Table 1.
Size and computational cost of Monorail runs. Processing wall-clock times are estimated from run logs and are approximate. Wall times are roughly “node hours”, where a typical node used here has 48 cores and 192 GB of RAM. Node types vary somewhat across clusters used. Input and output sizes are calculated from compressed files
| Collection | Input Size (TB) | Output Size (TB) | # Sequence Runs | # Studies | # Junctions (M) | # bigWigs (M) | Processing WallTime (h) |
|---|---|---|---|---|---|---|---|
| SRA Human v3 | 474 | 72 | 316,443 | 8,677 | 228 | 1.2 | 1,728 |
| SRA Mouse v1 | 362 | 62 | 416,859 | 10,088 | 148 | 1.7 | 1,608 |
| TCGA | 75 | 7 | 11,348 | 1 | 31.5 | 0.045 | 170 |
| GTEx v6 | 35 | 6.7 ∗ | 9,911 | 1 | 22 | 0.040 | 168 |
| GTEx v7 & v8 | 46 | 4.9 | 9,303 | 1 | 10.6 † | 0.037 | 123 |
| Total | 992 | 152.6 | 762,995 | 18,768 | 396 ‡ | 3.022 | 3,797 |
*Output size for GTEx v6 includes BAM files in addition to typical summaries
†Number of additional junctions beyond those in GTEx v6
‡Total after counting only the distinct junctions