Skip to main content

Table 2 Prediction accuracy of phenotypes using selected SNPs compared to all SNPs on Ψ data

From: Exploring the potential of incremental feature selection to improve genomic prediction accuracy

Species

Phenotype

#Selected SNPs (%)

R2 (selected SNPs)

R2 (all SNPs)

Maize

FT

446 (0.18%)

0.486

0.364

 

HT

4021 (1.64%)

0.237

0.203

 

YLD

1601 (0.65%)

0.346

0.278

Sorghum

HT

56,299 (100%)

0.287

 

MO

27,001 (47.96%)

0.391

0.389

 

YLD

8501 (15.10%)

0.003

0.030

Soy

HT

451 (10.65%)

0.213

0.201

 

R8

1001 (23.64%)

0.206

0.196

 

YLD

731 (17.27%)

0.373

0.370

Spruce

DBH

6930 (100%)

0.096

 

DE

2401 (34.65%)

0.141

0.142

 

HT

6930 (100%)

0.143

Switchgrass

AN

75,001 (34.54%)

0.767

0.754

 

HT

130,001 (59.87%)

0.519

0.513

 

ST

217,150 (100%)

0.512

Pig

T1

1 (0.003%)

− 0.021

0.013

 

T2

9501 (28.06%)

0.190

0.192

 

T3

3201 (9.45%)

0.080

0.079

 

T4

8501 (24.66%)

0.112

0.111

 

T5

7001 (20.31%)

0.172

0.163

Chicken

EW28

29,001 (9.84%)

0.046

0.046

 

EW36

41,001 (13.91%)

0.088

0.070

 

EW56

50,001 (16.97%)

0.091

0.079

 

EW66

43,001 (14.59%)

0.116

0.102

 

EW72

45,001 (15.27%)

0.015

− 0.012

 

EW80

60,001 (20.36%)

0.060

0.055

 

EWAFE

40,001 (13.57%)

− 0.010

− 0.014

  1. Number of selected SNP markers based on Φ data and prediction accuracy (measured as R2) of phenotypes on Ψ data using random forest trained on the selected SNP markers as well as on all SNPs. If the IFS approach selects all SNPs, the corresponding R2 value is given as –, since it would be the same value as in the last column