In order to avoid the arbitrary selection of genes, we selected all 20 oxidative stress-related SNPs available in the NAS database.
We construct an ensemble of B classification trees [ 5] using subsets of the genes we selected from the feature selection step.
From a list of genes, we selected those with increased or decreased changes of greater than 1.0 of Log2 ratio for comparison.
From the remaining genes we selected 258 genes of which the coefficient of bimodality was larger than 0.555.
For 11485 genes, we selected probes from the bottom of the clustering dendrogram from the 11-tissue dataset (see Table 1).
Two of the genes we selected to this study, LCT and APOBEC2, have not been previously associated with molecular pathogenesis of cardiovascular disease.
Among these genes, we selected survivin as one example, and confirmed its higher expression in SCAP by western blot (Figure 2A).
Among these promising genes, we selected five genes with non-related functions and spanning a range of expression levels: GSK3A, USP49, EEF2, ARHGEF2, and EIF2B2 (Table 1).
Of 50 genes we selected, 10 genes were not assayable for technical reasons (e.g., no SNPs or no RT-PCR product in brain).
The genes we selected fall into three groups genes involved in the Toll pathway, the IMD pathway and the Jnk pathway.
For 11485 genes, we selected the probes by cutting the clustering dendrogram near its bottom (h_final< = 0.4, see the definition of h_final in Materials and Methods).
