Trait-based machine learning modeling of soluble carbohydrate content in Trachyspermum ammi exposed to organic and chemical fertilization and salicylic acid.

Summary

This study developed trait-based machine learning models to enable rapid, non-destructive prediction of soluble carbohydrate content in Trachyspermum ammi, a medicinal plant with documented antioxidant and antimicrobial properties. Plants were grown under organic and chemical fertilization with salicylic acid treatments to induce trait variability, and Multilayer Perceptron and Gaussian Process Regression models were trained on 45 samples using morphological, biochemical, physiological, and macronutrient data. The findings suggest that biochemical and physiological traits alone provide cost-effective and reliable predictive accuracy, though the authors acknowledge the small dataset size and recommend validation with larger, independent cohorts.

Regional applicability

The study geography is not specified in the abstract; transferability to United Kingdom medicinal herb cultivation would depend on climate adaptation and whether Trachyspermum ammi is grown commercially in UK conditions. The machine learning methodology is geographically agnostic but would require local validation datasets if applied to UK-grown ajowan or other medicinal species.

Key measures

Soluble carbohydrate content, morphological traits, physiological traits, macronutrient concentrations; model performance assessed via five-fold cross-validation using MLP and GPR algorithms

Outcomes reported

The study developed machine learning models (MLP and GPR) to predict soluble carbohydrate content in Trachyspermum ammi non-destructively using morpho-physiological traits as input variables. Models achieved high predictive accuracy, with biochemical and physiological traits alone providing nearly equivalent performance to full feature sets.

Topic tags