Table 8.

Mean XGBoost model performance on five predicted CpG methylation levels with different sets of genetic and non-genetic features, predicted on an independent testing set.

	XGBoost
	1. GWAS of 75% of samples				2. GTEx	3. 41.5 Mb-43.6 Mb
	1.1. Before B-H correction		1.2. After B-H correction
	1.1.1, 1.1.3–1.1.5, 1.1.7. Individual CpG probes	1.1.9. Combined	1.2.1, 1.2.3–1.2.5, 1.2.7. Individual CpG probes	1.2.9. Combined
Mean RMSE train set (SD)	0.00656 (0.00308)	0.0172 (0.0124)	0.0301 (0.0182)	0.0294 (0.0176)	0.0283 (0.0163)	0.0279 (0.0193)
Mean R2 train set (SD)	0.936 (0.0604)	0.721 (0.154)	0.214 (0.0606)	0.246 (0.0576)	0.288 (0.0492)	0.348 (0.134)
Mean RMSE test set (SD)	0.0318 (0.0214)	0.0311 (0.0178)	0.0304 (0.0186)	0.0304 (0.0187)	0.0299 (0.0188)	0.0295 (0.0177)
Mean R2 test set (SD)	−0.0168 (0.145)	−0.0507 (0.129)	0.0292 (0.0882)	0.0308 (0.113)	0.0731 (0.0765)	0.0898 (0.0906)