e expression analysis platform. Additional analyses were performed using the R statistical tools framework. Gene Expression Omnibus The experimental design and the individual CEL files for the study have been deposited in the Gene Expression Omnibus, under data set GSE44903. 14 Gene Expression Profiling of Epileptogenesis Model Data standardization When required, standardization of profiles was performed for each probeset separately, by first log2-transforming the MAS5 intensity values for that probeset, then subtracting from the result the mean and dividing 
by the standard deviation of the log2transformed intensities across all experimental conditions. genes would be expected for two sets of comparable size chosen at random. The union of the two sets contained a total of 1,399 genes. Regulated KS analysis for gene set enrichment analysis of the hypoxia response Gene sets were scored for enrichment against the entire profile of log2-ratios of hypoxia to control intensities for 14,405 genes, using an algorithmic extension of the Kolmogorov-Smirnov test which we have denoted “regulated KS analysis”. In computing enrichment, the regulated KS analysis accounts for the sign of regulation of each gene as specified in the input gene set, as well as its identity. The algorithm generates a P-value, “left”and “right”enrichments scores CL and CR, and a distributional plot of log2 ratio values for gene set members relative to the profile being investigated. Briefly, consider a query gene set representing a biological pathway, with ku 11741201 genes known to be positive regulatees and kd genes known to be negative regulatees of the pathway, the gene set containing a total of k = ku + kd genes. For a given target profile of n expression values, for which enrichment of the query gene set is to be determined, the expression values are first transformed into ranks, with rank = 1 MRT-67307 corresponding to the largest value and rank = n to the smallest value in the profile. The empirical cumulative distribution functions of the ranks of the ku and kd positive and negative regulatees of the gene set are then independently estimated, and the maximum deviations du and dd from the CDF for a uniform distribution are computed. A single test statistic combining du and dd is then formed, dUpDown ~ ku du {kd dd: ku zkd Expression ratio estimation Expression ratios between two conditions with multiple replicates available for each condition were estimated using the Pfold algorithm, a robust Bayesian estimator which takes as input the mean intensity values for each condition, and the corresponding standard deviation of measurement of the mean intensities. The corresponding P-values were computed using an equal-variance ttest. Ratios were first computed at the probeset level, then mapped to genes; when a gene mapped to multiple probesets, with corresponding multiple ratios and P-values available, the ratio and P-value corresponding to the smallest P-value were assigned to that gene. Mapping of probesets to genes Mapping of Affymetrix probesets to genes was implemented using a correspondence file derived from the Rat230_2.na27.annot.csv library file available from the Affymetrix NetAffx service site. The resulting mapping is many-to-many, with generally each gene mapping to multiple probesets and occasionally one probeset mapping to multiple genes. In the assignment 14707029 of intensities, when a gene corresponded to multiple probesets, the largest intensity available was assigned as that gen