Table 3.
Exon-level accuracy comparison. The table shows the Pearson correlation between the true log2-FPKM expression values of exons and log2-FPKM values produced by each workflow. The Rsubread workflows give the best correlation with the true values
| Workflow | Correlation |
|---|---|
| subjunc-F + featureCounts | 0.982 |
| subjunc-G + featureCounts | 0.982 |
| STAR + featureCounts | 0.950 |
| STAR + dexseq_count | 0.927 |
| STAR + countOverlaps | 0.950 |
| TopHat2 + dexseq_count | 0.942 |
| TopHat2 + countOverlaps | 0.980 |
An offset of 1 was added to raw exon counts to avoid taking logarithms of zero values.