用r语言写程序：Calculate polygenic scores with both the marginnal and joint models on testing data. And compare their performance under low and high h^2

以下是一个用R语言写的计算多基因得分（polygenic scores）并分别采用边缘和联合模型在测试数据上评估其性能的示例程序。同时还会比较在低和高h^2情况下的表现，供您参考：

# Load required packages
library(rrBLUP)
library(plyr)

# Load training and testing data
train_data <- read.csv("train_data.csv")
test_data <- read.csv("test_data.csv")

# Define the phenotype and genotypes
pheno <- train_data$pheno
geno <- train_data[,4:ncol(train_data)]

# Define the relationship matrix
A <- A.mat(geno, n.core=4)

# Fit the marginal and joint models
marginal_model <- kin.blup(pheno, A, method="average")
joint_model <- mixed.solve(pheno ~ A, random=1)

# Calculate the polygenic scores
marginal_scores <- as.numeric(marginal_model$pred[test_data[,1]])
joint_scores <- as.numeric(joint_model$u[test_data[,1]])

# Calculate the correlation between true and predicted phenotypes
marginal_corr <- cor(test_data$pheno, marginal_scores)
joint_corr <- cor(test_data$pheno, joint_scores)

# Compare the performance under low and high h^2
h2_range <- seq(0.1, 0.9, by=0.1)
results <- data.frame(h2=NULL, marginal=NULL, joint=NULL)
for (h2 in h2_range) {
  # Simulate a new phenotype with the specified h2
  sim_pheno <- sim.giv(h2, A=A, K=A, n=1)
  
  # Calculate the polygenic scores
  marginal_scores <- as.numeric(marginal_model$pred[sim_pheno$id])
  joint_scores <- as.numeric(joint_model$u[sim_pheno$id])
  
  # Calculate the correlation between true and predicted phenotypes
  marginal_corr <- cor(sim_pheno$pheno, marginal_scores)
  joint_corr <- cor(sim_pheno$pheno, joint_scores)
  
  # Save the results
  results <- rbind(results, data.frame(h2=h2, marginal=marginal_corr, joint=joint_corr))
}

# Plot the results
library(ggplot2)
ggplot(results, aes(x=h2)) +
  geom_line(aes(y=marginal, color="Marginal")) +
  geom_line(aes(y=joint, color="Joint")) +
  scale_color_manual(values=c("blue", "red")) +
  labs(x="h2", y="Correlation") +
  theme_classic()

请注意：这只是一个示例程序，您需要根据您的具体数据和研究问题进行修改和调整。