Skip to main content

Table 4 Prediction performance in soybean for five traits of eight methods in terms of Pearson’s correlation (taken from Liu et al., 2019); A). Methods dualCNN, deepGS and singleCNN are different versions of CNN. Prediction performance in terms of Average Spearman Correlation (ASC) and mean square error (MSE) with genotype × environment interaction (I) and without genotype × environment interaction (WI) in a wheat dataset for trait Fusarium head blight (FHB) severity data (Montesinos-López et al., 2020; B)

From: A review of deep learning applications for genomic selection

A Method Yield Protein Oil Moisture Height
  dualCNN 0.452 0.619 0.668 0.463 0.615
  DeepGS 0.391 0.506 0.531 0.31 0.452
  Dense 0.449 0.603 0.657 0.427 0.612
  singleCNN 0.463 0.573 0.627 0.449 0.565
  rrBLUP 0.412 0.392 0.39 0.413 0.458
  BRR 0.422 0.392 0.39 0.413 0.458
  Bayes A 0.419 0.393 0.388 0.415 0.458
  BL 0.419 0.394 0.388 0.416 0.458
B Interaction Type ASC SE MSE SE
  I BRR 0.584 0.012 3.015 0.169
  I NDNN 0.626 0.013 1.891 0.088
  I GP 0.596 0.01 2.457 0.121
  I PDNN 0.627 0.012 1.912 0.073
  WI BRR 0.436 0.018 4.481 0.25
  WI NDNN 0.635 0.013 1.872 0.084
  WI GP 0.431 0.018 3.418 0.186
  WI PDNN 0.584 0.014 2.853 0.412