BACKGROUND: Aspiration pneumonia represents a significant but understudied cause of morbidity and mortality, particularly among aging populations with neurological comorbidities. Current risk stratification tools rely on single-dimensional clinical data, limiting their predictive accuracy and population health utility. This study aimed to develop comprehensive risk prediction models and identify genetic determinants of aspiration pneumonia using integrated epidemiological, computational, and genomic approaches. METHODS: We conducted a nested case-control study within the UK Biobank (n = 501,968), identifying 3,659 aspiration pneumonia cases (ICD-10 J69.0) and selecting 1:4 controls initially matched on age, sex, and assessment center. Propensity scores were estimated using logistic regression based on demographic, lifestyle, functional, biomarker, and comorbidity variables. A 1:4 nearest-neighbor propensity score matching without replacement was then performed using a caliper of 0.2 standard deviations of the logit of the propensity score. We performed multivariable logistic regression and sensitivity analyses for comorbidity profiling, machine-learning risk prediction with model interpretation, and genome-wide association analysis using REGENIE. RESULTS: Neurological conditions were the strongest risk factors, including dysphagia (OR 11.71, 95% CI 10.52-13.03; P < 0.001), Parkinson's disease (OR 10.94, 95% CI 9.20-13.02; P < 0.001), and dementia (OR 9.45, 95% CI 8.35-10.69; P < 0.001). In joint comorbidity models, dysphagia, stroke, Parkinson's disease, and dementia remained independently associated with aspiration pneumonia. The LightGBM model showed high discrimination (AUC 0.867, 95% CI 0.854-0.880) with balanced sensitivity (75.3%) and specificity (78.9%). GWAS identified genome-wide significant signals at the APOE locus, with rs429358 as the lead variant (P = 2.29e-13; OR 2.22, 95% CI 2.09-2.36). CONCLUSIONS: This population-based analysis shows that aspiration pneumonia is closely associated with neurological morbidity, functional vulnerability, inflammatory markers, and an APOE-centered genetic signal. The prediction model demonstrated strong discriminative performance. Because dementia-excluded and comorbidity-stratified genotype-level analyses were not performed, the APOE association should be interpreted as a neurodegeneration-related susceptibility signal rather than definitive evidence of an effect independent of dementia. CLINICAL TRIAL NUMBER: Not applicable.