Singh M, Zhang M, Espinal-Ruiz M, Rathnayake S, Xue J, Shi J, Liu X, Hanner R, Corradini M G
Department of Food Science, University of Guelph, Guelph, ON, Canada.
Department of Integrative Biology, University of Guelph, Guelph, ON, Canada.
J AOAC Int. 2025 Mar 26. doi: 10.1093/jaoacint/qsaf029.
Maple syrup is often adulterated by dilution or substitution with other syrups due to its high demand and price. Fingerprinting techniques, e.g., DNA barcoding, detect adulteration in other foods. However, extensive processing during the transformation of sap into syrup degrades the genetic material, lowering the efficacy of this approach. In contrast, fluorescence fingerprints (EEMs) rely on a sample's intrinsic fluorophores to provide valuable information for detecting adulteration.
This study evaluates the capabilities and limitations of EEMs to scout for adulteration markers and discriminate between pure and adulterated maple syrup samples.
EEMs of pure amber and dark maple syrups and admixtures with common adulterants (beet, corn, and rice syrups at 1-50%) were obtained using a spectrophotometer (λex=250-500 nm, and λem=280-650 nm). The major components of the EEMs were identified using PARAFAC and confirmed by LC-MS/MS. The ratio of intensities of the two most prevalent EEM features was calculated. An artificial neural network (ANN) and a convolutional neural network (CNN) were developed to analyze the EEMs based on emissions at two selected excitation wavelengths and the full EEM image, respectively, to discriminate presence and level of adulteration.
EEMs of the samples allowed identifying valuable discriminatory information. The efficacy of the ratio of the emission intensities at λem=350 and 425 (I425/I350) when λex= 290 nm to identify potential fraud (70-86% correct identifications) depended on the adulterant. This ratio was particularly effective for beet syrup adulteration, even at concentrations <2%. Applying machine learning algorithms improved detection for all adulterants. ANN correctly identified adulteration type and level (90 & 82%). The CNN approach accurately classified 75-99% of adulterated syrups but required additional computational power and denser data sets.
This study aids in providing a quick, non-destructive and green monitoring tool for maple syrup adulteration based on its intrinsic fluorophores.
Maple syrup is often adulterated with other syrups due to high demand and price. DNA barcoding is ineffective in detecting maple syrup adulteration due to DNA degradation. Fluorescence fingerprints or EEMs allow scouting for discriminatory markers in maple syrup. Machine learning algorithms (ANN and CNN) applied to EEM data can aid detection.
由于枫糖浆需求高且价格昂贵，常被稀释或用其他糖浆替代而掺假。指纹识别技术，如DNA条形码技术，可检测其他食品中的掺假情况。然而，在树液转化为糖浆的过程中，大量加工会使遗传物质降解，降低这种方法的有效性。相比之下，荧光指纹（EEMs）依靠样品的固有荧光团提供检测掺假的有价值信息。
本研究评估EEMs寻找掺假标志物以及区分纯枫糖浆和掺假枫糖浆样品的能力和局限性。
使用分光光度计（λex = 250 - 500 nm，λem = 280 - 650 nm）获取纯琥珀色和深色枫糖浆以及与常见掺假物（甜菜、玉米和大米糖浆，含量为1 - 50%）混合后的EEMs。使用平行因子分析（PARAFAC）确定EEMs的主要成分，并通过液相色谱 - 串联质谱（LC - MS/MS）进行确认。计算两个最普遍的EEM特征的强度比。分别基于在两个选定激发波长处的发射以及完整的EEM图像开发人工神经网络（ANN）和卷积神经网络（CNN）来分析EEMs，以区分掺假的存在和程度。
样品的EEMs能够识别有价值的鉴别信息。当λex = 290 nm时，λem = 350和425处发射强度之比（I425/I350）识别潜在掺假的功效（正确识别率为70 - 86%）取决于掺假物。该比值对甜菜糖浆掺假特别有效，即使浓度低于2%。应用机器学习算法可提高对所有掺假物的检测能力。ANN能正确识别掺假类型和程度（分别为90%和82%）。CNN方法能准确分类75 - 99%的掺假糖浆，但需要额外的计算能力和更密集的数据集。
本研究有助于基于枫糖浆的固有荧光团提供一种快速、无损且绿色的枫糖浆掺假监测工具。
由于需求高和价格贵，枫糖浆常被其他糖浆掺假。由于DNA降解，DNA条形码技术在检测枫糖浆掺假方面无效。荧光指纹或EEMs可用于在枫糖浆中寻找鉴别标志物。应用于EEM数据的机器学习算法（ANN和CNN）有助于检测。
