Natural Preservation of Refrigerated Chicken Meat Using Asphodelus microcarpus Root Extract With Machine‐Learning‐Based Shelf‐Life Prediction

ABSTRACT The demand for natural and multifunctional food preservatives has increased due to growing concerns about the safety and sustainability of synthetic additives. In this study, the ethyl acetate extract of Asphodelus microcarpus roots ( Am EA) was evaluated as a preservative, integrating antioxidant and antibacterial evaluation with an artificial intelligence (AI)‐assisted shelf‐life prediction approach. The extraction yield of Am EA was 1.15% ± 0.06% (obtained from 100 g of dried root powder using a 1:5 w/v ratio of ethyl acetate over a 24 h triple maceration period), with a high total polyphenol content (444.6 ± 0.4 mg GAE/g extract) and a comparatively low total flavonoid content, indicating enrichment in non‐flavonoid phenolic compounds. HPLC‐DAD and GC–MS analyses revealed a complex phytochemical profile dominated by phenolic acids (vanillic and ferulic acids), flavonoids (apigenin‐7‐glucoside and naringenin), and anthraquinone derivatives (aloe‐emodin and chrysophanol). The extract exhibited a radical‐scavenging capacity in the DPPH assay (IC 50 = 5.24 µg/mL), indicating effective hydrogen‐donating ability under in vitro conditions. Am EA demonstrated broad‐spectrum antibacterial activity against both Gram‐positive and Gram‐negative bacteria, with inhibition zones up to 26 ± 0.44 mm and minimum inhibitory concentrations (MICs) as low as 0.3 mg/mL. Bactericidal activity (MBC/MIC ≤ 4) was observed against Salmonella enterica , Listeria monocytogenes , and Bacillus cereus . When applied to refrigerated minced chicken meat at concentrations of 1.3% and 2.6%, Am EA significantly delayed microbial growth, lipid and protein oxidation, and sensory deterioration over 14 days of storage at 4°C. The 2.6% treatment maintained aerobic plate counts below the spoilage threshold, reduced TBARS to 1.14 ± 0.05 mg MDA/kg, limited metmyoglobin formation to 33.65%, and preserved sulfhydryl content at 54.6 µmol/g protein. Sensory acceptability remained above the rejection limit throughout storage, extending shelf life by 4–7 days compared to untreated samples. A machine‐learning‐based shelf‐life prediction framework was further developed and validated on an independent test set, comparing a random forest (RF) algorithm against Support Vector Regression and Multiple Linear Regression baselines. The RF model achieved the best predictive performance, outperforming both alternative algorithms and confirming its ability to capture the nonlinear degradation dynamics of the meat system. Shapley Additive Explanations analysis identified lipid oxidation (thiobarbituric acid reactive substances) and aerobic plate counts as the primary predictors of shelf‐life reduction, followed by psychrotrophic bacterial counts and protein oxidation markers, a hierarchy corroborated by predicted‐versus‐observed, residual, and confusion‐matrix diagnostics. Together, these chemical, microbiological, sensory, and AI‐validated findings support the potential of A. microcarpus root ethyl acetate extract as a candidate natural preservative and demonstrate the value of interpretable machine learning as a complementary, data‐driven tool for shelf‐life assessment, pending comprehensive toxicological and regulatory evaluation under controlled experimental conditions.