Evaluation Results

Performance Score Table

Model	AUC
Pre-Trained PRS	0.829
AB-PRS	0.895

Performance Score Comparisons

This barplot compares the AUC performance score between different sets of PRSs derived from different models. The models are ordered by increasing performance score.

 
   plot <- ggplot(data=Scores, aes(x=reorder(Model, +Score), y=Score, fill=Model)) +
    geom_bar(stat="identity", width=0.5) +
    geom_text(aes(label=round(Score, 2)), size=3.5, vjust=-0.3) +
    scale_y_continuous(expand = expansion(mult = c(0, .1))) + 
    labs(x="Model", y=ylab, fill="Model") +
    theme(panel.grid.major = element_blank(), panel.grid.minor = element_blank(),
          panel.background = element_blank(), axis.line = element_line(colour = "black")) +
    scale_fill_manual(values = setNames(c("#ef8a62", "#67a9cf"), c(name1, name2)))

Polygenic Risk Score Distributions

The figure shows the density curves of the PRS of control (0) and case (1) phenotype for each set of PRS. The PRSs are standardized with a mean of 0 and standard deviation of 1 for each set. The goal of this figure is to see how well the polygenic risk scores can distinguish between case and control.

 
    StandardizedPRS$Method <- factor(StandardizedPRS$Method, levels = c(name1, name2))
    plot<-ggplot(StandardizedPRS,aes(x=PRS, fill=Phenotype))+
      geom_density(alpha=0.5)+
      scale_fill_manual(values = c("#67a9cf", "#ef8a62"))+
    labs(x="Standardized PRS", y="Density") +
    theme(panel.grid.major = element_blank(), panel.grid.minor = element_blank(),
            panel.background = element_blank(), axis.line.x = element_line(colour = "black")) + 
    facet_grid(cols=vars(Method),scales="free_x")

Phenotype vs PRS Percentile

The figure plots percentage of cases(prevalence) against the risk score percentile for PRSs derived from different models. For each model, 15 quantiles are plotted in the graph. A model that performs better should have a higher prevalence in the higher risk score percentiles and a lower prevalence in the lower percentiles.

 
  plot <- ggplot(PrevalenceData, aes(x = Percentile, y = Prevalence, color=Model)) +
    geom_point(size = 3) +
    labs(x = "Risk Score Percentile", y = "Case Prevalence",
         title = "Prevalence vs. Risk Score Percentile") +
    theme(panel.grid.major = element_blank(), panel.grid.minor = element_blank(), 
          panel.background = element_blank(), axis.line = element_line(colour = "black"))