Analyzing the results
Evaluation metrics
The base-maf (base minor allele frequency) and maf (target minor allele frequency) filters were toggled with. However, there was no difference in the evaluation metrics obtained.
R2 was obtained from the summary file generated by PRSice as described previously.
Using a merged dataframe containing phenotype, age, sex, and PRS, AUC is computed as follows in R:
# Compute AUC for only PRS
auc_prs <- roc(df$PHENO, df$PRS)$auc
# Create GLM for PRS, age, and sex
prs_model <- glm(PHENO ~ PRS + Age + Sex, data = df, family = binomial)
# Get predicted probabilities from the model
predicted_probs <- predict(prs_model, type = "response")
# Compute combined AUC
auc_combined <- roc(df$PHENO, predicted_probs)$auc
# Print combined AUC
print(auc_combined)
Evaluation metrics | PGS000074 | PGS000785 |
---|---|---|
R2 | 0.0505 | 0.0576 |
AUC (only PRS) | 0.5783 | 0.6058 |
AUC (PRS + age + sex) | 0.6976 | 0.7095 |
Distribution plots
PGS000074 | PGS000785 |
---|---|
Odds ratio plots
PGS000074 | PGS000785 |
---|---|