-glutamine plus 1% pen/strep.
Extracted molecule
total RNA
Extraction protocol
Total RNA was extracted using the RNeasy kit (Qiagen), followed by DNase treatment (Ambion).
Label
biotin
Label protocol
Labeling was performed according to the Affymetrix (Santa Clara, CA) GeneChip Whole Transcript (WT) Sense Target Labelling Assay manual.
Hybridization protocol
Microarrays were hybridized overnight with 5 mg biotin labelled ss-cDNA, and washed in fluidics station wash protocol: MES_EukGE-WS2v5_450.
The input files were normalized with full quantile normalization (Irizarry et al 2003). For each input array, for each probe expression value, the array ith percentile probe value was replaced with the average of all array ith percentile points. Next, the 6,553,590 probes were manipulated into the analysis values as follows. Probes with GC count less than 6 and greater than 17 were excluded from the analysis. Probe scores were then transformed by taking the Natural Logarithm of 0.1 plus the probe score. Background CorrectionExon arrays do not use individual mis-match probes. Background is established from a pool of probes designed for that purpose. Background probes are stratified by CG content and are defined in the HumanExon10ST_antigenomic.bgp file. BGP files can also be downloaded from www.affymetrix.com. Each probe score was corrected for background by subtracting the median expression score of background probes with similar GC content. The HumanExon10ST array contains 1,404,693 probe-sets (typically, but not always, groups of four probes). Probe-set Expression Scores and Annotation FilteringThe expression score for a probe-set was defined to be the median of its probe expression scores and probe-sets with fewer than 3 probes (that pass all of the tests defined above) are excluded from further analysis. Exon Array Probes are designed off of genomic sequence and hence the reliability of probes and probe-sets correspond to the quality of their parent genomic annotations. Probe-set reliability is ranked from more to less reliable as Core, Extended, or Full. For example 'Core' probe-sets include probe-sets that correspond to high quality genomic features like RefSeq (www.ncbi.nlm.nih.gov) or Ensembl (www.ensembl.org) transcripts while 'full' and 'extended' probe-sets match less reliable annotations like EST hits and gene prediction algorithms. For this analysis, only 'Core' probe-sets were analyzed.Probe-set Presence/Absence and the Removal of Non-expressed Probe-setsNon-expressed probes can cause tests for alternative splicing to find false positives (because they cause 'non-parallel' expression patterns across the gene). A probe-set is judged to be expressed above background if for any group: Integral from T0 to Infinity of the standard normal distribution less than Significance (0.001) Where T0 equals Sqr(GroupSize) (T minus P) divided by Sqr(Pvar), GroupSize equals Number of CEL files in the group, T equals average of probe scores in probe-set, P equals average of background probes averages of GC content, and Pvar equals sum of background probe variances divided by (Number of probes in probe-set) to the second power. Hence, we test that the average of probe-sets in a group is greater than the average expression of background probes of similar GC content as the probe-set probes as the center of background for the probe-set and derive its dispersion from the background probe-set variance. Low-variance probe-sets are excluded from the analysis via a Chi-Square test. A probe-set is considered to be low-variance if its transformed variance is to the left of the 90 percent confidence interval of the Chi-Squared distribution with (N-1) degrees of freedom.(N-1) * Probe-set Variance / (Gene Probe-set Variance) ~ Chi-Sq(N-1) where N is the number of input CEL files, (N-1) is the degrees of freedom for the Chi-Squared distribution, and the 'probe-set variance for the gene' is the average of probe-set variances across the gene. Although, in practice, this method works well, it should be noted that the Chi-Square test of variance is usually applied to test a variance against a constant value and we are using it to test probe-set variance against a random variable (probe-set variance across gene) ; furthermore, the probe-set and probe-set across gene are not independent.
