Table 1.
MILANC rate and average local ancestry across methods on the GALA trios data
| Method | Average MILANC | Reported error rate | Average local ancestry |
||
|---|---|---|---|---|---|
| EUR | NAM | AFR | |||
| GALA Mexican | |||||
| WINPOP | 2.9 (1.3) | 12.8 | 45.4 (1.8) | 49.9 (1.8) | 4.7 (0.9) |
| LAMP-LD | 3.2 (1.5) | 9.9 | 48.5 (2.2) | 46.5 (2.2) | 5.1 (0.9) |
| ALLOY | 3.0 (1.3) | 12.5 | 62.3 (3.3) | 32.2 (3.3) | 5.5 (1.2) |
| PCAdmix | 17.1 (3.4) | 4.2a | 52.1 (6.0) | 42.3 (5.7) | 5.5 (1.5) |
| GALA Puerto Rican | |||||
| WINPOP | 3.2 (1.3) | 9.0 | 66.6 (2.5) | 13.3 (1.9) | 20.1 (1.9) |
| LAMP-LD | 2.5 (1.1) | 6.4 | 67.0 (2.6) | 11.4 (2.0) | 21.5 (2.0) |
| ALLOY | 4.3 (1.4) | 12.5 | 67.9 (2.9) | 11.2 (2.3) | 21.0 (2.2) |
| PCAdmix | 11.4 (2.4) | 4.2a | 69.2 (3.9) | 9.3 (3.0) | 21.5 (2.7) |
Note: Results were averaged in non-overlapping contiguous loci 1 Mb in size. Standard deviation of the MILANC rate and the average local ancestry across loci is denoted in parenthesis. Reported Error Rates are obtained as follows: Table 1 of Baran et al. (2012) for WINPOP/LAMP; Figure 3 of Bercovici et al. (2012) for ALLOY; Maya-Yoruba-French simulations from Supplementary Table S4 of Brisbin et al. (2012) for PCAdmix (aaccuracy for PCAdmix is only reported at calls confidently called at threshold 0.8, whereas all other methods report accuracy at all loci).