The package provides a list of options for classifier gene set, cross-platform normalization method and data set to train the model along with reasonable default settings.
Probesets were mapped to the human genome in the annotate and hgu133plus2.db software packages.> The COMBAT (empirical Bayes [ 33]) batch correction and cross-platform normalization method which is implemented in the inSilicoMerging package [ 34], was used to merge all normalized microarray datasets into one global merged dataset.
As Figure 3 shows, cross-platform normalisation methods bring the data significantly closer together, compared to simple standardisation.
We compared four cross-platform normalization methods, Z-score, Rank, XPN and DWD.
Another recently published comparison of cross-platform normalization methods also found XPN to have the highest inter-platform concordance [ 37].
Z-score, rank and two more sophisticated methods, XPN and DWD [ 47, 48] implemented in the CONOR package [ 49], were used to examine the effect of different cross-platform normalization methods.
A number of batch-correction and cross-platform normalisation methods were evaluated, including mean centering [ 5], ComBat [ 21], Distance Weighted Discrimination [ 20] and cross-platform normalisation (XPN) [ 22] in order to determine the most effective method for reducing the bias imposed by the different platforms.
A similar test of the training set effect on GSE4475_p1 is shown in Additional file 4. Following the above investigations, we trained a 28 gene-based SVM, the BL and DLBCL classifier BDC, on the GSE4475 data set with a BL probability threshold of 0.95, and applied it to our Illumina data sets (Table 1) using several cross-platform normalization methods.
Good overall cross-laboratory, cross-platform, and counting method agreement were achieved in CCQM-P102, with mean CD4+ cell counts in good agreement.
There was good overall cross-platform and counting method agreement with a grand mean of the laboratory calculated means of (301.7 ± 4.9) μL−1 cellscells.
To merge the microarray datasets measured with multiple platform chips, we selected genes from all platforms based on the NIH Entrez Gene ID and used the Cross-Platform Normalization (XPN) method of Shabalin et al. [ 40] implemented in the R package: "CONOR" [ 41].
