Employing genome-wide association studies and machine learning to accurately identify Eastern and Western migratory pathways of Spodoptera frugiperda in China via key molecular markers

The East Asian Insect Flyway serves as a critical migration corridor for major agricultural pests including the highly destructive fall armyworm (FAW) ( Spodoptera frugiperda ), which poses severe threats to crop security worldwide. Two primary migratory routes—eastern and western—extend from the Indochina Peninsula to northern China, enabling its widespread dispersal and enhancing its potential for damage. However, whether the population structure differs between populations from the eastern and western pathways and whether molecular markers can be used to distinguish the insect sources of these two migratory pathways remain largely unknown. Herein, FAW samples from the Indochina Peninsula to northern China collected over 2 years (2019 and 2023) were used to screen 176 key genomic loci (124 single nucleotide polymorphisms (SNPs) and 52 insertions and deletions (InDels)) through genome resequencing and genome-wide association studies (GWAS) analysis. Principal component analysis and phylogenetic trees based on these genomic loci clearly distinguished the eastern and western lineage, grouping all samples into two corresponding clusters. LASSO-regularized logistic regression identified eight stable features (6 SNPs and 2 InDels). Accordingly, machine learning models were constructed to accurately distinguish FAW ofrom the eastern and western migratory pathways. The Multi-Layer Perceptron model achieved the highest performance, with precision (90.00 %), F1-score (0.8730), accuracy (87.50 %), ROC-AUC (89.60 %) and PR-AUC (87.00 %). Using those eight stable features, six UDP-glycosyltransferase (UGT) genes were identified which were enriched in several detoxification-related pathways. Our findings establish a robust molecular toolkit for accurate source identification of FAW individuals, which can significantly enhance the monitoring and early warning systems against this migratory pest. These results also provide a new method for elucidating the migration pathways of other migratory insects worldwide. • Spodoptera frugiperda from the Indochina Peninsula to northern China were used to screen out 124 SNPs and 52 InDels. • These 176 genomic loci revealed a marked ability to distinguish S. frugiperda from the eastern and western pathways. • Machine learning based models distinguish S. frugiperda from the eastern and western migratory pathways with high accuracy. • Six UGT genes were identified from eight stable features which were enriched in several detoxification-related pathways