gender: male strain: C57BL/6J genotype/variation: Scd1 knockout (SKO) tissue: Dorsal skin age: 8-9 wks old
Treatment protocol
Male mice were sacrificed in the non-fasted state by isoflurane overdose, shaved with electric clippers, and whole dorsal skin removed and flash frozen in liquid nitrogen
Growth protocol
Male Lox and SKO mice on a C57BL/6J background were weaned at 3 weeks of age and maintained on a standard rodent diet (Purina Formulab 5008) until 8-9 weeks of age
Extracted molecule
total RNA
Extraction protocol
RNA was extracted from skin with TRI reagent (Molecular Research Center), DNase treated with Turbo Dnase (Ambion), and re-extraced with TRI reagent to inactivate Dnase
Label
Biotin
Label protocol
Samples were labeled using the Ambion MessageAMP Biotin II-Enhanced IVT kit using 1.5 ug of RNA for all samples and processed according to kit instruction manual. 14 hour IVT reaction at 37C.
Hybridization protocol
Hybridization cocktail prepared acording to the protocols and procedures in the AFX Expression Analysis Technical Manual. 10 ug of aRNA applied to AFX array and hybridized overnight at 45C for 16 hours in AFX 640 Hybridization oven. AFX 430 2.0 gene chips were post processed on the AFX 450 Fluidics Station according to all AFX protocols and procedures defined for the array (EukGE_WS2v5_450)
Scan protocol
All AFX 430 2.0 gene chips scanned on AFX GC3000 G7 scanner using Command console. Data extracted from scanned images using Expression console. All recommended AFX control parameters ok.
Description
Gene expression data of whole skin
Data processing
Expression measurements were pre-processed to provide background correction, normalization and log base 2 transformation using RMA (Robust Multi-array Average) (Irizarry RA, et al. Biostatistics 4: 249-264, 2003). We used two analytical approaches to generate lists of differentially expressed probe sets, loosely referred to as genes. First, we calculated an un-moderated t-statistic for every probe set by applying Welch’s t-test. The resulting p-values were used to calculate q-values, which account for multiple tests and provide thresholds so that lists can be generated for a target false discovery rate (FDR) (Storey JD, and Tibshirani R. Proc Natl Acad Sci U S A 100: 9440-9445, 2003). Using this method, we targeted a FDR of 5% and probe sets with q-values below 0.05 are considered to be of interest. We also applied EBarrays (Kendziorski CM, et al. Stat Med 22: 3899-3914, 2003; Newton MA, et al. J Comput Biol 8: 37-52, 2001), which along with RMA is implemented in R, a publicly available statistical analysis environment (R. Development Core Team. R: A language and environment for statistical computing. http://www.r-project.org. Vienna, Austria: R Foundation for Statistical Computing, 2005) and available at Bioconductor (Gentleman RC, et al Genome Biol 5: R80, 2004.). EBarrays is an empirical Bayes approach which models the probability distribution of a set of expression measurements (Kendziorski CM, et al. Stat Med 22: 3899-3914, 2003, 50). It accounts generally for differences among probe sets in their true underlying expression levels, measurement fluctuations and distinct expression patterns for a given probe set among conditions (Kendziorski CM, et al. Stat Med 22: 3899-3914, 2003). An expression pattern is an arrangement of the true underlying intensities (u1) in each condition. The number of patterns possible depends on the number of conditions from which the expression measurements were obtained. For example, when measurements are taken from two conditions, two patterns of expression are possible: equivalent expression (EE; u1 = u2) and differential expression (DE; u1 ≠ u2). Since we do not know a priori which probe sets are in which patterns, the marginal distribution of the data is a mixture over the possible patterns with model parameters determined by the full set of array data. In this way, the approach utilizes information across a set of arrays to optimize model fit and is thus more efficient than a number of methods that make inferences one probe set at a time (Kendziorski CM, et al. Stat Med 22: 3899-3914, 2003). The approach also naturally controls for both type I and type II errors (Kendziorski CM, et al. Stat Med 22: 3899-3914, 2003). The fitted model parameters provide information on the number of probe sets expected in each expression pattern. Furthermore, the fitted model is used to assign posterior probability distributions to every probe set. Each probe set specific distribution gives the posterior probability of that probe set's individual expression pattern. The posterior expected FDR is controlled by thresholding the posterior probabilities. Two approaches were used as discussed in (Newton MA, et al. Detecting differential gene expression with a semiparametric hierarchical mixture method. Biostatistics 5: 155-176, 2004.). The most conservative approach to control the FDR at 5%, for example, is to only consider probe sets with specific posterior probability of EE less than 0.05 (hard threshold). We also report results from the less conservative approach that determines the exact threshold required (soft threshold) so that the average posterior probability of EE for all probe sets on a list is less than 0.05.
