| Prostate | Smokers | Male/female |
---|
n
| Up | Down | Up | Down | Up | Down |
20 | 95% | 100% | 95% | 100% | 100% | 100% |
100 | 92% | 97% | 97% | 99% | 91% | 98% |
250 | 82% | 94% | 97% | 92% | 86% | 95% |
500 | 74% | 86% | 89% | 84% | 79% | 86% |
1000 | 55% | 69% | 78% | 72% | 75% | 64% |
2000 | 43% | 48% | 62% | 52% | 74% | 54% |
- Two ranked lists of differentially expressed genes were calculated: one using Affymetrix' original CDF file and one using our CDF file. Genes were sorted in descending order according to their normalized difference in rank (|rank1–rank2|/minimum(rank1, rank2)). The top n genes are those which are most strongly affected by the probe set redefinition. The percentage of genes that obtained a higher rank using our refined CDF file compared to using the original CDF file is given. For random redefinitions, a percentage of 50 would be expected. Separate analyses for up- and downregulated genes were done. It can be seen that among the most strongly affected probe sets the large majority shows improved differential expression results when analyzed using the refined CDF file.