Dataset | Accuracy | %False positive |
---|
| Random | Unique | BowStrap | Random | Unique | BowStrap |
---|
50 M
| 0.97 | 0.91 | 0.98 | 2.70 | 0.0 | 0.43 |
25 M
| 0.97 | 0.91 | 0.97 | 2.67 | 0.0 | 0.21 |
10 M
| 0.97 | 0.90 | 0.92 | 2.72 | 0.0 | 0.08 |
1 M
| 0.97 | 0.88 | 0.74 | 3.44 | 0.0 | 0.00 |
0.5 M
| 0.97 | 0.87 | 0.66 | 3.40 | 0.0 | 0.00 |
- Three ‘Bowtie’-based methods for gene model expression were considered, using gene model expression values from synthetic RNAseq data of 50, 25, 10, 1, and 0.5 million total 46-mer sequence reads. “Random” uses ‘Bowtie’s default random assignment of multiply aligning reads. “Unique” uses only reads with a single, unambiguous alignment location. “Accuracy” is the proportion of gene models correctly identified as either expressed or absent. For “Random” and “Unique”, detection of expression is defined as an RPKM value greater than 0. For “BowStrap”, detection of expression is defined as Benjamini-Hochburg corrected p-value < 0.05. “% False Positive” is the percent of gene models identified as expressed that are absent in synthetic RNAseq data.