An Evaluation of the MiDCoP Method for Imputing Allele Frequency in Genome Wide Association Studies - Software Engineering, Artificial Intelligence, Networking and Parallel/Distributed Computing - page 62

Table 3 Summary of linear regression fit of p-values between imputed (Y) and actual (X)
SNPs for 2/3 matched scenario

LD group Reg. Statistics Mean Sta. Dev Min Max

R 2 0.9427 0.0077 0.9236 0.9558

1

Intercept 0.0119 0.0031 0.0037 0.0178

Slope 0.9713 0.0090 0.9515 0.9903

R 2 0.8893 0.0178 0.8522 0.9269

[0.9-1.0)

Intercept 0.0237 0.0071 0.0103 0.0393

Slope 0.9437 0.0203 0.8954 0.9832

R 2 0.7904 0.0319 0.7159 0.8432

[0.8-0.9)

Intercept 0.0462 0.0113 0.0213 0.0736

Slope 0.8859 0.0332 0.7836 0.9603

R 2 0.6828 0.0408 0.6077 0.8001

[0.7-0.8)

Intercept 0.0652 0.0147 0.0365 0.0978

Slope 0.8289 0.0438 0.7103 0.9372

R 2

0.6452 0.0607 0.4916 0.7944

[0.6-0.7)

Intercept 0.0872 0.0242 0.0368 0.1465

Slope 0.8040 0.0611 0.6443 0.9406

R 2 0.4555 0.0772 0.2568 0.6204

[0.5-0.6)

Intercept 0.1210 0.0311 0.0566 0.2116

Slope 0.6718 0.0940 0.4287 0.8971

R 2

0.3084 0.0581 0.1825 0.4124

[0,0.5)

Intercept 0.1656 0.0225 0.1115 0.2160

Slope 0.5673 0.0653 0.4299 0.7024

R 2 0.7902 0.0144 0.7548 0.8216

Over All

Intercept 0.1656 0.0225 0.1115 0.2160

Slope 0.8904 0.0137 0.8587 0.9166

6

Summaries and Conclusion

GWAS studies have been an important area of studies to investigate the
associations of SNPs between diseases and genetic markers. Due to technological
limitation, a large fraction of SNPs in the genome are usually not genotyped. In
order to increase the opportunity of identifying the SNPs with potential high
association with diseases and increase the power of the association tests, several
methods for imputing the missing SNPs at the individual level data have been
developed and successfully applied in GWAS studies. However, due to
computational intensity and cost, there is a need for developing methods to impute