Table 1.
SV size | Single cutoff | Multi-cutoff | Simplified multi-cutoff | Multi-cutoff* | Simplified multi-cutoff* |
1000 | 3 (4) | 3 (4) | 3 (4) | 3 (4) | 3 (4) |
2000 | 12 (13) | 23 (26) | 21 (23) | 11 (13) | 6 (6) |
3000 | 52 (57) | 61 (68) | 61 (68) | 49 (52) | 44 (46) |
4000 | 84 (85) | 85 (86) | 85 (86) | 80 (82) | 80 (82) |
5000 | 91 (93) | 91 (93) | 91 (93) | 91 (93) | 91 (93) |
6000 | 92 (92) | 92 (92) | 92 (92) | 92 (92) | 92 (92) |
10000 | 88 (91) | 88 (91) | 88 (91) | 88 (91) | 88 (91) |
Total | 422 (435) | 443 (460) | 441 (457) | 414 (427) | 404 (414) |
False positives (chromosome 2) | 31 (31) | 31 (31) | 26 (31) | 5 (4) | 2 (1) |
Each row displays statistics for reconstructed heterozygous deletions of a particular size, derived from simulations of 454/Roche based PEM data. Columns show numbers of identified SVs for each reconstruction strategy. Numbers in parentheses correspond to simulated SV reconstructions without sequencing error. All SVs were reconstructed at an effective span coverage λ = 5× (where, λ was assessed after optimal paired-end placement) of a simulated diploid chromosome 2. Note that for the single cutoff strategy events ≥ 5 kb were reconstructed at a level near the theoretical maximum of 95% derived from the Poisson approximation (see supplementary methods and notes in Additional data file 1). However, the multi-cutoff and simplified multi-cutoff strategies outperformed the single cutoff strategy in detecting SVs < 4 kb. *We also applied alternative - that is, optimal - cutoff parameters, for which the sensitivity is similar to the single cutoff, but for which a false positive rate of approximately 5% was observed.