Methodology article
Discovering biological connections between experimental conditions based on common patterns of differential gene expression
1 Bioinformatics Program, Boston University, 44 Cummington Street, Boston, Massachusetts, 02215, USA
2 Division of Computational Biomedicine, Department of Medicine, Boston University School of Medicine, 72 East Concord Street, Evans 631, Boston, Massachusetts, 02118, USA
BMC Bioinformatics 2011, 12:381 doi:10.1186/1471-2105-12-381
Published: 27 September 2011Additional files
Additional file 1:
A gene expression signature of E2 treatment. The expression of these 111 genes was previously reported to be increased or decreased > 2.5-fold by the treatment of MCF7 cells with 10 nM 17-β-estradiol (E2) for either 8 or 48 hours (or both) relative to treatment with vehicle[20].
Format: XLS Size: 44KB Download file
This file can be viewed with: Microsoft Excel Viewer
Additional file 2:
Results of openSESAME query using E2 treatment signature. In an openSESAME query using the signature of E2 treatment, these 56 GEO series were assigned an FDR q < 0.25 with either the K-S or Fisher's exact tests. Series are sorted in ascending order by K-S q value.
Format: XLS Size: 471KB Download file
This file can be viewed with: Microsoft Excel Viewer
Additional file 3:
A gene expression signature of p63 silencing. The expression of these 77 genes was previously reported to be increased or decreased > 2-fold in at least three of five cell lines treated with an shRNA targeting p63, relative to the same cell lines treated with an shRNA targeting GFP[28].
Format: XLS Size: 39KB Download file
This file can be viewed with: Microsoft Excel Viewer
Additional file 4:
Results of openSESAME query using p63 silencing signature. In an openSESAME query using the signature of p63 silencing, these 123 GEO series were assigned an FDR q < 0.25 with either the K-S or Fisher's exact tests. Series are sorted in ascending order by K-S q value.
Format: XLS Size: 485KB Download file
This file can be viewed with: Microsoft Excel Viewer
Additional file 5:
Receiver Operating Curves (ROCs) using E2 treatment signature in Connectivity Map (CMap) MCF7 dataset. An openSESAME query was performed using the original 189-Affymetrix-probeset E2 treatment signature on the ranked fold changes from all instances of treatment of MCF7 cells in the CMap build 2.0 dataset. A ROC was constructed (orange) in which instances of treatment of MCF7 with 17-β-estradiol (E2) were considered true positives. Another ROC was constructed (purple) using the S scores from a web query of the Connectivity Map using the same signature.
Format: PDF Size: 176KB Download file
This file can be viewed with: Adobe Acrobat Reader
Additional file 6:
Variation of openSESAME p values with sample size in GEO series GSE2225 and GSE21653. A, B. For each sample size, 1000 subsets of each GEO series were obtained by permutation and SA scores were computed. Fisher's exact test or a two-sided Kolmogorov-Smirnov (K-S) test were used to compute p values for each permutation. C, D. The expression values of each gene were shuffled independently 100 times, and for each shuffled dataset, 10 subsets were obtained for each sample size and SA scores and p values were computed. E, F. A total of 100 simulated datasets were obtained by generating random values from a standard normal distribution and z-normalizing each row ("gene") across all columns ("samples"). For each simulated dataset, 10 subsets were obtained for each sample size and SA scores and p values were computed. Below all panels, the fraction of permutations with p values below each threshold is shown.
Format: PDF Size: 32KB Download file
This file can be viewed with: Adobe Acrobat Reader
Additional file 7:
Variation of openSESAME p values with signature size in GEO series GSE2225 and GSE21653. A, B. For each subset size, 1000 subsets of the signature genes were obtained by permutation, maintaining the same proportion of up- and down-regulated genes in the original signature, and SA scores were computed using each GEO series. Fisher's exact test or a two-sided Kolmogorov-Smirnov (K-S) test were used to compute p values for each permutation. C, D. The expression values of each gene were shuffled independently 100 times, and for each shuffled dataset, 10 subsets of the signature genes were obtained for each subset size, and SA scores and p values were computed. E, F. A total of 100 simulated datasets were obtained by generating random values from a standard normal distribution and z-normalizing each row ("gene") across all columns ("samples"). For each simulated dataset, 10 subsets of the signature genes were obtained for each subset size and SA scores and p values were computed.
Format: PDF Size: 34KB Download file
This file can be viewed with: Adobe Acrobat Reader
Additional file 8:
Variation of openSESAME p values with signature composition in GEO series GSE2225 and GSE21653. A, B. For each subset size, 1000 subsets of the up-and down-regulated genes in the signature were obtained by permutation, maintaining a constant signature size, and SA scores were computed using each GEO series. Fisher's exact test or a two-sided Kolmogorov-Smirnov (K-S) test were used to compute p values for each permutation. C, D. The expression values of each gene were shuffled independently 100 times, and for each shuffled dataset, 10 subsets of the up- and down-regulated genes were obtained for each subset size and SA scores and p values were computed. E, F. A total of 100 simulated datasets were obtained by generating random values from a standard normal distribution and z-normalizing each row ("gene") across all columns ("samples"). For each simulated dataset, 10 subsets of the up- and down-regulated genes were obtained for each subset size and SA scores and p values were computed.
Format: PDF Size: 49KB Download file
This file can be viewed with: Adobe Acrobat Reader
Additional file 9:
Comparison of MARQ and openSESAME query results using the E2 treatment signature. In queries using the E2 treatment signature, these GEO series were assigned an FDR q < 0.05 by openSESAME with either the K-S or Fisher's exact tests or contained an experimental comparison to which MARQ assigned an FDR-corrected p < 0.05. Series are sorted in ascending order by Fisher q value.
Format: XLS Size: 36KB Download file
This file can be viewed with: Microsoft Excel Viewer
Additional file 10:
Comparison of MARQ and openSESAME query results using the p63 silencing signature. In queries using the p63 silencing signature, these GEO series were assigned an FDR q < 0.05 by openSESAME with either the K-S or Fisher's exact tests or contained an experimental comparison to which MARQ assigned an FDR-corrected p < 0.05. Series are sorted in ascending order by Fisher q value.
Format: XLS Size: 48KB Download file
This file can be viewed with: Microsoft Excel Viewer
