DL_drawing.Rmd

---
title: "Deep learning plot drawing"
author: "Ye Bi"
date: "`r Sys.Date()`"
output: html_document
---

```{r}
library(tidyverse)
library(ggplot2)
library(plyr)
library(ggpubr)
library(ggrepel)
library(tidyverse)
library(stargazer)
library(patchwork)
library(ggplot2)
```

## Drawing plots for h2, varE, varG
```{r}
source("../Functions/functions.R")
met_names = met_name_func(mode = 'name')
```



##Loading pred corr 
```{r}
path.bglr = "../../ARC_outputs/Met_same_genotypes_per_treatment/GBLUP/outputs_GBLUP/"
PreCorr = load_pred_corr_func(path=path.bglr, kernel = "BGLR", namm=NULL)
# PreCorr = PreCorr %>% filter(Met != "ribitol")%>% droplevels()


path.GK = "../../ARC_outputs/Met_same_genotypes_per_treatment/GK/outputs_GK/"
PreCorr_GK = load_pred_corr_func(path=path.GK, kernel = "GK", namm=NULL)
# PreCorr_GK = PreCorr_GK %>% filter(Met != "ribitol") %>% droplevels()
```


#### Barplot

##Load DL models
```{r}
DLs_model_names = c("VGG16", "ResNet50", "EfficientNetB7","InceptionV3", "MobileNetV2", "DenseNet201")
DLs_RR_mu_L = list()
PreCorr_DL_L = list()
for (i in 1:6){
  model_name = DLs_model_names[i]
  DLs_RR_mu_L[[i]] = DLs_RR_mu_generator(model = model_name)
  PreCorr_DL_L[[i]] = DLs_RR_PreCorr_generator(model = model_name)
}
names(DLs_RR_mu_L) = DLs_model_names
names(PreCorr_DL_L) = DLs_model_names
```


```{r}
# mm <- plyr::ddply(G_GK_dff, c("Treatment"), summarise, corr.mean=mean(Corr_diff, na.rm=T), corr.me = median(Corr_diff, na.rm=T))
BGLR_G_mu$corr.mean[BGLR_G_mu$corr.mean < 0.1]=0

DLs_model_names_selected = c("VGG16", "ResNet50", "EfficientNetB7","InceptionV3", "MobileNetV2", "DenseNet201")
dff_DLs_L = list()
for (i in 1:length(DLs_model_names_selected)){
  dff_DLs_L[[i]] = pred_dff_generator(df1 = BGLR_G_mu,
                                     df2 = DLs_RR_mu_L[[DLs_model_names_selected[i]]],
                                     Method=DLs_model_names_selected[i])
}

dff_bglr_gk=pred_dff_generator(df1=BGLR_G_mu,
                   df2=BGLR_GK_mu,
                   Method="RKHS")
dff_gblup = pred_dff_generator(df1=BGLR_G_mu,
                               df2=BGLR_G_mu,
                               Method="GBLUP")  
my_colors <- RColorBrewer::brewer.pal(11, 'Spectral')[c(1:4, 9:11)]
ccc = rbind(do.call(rbind, dff_DLs_L),
            dff_bglr_gk)
            
ccc$Method = factor(ccc$Method, levels=c(DLs_model_names, "RKHS", "GBLUP"))

ttt = ccc[!is.infinite(ccc$value), ]
ttt = ttt %>% filter(value >= 0) %>% droplevels()
limm = c(floor(min(ttt$value)), ceiling(max(ttt$value)))
dp_diff <- ggplot(ttt, aes(y=value, x=Met, fill=Method)) +
            # geom_line()+
            geom_bar(stat="identity",position="dodge", width = 1)+
            facet_grid(rows = vars(Treatment))+
            labs(x="Metabolite accumulation", y = "Percentage difference (%)") +
            theme_bw()+
            geom_hline(aes(yintercept = 0), color = 'darkgrey', linetype='solid', linewidth = 0.3)+
            # scale_y_continuous(limits=c(0, 90), breaks=seq(0,90,10)) +
            # scale_y_continuous(limits=c(0,400)) +
            scale_fill_manual(values = alpha(my_colors, 1))+
            # scale_color_manual(values = my_colors)+
            theme(axis.text.x=element_text(size=8, angle=90, hjust=0.95, vjust=0.2),
                  axis.text.y=element_text(size=8))

dp_diff

dev.print(pdf, file="../../temp/DL_dff_barplot_0.1_positive.pdf", height = 6, width = 10)
```