Table 7.
Performance of exon–exon junction detectors in junction detection and read mapping from using SEQC RNA-seq data
| Method | Number of junctions (’000) |
Known junctions (%) |
Supporting junction reads (%) |
Time (h) | Memory (Gb) | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| A | B | C | D | A | B | C | D | A | B | C | D | |||
| Subjunc | 152 | 142 | 155 | 157 | 84.4 | 86.6 | 85.6 | 85.8 | 95.8 | 95.1 | 95.7 | 95.3 | 1.4 (1.9) | 8.4 (4.7) |
| MapSplice | 171 | 157 | 173 | 175 | 78.3 | 81.4 | 80.1 | 80.2 | 94.4 | 93.5 | 94.2 | 93.8 | 5.6 | 4.3 |
| TopHat | 156 | 145 | 159 | 161 | 82.5 | 84.9 | 83.8 | 84.0 | 93.8 | 93.5 | 93.8 | 93.6 | 9.2 | 2.9 |
| TopHat 2 | 152 | 141 | 155 | 157 | 83.8 | 85.9 | 85.0 | 85.2 | 94.1 | 93.5 | 93.9 | 93.7 | 9.9 | 3.5 |
Columns give the number of reported exon–exon junctions, percentage of reported junctions that span known exons, percentage of reported junction reads that support those known junctions, for each of the four samples included in the SEQC project (A, B, C and D). Sample A is Universal Human Reference RNA made up from 10 cancer cell lines and sample B is Human Brain Reference RNA. C and D are the mixtures of A and B. Chromosomal coordinates of annotated exons from NCBI RefSeq mouse annotation build 37.2 are used to determine whether or not a junction spans known exons. Running time and peak memory usage of Subjunc when set to use less memory are given in parenthesis.