부산대학교 도서관

This research used the simulation program to identify individuals resistant to foot-and-mouth disease (FMD) using the QMSim program. The simulation program was utilized to generate genetic and phenotypic data for individuals with and without FMD immunity. Subsequently, based on the simulated data, a genome-wide association study (GWAS) was performed to detect quantitative trait loci (QTL) associated with FMD immunity. Additionally, the QTLs identified by GWAS were compared with Random Forest (RF) and XGBoost. Out of the 41,461 SNPs, which included QTLs generated from the simulation, a total of 20 markers were found to be associated with FMD immunity. When comparing the performance of GWAS, RF, and XGBoost, RF identified the highest number of QTLs (7), followed by GWAS (6) and XGBoost (3). Furthermore, GBLUP, RF, and XGBoost were employed to classify individuals as either having or lacking FMD immunity. The classification accuracy, sensitivity, and specificity were evaluated using a confusion matrix, and the results were compared. The overall accuracy of the classification was as follows: XGBoost 0.53, RF 0.52, GBLUP 0.51. Sensitivity values were RF 0.98, XGBoost 0.97, GBLUP 0.19, and specificity values were GBLUP 0.83, XGboost 0.08, RF 0.05. XGBoost consistently outperformed the other methods in the overall accuracy and sensitivity, while GBLUP exhibited the lowest performance. Therefore, the research suggests that combining various methods in an ensemble approach, rather than relying solely on GBLUP, can lead to better predictions of FMD-resistant individuals. This approach has the potential to help mitigate the damages caused by future FMD outbreaks.