Figure 3 | Scientific Reports

Figure 3

From: Machine learning approaches for the prediction of bone mineral density by using genomic and phenotypic data of 5130 older men

Figure 3

Mean squared error loss of various models with the number of training iterations for BMD prediction in the test dataset (\(\mathrm{n}=1026)\). The upper panel shows the performance of each model with phenotype covariates and GRS as predictors in predicting BMD at the femoral neck (A), total hip (B), and total spine (C) in the testing dataset at different BMD sites. The lower panel shows the performance of each model with phenotype covariates and 1103 individual SNPs in predicting BMD at the femoral neck (D), total hip (E), and total spine (F). Lasso regularization with the penalized value of 0.01 was used in the linear regression model for 1103 individual SNPs and phenotype covariates.

Back to article page