Skip to main content

Table 1 Computational time (in second) for the four models for predicting gene expression measurements

From: Prediction of gene expression with cis-SNPs using mixed models and regularization methods

#SNP PVE Lasso ENET LMM BSLMM
510 0.118 4.117 (0.203) 2.937 (0.073) 0.159 (0.148) 1.780 (1.789)
1375 0.002 6.594 (0.273) 5.345 (0.110) 0.560 (0.021) 3.895 (0.917)
2011 0.000 5.805 (0.172) 5.134 (0.100) 0.727 (0.076) 1.502 (0.841)
3045 0.357 8.623 (0.177) 7.992 (0.234) 1.097 (0.011) 8.286 (8.159)
4120 0.046 8.649 (0.282) 8.385 (0.227) 1.412 (0.073) 16.129 (8.792)
4953 0.523 10.019 (0.248) 9.772 (0.285) 1.621 (0.182) 7.626 (3.406)
5818 0.124 13.492 (0.199) 13.077 (0.237) 1.957 (0.057) 2.269 (0.854)
  1. #SNP denotes the number of cis-SNPs included in this gene; PVE is the proportion of variance of gene expression explained by cis-SNPs; the tuning parameters of LASSO ENET are selected using 100-fold cross validation; BSLMM uses 10,000 Monte Carlo samplings after 2,000 burn-in samplings. The times are averaged across 20 replicates, and values in parentheses are the standard deviations