Machine learning-assisted carbon dot sensing for iron speciation in Auricularia auricula soaking solutions under various processing conditions.

Accurate iron speciation in Auricularia auricula soaking solution is essential for evaluating nutritional quality, as bioavailability depends on its valence state (Fe/ Fe). Herein, a machine learning-assisted fluorescent sensing strategy based on bamboo-derived carbon quantum dots (CQDs) was developed for iron speciation analysis in Auricularia auricula soaking solutions. The CQDs, synthesized via a one-pot hydrothermal method, acted as a selective probe for Fe through static fluorescence quenching, while Fe was indirectly quantified following HO oxidation to Fe. The method exhibited a wide linear range of 1-500 μM with a low limit of detection (LOD) of 0.30 μM. Application to soaking solutions showed iron predominantly existed as Fe, while the bioaccessible Fe fraction decreased to below 1% after 30 min at 100 °C. Furthermore, a random forest model revealed temperature- and origin-dependent iron leaching behavior, achieving excellent predictive performance with a coefficient of determination R = 0.99 between measured and predicted values.