Bayesian DNA copy number analysis

BMC Bioinformatics

Table 3 Comparison among the boundary estimators on Simulated Chromosomes

method	SSQ	MAD	accuracy	accuracy inside aberrations	accuracy outside aberrations
${\hat{T}}_{01}$ , ${\hat{ρ}}^{2}$	14.23	0.00877	0.889	0.961	0.883
${\hat{T}}_{j o i n t}$ , ${\hat{ρ}}^{2}$	2.22	0.00840	0.904	0.992	0.892
${\hat{T}}_{B i n E r r A k}$ , ${\hat{ρ}}^{2}$	1.70	0.00733	0.936	0.992	0.932
${\hat{T}}_{01}$ , ${\hat{ρ}}_{1}^{2}$	9.74	0.00952	0.881	0.960	0.877
${\hat{T}}_{j o i n t}$ , ${\hat{ρ}}_{1}^{2}$	2.67	0.00970	0.882	0.993	0.867
${\hat{T}}_{B i n E r r A k}$ , ${\hat{ρ}}_{1}^{2}$	1.85	0.00781	0.929	0.993	0.920

This table shows the comparison among the error measures for profile estimation, obtained on dataset Simulated Chromosomes, for all estimators of t⁰ (apart ${\hat{T}}_{B i n E r r}$ ). We always used ${\hat{K}}_{2}$ , $\hat{ν}$ , ${\hat{σ}}^{2}$ and both ρ² estimators as estimators of the other parameters involved. The estimator ${\hat{T}}_{B i n E r r A k}$ always had the lowest SSQ and MAD errors and the highest accuracy both inside and outside the regions of aberration. Using ${\hat{ρ}}^{2}$ the performance was slightly better, because σ² <ρ².

ISSN: 1471-2105