Genome-wide prediction using Bayesian additive regression trees

Table 1 Mean squared prediction error (MSPE) for the LASSO, Bayesian LASSO (BLASSO), genomic BLUP (GBLUP), reproducing kernel Hilbert space (RKHS) regression, random forests (RF) and Bayesian additive regression trees (BART) methods evaluated on the simulated original QTLMAS2010 data

Method	Mean squared prediction error (MSPE)
LASSO
minMSE	62.020
minMSE + 1SE	63.404
BLASSO	66.209
GBLUP	66.949
RKHS
\(h = 0.05\)	66.910
\(h = 0.1\)	66.821
\(h = 0.25\)	67.200
RF	M = 10	M = 25	M = 50	M = 100	M = 200	M = 400	M = 600
	82.108	79.772	77.794	77.274	77.149	76.141	76.419
BART	M = 10	M = 25	M = 50	M = 100	M = 200	M = 400	M = 600
\(q = 0.9\)
\(\kappa\) = 2	76.231	69.974	65.703	64.967	64.324	64.213	64.574
\(\kappa\) = 3	71.325	68.537	66.755	63.772	62.782	62.919	63.476
\(\kappa\) = 4	79.264	66.554	66.376	63.596	62.595	63.119	63.790
\(\kappa\) = 5	72.344	70.608	65.467	62.705	62.715	63.997	64.982
\(q = 0.95\)
\(\kappa\) = 2	78.656	76.734	68.282	64.126	64.218	63.697	64.566
\(\kappa\) = 3	74.893	68.379	64.858	63.762	62.884	63.108	63.402
\(\kappa\) = 4	74.128	66.817	64.788	63.836	62.596	63.175	63.807
\(\kappa\) = 5	76.757	66.284	64.512	62.648	62.823	63.912	64.976

The lowest MSPE obtained with each method is highlighted in italics. M is the number of trees for RF and BART, and \(q\) and \(\kappa\) are hyperparameters of the BART priors. The stopping criteria for the regularization coefficient λ in LASSO were obtained based on tenfold cross-validation both at minimum MSE and minimum MSE plus 1 standard error [42]

ISSN: 1297-9686