Table 4. Time used by a bit-block based algorithm and the conventional algorithm for computing Pearson correlation of genotypes.
Number of variants | Bit-block | Conventional | Ratio |
---|---|---|---|
4000 | 0:0:2.59 | 50:01.36 | 1:1159 |
7000 | 0:0:7.59 | 2:31:56.48 | 1:1201 |
10 000 | 0:0:14.99 | 5:08:54.14 | 1:1236 |
13 000 | 0:0:23.60 | 8:43:17.78 | 1:1330 |
16 000 | 0:0:35.18 | 13:18:33.96 | 1:1362 |
Note: The variants were randomly drawn from chromosome 22 and genotypes of 2504 subjects were used to calculate the correlation. The time format a:b:c.d stands for a hour(s), b minute(s) and c.d second(s), respectively. The version of Java running environment was 1.7. CPU frequency was 3.457 GHz. a: For the conventional approach, the genotypes were encoded according to the number of alternative alleles, i.e. 0, 1 and 2 first. The Pearson correlation coefficients were calculated for each pair of variants by a fast Java package Colt (https://dst.lbl.gov/ACSSoftware/colt/).