Troubleshooting: MA-GBLUP

Comparing MA estimates using meta vs target genetic variance on selected traits

Published

February 25, 2026

Selected traits (MT, PH, and TKW) exhibited higher RMSE for MA-GBLUP using meta genetic variance compared to target RMSE
For MA-GBLUP: SS=summary statistics; MetaPAs= using SS of PYT-S and AYT-S; MetaPAo= using SS of PYT-S and AYT-O; MetaAAs= using SS of AYT-O and AYT-S
Plot: Top=Meta estimates uses meta genetic variance; bottom=Meta estimates uses target genetic variance

MT

Code

par(mfrow= c(2, 3))
par(mar=c(5,7,6,0))
plot(Gh_AYT_O, mt_AYT_O, xlab = expression(hat(g)[AYT_O]), 
     ylab = expression(hat(g)[MetaPAs]), cex.lab= 1.5, col="orangered3")
plot(Gh_AYT_S, mt_AYT_S, xlab = expression(hat(g)[AYT_S]), 
     ylab = expression(hat(g)[MetaPAo]), cex.lab= 1.5, col="gold3")
plot(Gh_PYT_S, mt_PYT_S, xlab = expression(hat(g)[PYT_S]), 
     ylab = expression(hat(g)[MetaAAs]), cex.lab= 1.5, col="forestgreen")

plot(Gh_AYT_O, mt_AYT_O2, xlab = expression(hat(g)[AYT_O]), 
     ylab = expression(hat(g)[MetaPAs]), cex.lab= 1.5, col="orangered3")
plot(Gh_AYT_S, mt_AYT_S2, xlab = expression(hat(g)[AYT_S]), 
     ylab = expression(hat(g)[MetaPAo]), cex.lab= 1.5, col="gold3")
plot(Gh_PYT_S, mt_PYT_S2, xlab = expression(hat(g)[PYT_S]), 
     ylab = expression(hat(g)[MetaAAs]), cex.lab= 1.5, col="forestgreen")
mtext("MT \n Within Estimated SNP effect vs Meta-SNP effect", side =3, line= -4,
      outer = TRUE)

Code

df_var <- data.frame(Within=c(GP_AYT_O, GP_AYT_S, GP_PYT_S), Meta = c(mt2_AYT_O,
                                                                      mt2_AYT_S, mt2_PYT_S),
                     n = c(1698, 850, 7800)) %>%
  `row.names<-`(c("AYT-O", "AYT-S", "PYT-S"))

kable(df_var,caption="Estimated genetic variance") %>% 
  kable_styling(latex_options = "scale_down")%>% 
  add_footnote(c("n= number of lines"))

Estimated genetic variance
	Within	Meta	n
AYT-O	40.14131	43.52643	1698
AYT-S	20.48739	44.97834	850
PYT-S	46.03035	33.59257	7800
^a n= number of lines

Code

df_rmse <- data.frame(Meta=c(7.37, 5.62, 6.27), Target = c(7.02, 4.21, 8.42))%>%
  `row.names<-`(c("AYT-O", "AYT-S", "PYT-S"))

kable(df_rmse,caption="RMSE") %>% 
  kable_styling(latex_options = "scale_down")%>% 
  add_footnote(c("Meta=using meta genetic variance","Target=using target genetic variance"))

RMSE
	Meta	Target
AYT-O	7.37	7.02
AYT-S	5.62	4.21
PYT-S	6.27	8.42
^a Meta=using meta genetic variance
^b Target=using target genetic variance

PH

Code

par(mfrow= c(2, 3))
par(mar=c(5,7,6,0))
plot(Gh_AYT_O, mt_AYT_O, xlab = expression(hat(g)[AYT_O]), 
     ylab = expression(hat(g)[MetaPAs]), cex.lab= 1.5, col="orangered3")
plot(Gh_AYT_S, mt_AYT_S, xlab = expression(hat(g)[AYT_S]), 
     ylab = expression(hat(g)[MetaPAo]), cex.lab= 1.5, col="gold3")
plot(Gh_PYT_S, mt_PYT_S, xlab = expression(hat(g)[PYT_S]), 
     ylab = expression(hat(g)[MetaAAs]), cex.lab= 1.5, col="forestgreen")

plot(Gh_AYT_O, mt_AYT_O2, xlab = expression(hat(g)[AYT_O]), 
     ylab = expression(hat(g)[MetaPAs]), cex.lab= 1.5, col="orangered3")
plot(Gh_AYT_S, mt_AYT_S2, xlab = expression(hat(g)[AYT_S]), 
     ylab = expression(hat(g)[MetaPAo]), cex.lab= 1.5, col="gold3")
plot(Gh_PYT_S, mt_PYT_S2, xlab = expression(hat(g)[PYT_S]), 
     ylab = expression(hat(g)[MetaAAs]), cex.lab= 1.5, col="forestgreen")
mtext("PH \n Within Estimated SNP effect vs Meta-SNP effect", side =3, line= -4,
      outer = TRUE)

Code

df_var <- data.frame(Within=c(GP_AYT_O, GP_AYT_S, GP_PYT_S), Meta = c(mt2_AYT_O,
                                                                      mt2_AYT_S, mt2_PYT_S),
                     n = c(1698, 850, 7800)) %>%
  `row.names<-`(c("AYT-O", "AYT-S", "PYT-S"))

kable(df_var,caption="Estimated genetic variance") %>% 
  kable_styling(latex_options = "scale_down")%>% 
  add_footnote(c("n= number of lines"))

Estimated genetic variance
	Within	Meta	n
AYT-O	0.9740531	1.1405037	1698
AYT-S	0.7472458	1.1458741	850
PYT-S	1.1832437	0.8984803	7800
^a n= number of lines

Code

df_rmse <- data.frame(Meta=c(1.30, 1.03, 1.00), Target = c(1.18, 0.86,1.16))%>%
  `row.names<-`(c("AYT-O", "AYT-S", "PYT-S"))
kable(df_rmse,caption="RMSE") %>% 
  kable_styling(latex_options = "scale_down")

RMSE
	Meta	Target
AYT-O	1.30	1.18
AYT-S	1.03	0.86
PYT-S	1.00	1.16

TKW

Code

par(mfrow= c(2, 3))
par(mar=c(5,7,6,0))
plot(Gh_AYT_O, mt_AYT_O, xlab = expression(hat(g)[AYT_O]), 
     ylab = expression(hat(g)[MetaPAs]), cex.lab= 1.5, col="orangered3")
plot(Gh_AYT_S, mt_AYT_S, xlab = expression(hat(g)[AYT_S]), 
     ylab = expression(hat(g)[MetaPAo]), cex.lab= 1.5, col="gold3")
plot(Gh_PYT_S, mt_PYT_S, xlab = expression(hat(g)[PYT_S]), 
     ylab = expression(hat(g)[MetaAAs]), cex.lab= 1.5, col="forestgreen")

plot(Gh_AYT_O, mt_AYT_O2, xlab = expression(hat(g)[AYT_O]), 
     ylab = expression(hat(g)[MetaPAs]), cex.lab= 1.5, col="orangered3")
plot(Gh_AYT_S, mt_AYT_S2, xlab = expression(hat(g)[AYT_S]), 
     ylab = expression(hat(g)[MetaPAo]), cex.lab= 1.5, col="gold3")
plot(Gh_PYT_S, mt_PYT_S2, xlab = expression(hat(g)[PYT_S]), 
     ylab = expression(hat(g)[MetaAAs]), cex.lab= 1.5, col="forestgreen")
mtext("TKW \n Within Estimated SNP effect vs Meta-SNP effect", side =3, line= -4,
      outer = TRUE)

Code

df_var <- data.frame(Within=c(GP_AYT_O, GP_AYT_S, GP_PYT_S), Meta = c(mt2_AYT_O,
                                                                      mt2_AYT_S, mt2_PYT_S),
                     n = c(1698, 850, 7800)) %>%
  `row.names<-`(c("AYT-O", "AYT-S", "PYT-S"))

kable(df_var,caption="Estimated genetic variance") %>% 
  kable_styling(latex_options = "scale_down")%>% 
  add_footnote(c("n= number of lines"))

Estimated genetic variance
	Within	Meta	n
AYT-O	11.90081	19.95556	1698
AYT-S	10.47922	19.36336	850
PYT-S	20.98716	11.42676	7800
^a n= number of lines

Code

df_rmse <- data.frame(Meta=c(5.11, 4.81, 4.12), Target = c(3.80, 3.49,6.93))%>%
  `row.names<-`(c("AYT-O", "AYT-S", "PYT-S"))
kable(df_rmse,caption="RMSE") %>% 
  kable_styling(latex_options = "scale_down")

RMSE
	Meta	Target
AYT-O	5.11	3.80
AYT-S	4.81	3.49
PYT-S	4.12	6.93