¡¡

Improving identification of differentially expressed genes in microarray studies with small number of replicates using information from public databases

¡¡

¡¡

Mingoo Kim^1,2¢Ó, Sung Bum Cho^1,2¢Ó and Ju Han Kim^1,2*

¹Seoul National University Biomedical Informatics (SNUBI),

²Human Genome Research Institute, Seoul National University College of Medicine, Seoul 110-799, Korea

¡¡

¡¡

1.  A supplementary table for GSE IDs of 58 reference datasets

¡¡

¡¡

2.  A supplementary figure for various top K genes parameters

¡¡

¡¡

¡¡

¡¡

¡¡

¡¡

¡¡

¡¡

¡¡

¡¡

¡¡

¡¡

¡¡

¡¡

¡¡

            In the supplementary Fig. 1, x and y axis indicate number of samples and number of true top 100 genes present in the differentially expressed  

            genes (DEGs) identified by each methods, respectively.

            Test methods are denoted with single letters: FMean, F; TStan, S; TReg, R; GGlobal, G; GPooled, P; GMixBayes, B.

¡¡

             Supplementary Figure 1.

¡¡

¡¡

3.  A supplementary figure for various prior confidence parameters

¡¡

¡¡

¡¡

             In the supplementary Fig. 2, x and y indicate the number of k and the number of top 50 genes present in the DEG by the benchmark methods.

             Supplementary Figure 2.

¡¡

4. BIC convergence graph

¡¡

¡¡

¡¡

¡¡

¡¡

5. Magnitude of improvement

    Magnitude of Improvement.

    This excel file contains magnitude of improvement of the GMixBayes method compared with other benchmark method. The average peformance of

    each method was presented with original performance and magnitude of improvement. The excel file also contains the original performance of

    GGlobal, GPooled, GMixBayes and regularized T test that was hybridized with the methods.

¡¡

6. Benchmark test of   GGlobal, GPooled and regularized T test

¡¡

    Similar graph with the Figure 4 of the Kim and Park's study (2006)

¡¡

    1) 2v2 comparison

    2) 3v3 comparison

    3) 4v4 comparison

¡¡

        4) 5v5 comparison