Table 3.
Error, time and memory comparisons on real and simulated data
| Error rate | Time (s) | Memory (GB) | |
|---|---|---|---|
| Sim 10K | |||
| MendelImpute | 3.00E−04 | 10 | 1.6 |
| Impute 5 | 2.82E−05 | 43 | 8.9 [6.6] |
| Beagle 5.1 | 2.81E−05 | 189 | 8.8 |
| Minimac 4 | 2.38E−05 | 271 [177] | 1.0 [6.6] |
| Sim 100K | |||
| MendelImpute | 2.19E−05 | 14 | 1.6 |
| Impute 5 | 9.36E−06 | 43 [177] | 9.9 [14.5] |
| Beagle 5.1 | 8.22E−06 | 279 | 20.1 |
| Minimac 4 | 7.91E−06 | 3032 [253] | 2.6 [14.5] |
| Sim 1M | |||
| MendelImpute | 2.21E−05 | 27 | 4.4 |
| Impute 5 | 1.33E−05 | 153 [752] | 12.6 [25.6] |
| Beagle 5.1 | 7.01E−06 | 769 | 25.6 |
| Minimac 4 | NA | NA | NA |
| 1000G chr10 | |||
| MendelImpute | 7.52E−03 | 34 | 4.1 |
| Impute 5 | 4.79E−04 | 35 [151] | 5.6 [7.3] |
| Beagle 5.1 | 4.82E−04 | 178 | 26.2 |
| Minimac 4 | 4.89E−04 | 298 [151] | 3.8 [7.3] |
| 1000G chr20 | |||
| MendelImpute | 7.49E−03 | 14 | 2.2 |
| Impute 5 | 4.48E−03 | 21 [74] | 5.5 [5.3] |
| Beagle 5.1 | 4.53E−03 | 88 | 6.4 |
| Minimac 4 | 4.48E−03 | 243 [74] | 2.8 [5.3] |
| HRC chr10 | |||
| MendelImpute | 1.70E−03 | 183 | 8.4 |
| Impute 5 | 8.00E−04 | 754 [2619] | 47.6 [13.0] |
| Beagle 5.1 | 8.34E−04 | 3191 | 33.2 |
| Minimac 4 | 8.57E−04 | 17 100 [2619] | 11.5 [13.0] |
| HRC chr20 | |||
| MendelImpute | 1.89E−03 | 84 | 5.1 |
| Impute 5 | 8.57E−04 | 645 [1319] | 39.0 [13.3] |
| Beagle 5.1 | 9.10E−04 | 1549 | 22.1 |
| Minimac 4 | 9.18E−04 | 8640 [1319] | 15.1 [13.3] |
Note: The best number in each cell is bold faced. The displayed error rate is the proportion of incorrectly imputed genotypes. Minimac 4 and Impute 5’s runs use the results of a prephasing step done by Beagle 5.1, whose time and memory usage are reported in brackets. For the sim 1M data, the m3vcf reference panel required for Minimac could not be computed due to excessive memory requirements.