Table 2

Impact on imputation accuracy of splitting chromosomes into chunks

Accuracy


NNC (split)

NNC (non split)

SCA (split)

SCA (non split)


Complete missing

Overall

97.42%

(0.01, 0.93, 0.99, 1)

97.42%

(0.01, 0.93, 0.99, 1)

88.29%

(0.01, 0.79, 0.95, 1)

88.29%

(0.01, 0.79, 0.95, 1)


0.95 P.P.

99.23%

(0, 0.98, 1, 1)

99.23%

(0.00, 0.98, 1.00, 1)

97.30%

(0.00, 0.94, 1.00, 1)

97.30%

(0.00, 0.94, 1.00, 1)


80% missing

Overall

97.24%

(0.01, 0.93, 0.99, 1)

97.70%

(0.01, 0.93, 0.99, 1)

88.76%

(0.01, 0.80, 0.96, 1)

88.76%

(0.01, 0.80, 0.96, 1)


0.95 P.P.

99.28%

(0.00, 0.98, 1.00, 1)

99.30%

(0.00, 0.98, 1.00, 1)

97.44%

(0.00, 0.94, 1.00, 1)

97.44%

(0.00, 0.94, 1.00, 1)


No obvious impact of splitting chromosome 2 into small chunks of 10 Mb on imputation accuracy while using the data from the NNC and SCA sets. In all tests, 10% of the SNPs on chromosome 2 were randomly selected and their genotype data were either completely removed (Complete missing), or only 80% randomly removed (80% missing).

Zhao et al. BMC Genetics 2008 9:85   doi:10.1186/1471-2156-9-85

Open Data