MicroRNA profiling as novel biomarkers for detecting gutter oil using machine learning

Researchers from Nanjing University have published a research article in ExRNA, presenting a novel approach for detecting gutter oil using microRNAs (miRNAs) as biomarkers. Researchers demonstrated that miRNAs are present in edible oils and can be used to distinguish between pure and recycled oils. By combining qRT-PCR with machine learning techniques, they characterized miRNA profiles across commercial vegetable oils, animal oils, and gutter oil. Notably, the relative abundances of miR-16 and let-7a differed significantly among these oils, enabling accurate differentiation using a support vector machine (SVM) model. The findings suggest that miRNAs such as miR-16 and let-7a serve as reliable biomarkers, allowing for the classification of gutter oil even when it meets national standards. This study presents a feasible and effective method for detecting gutter oil, with potential applications in enhancing food safety and public health.

Gutter oil is a broad term encompassing various types of substandard oils found in daily life, including recycled kitchen waste oil, repeatedly used frying oil, and discarded animal fats. These "problematic edible oils" pose a significant public health risk to billions of people. As of 2012, an estimated 2 to 3 million tons of gutter oil entered the Chinese market annually, potentially contaminating up to one-tenth of the nation's food supply. In recent years, large-scale illegal gutter oil operations have become rare due to stricter regulations on waste edible oil management. However, emerging issues such as "saliva oil" (oil repeatedly used by multiple customers) and adulterated oils resulting from the improper transportation of non-edible oils in tanker trucks have led to the resurgence of new forms of gutter oil. These variants are more covert and dispersed, making their detection increasingly challenging. As a result, the effective identification of gutter oil and other substandard oils remains a critical issue.

Fundamentally, gutter oil is a mixed oil, comprising various oils from different sources with inconsistent chemical compositions and lipid ratios, often undergoing complex post-processing treatments. "Contrary to common perception, certain refining and alkaline neutralization techniques may be employed during the processing of gutter oil to circumvent regulatory standards, rendering most conventional detection methods specified in the national standard (GB 2716-2018) ineffective against it," says Jiaxin Li, lead author. Existing detection approaches suffer from significant limitations, including susceptibility to processing-related modifications, high testing costs, and interference from cooking and dietary habits.

This study found that microRNAs (miRNAs) can persist in processed gutter oil. Through a systematic investigation, researchers identified seven specific miRNAs—MIR162a, MIR168a, MIR166, MIR156a, let-7a, miR-223, and miR-16—as potential biomarkers. Using qRT-PCR screening, they further determined that four of these miRNAs—let-7a, MIR162a, MIR156a, and miR-16—exhibited significant abundance differences between commercial oils and simulated edible oils (blends of animal and vegetable oils). Moreover, this abundance variation was found to be concentration-dependent.

To validate the ability of let-7a, MIR162a, MIR156a, and miR-16 to distinguish gutter oil, researchers analyzed 37 real-world gutter oil samples. The results showed that MIR162a and MIR156a displayed significant abundance differences between animal oils and other edible oils, but their levels did not differ significantly between vegetable oils and gutter oils. In contrast, miR-16 and let-7a exhibited significant abundance differences across all three oil types in pairwise comparisons.

Subsequently, researchers performed hierarchical clustering by categorizing oils based on two animal-derived miRNAs (miR-16 and let-7a) and two plant-derived miRNAs (MIR162a and MIR156a). The clustering results indicated that the animal-derived miRNAs could correctly classify all oils into three categories: animal oil, gutter oil, and vegetable oil. However, clustering based on plant-derived miRNAs resulted in mixed classifications of vegetable oil and gutter oil, suggesting a lower discriminatory power.

To further enhance the method’s reliability and applicability, the researchers plotted all oil sample data on a two-dimensional plane using miR-16 and let-7a abundance as the x- and y-axes. To identify key differences among oil types, they applied the concept of the maximum-margin hyperplane and employed a support vector machine (SVM) algorithm for machine learning. This approach enabled the derivation of a classification “formula” (decision boundary) that could distinguish oil types based on miRNA abundance. By inputting qRT-PCR-derived abundance values of miR-16 and let-7a into this formula, an oil sample could be accurately classified as animal oil, gutter oil, or vegetable oil. The algorithm was validated on a test dataset, achieving a classification accuracy of 100%.

By integrating qRT-PCR with the SVM algorithm, this study introduces a novel technological approach for identifying gutter oil based on microRNA abundance, offering a powerful tool for ensuring food safety and combating food-related crimes. "Given the nationwide deployment of qRT-PCR equipment during the COVID-19 pandemic, this method is not only more cost-effective than traditional approaches but also more convenient. Moreover, as microRNAs directly reflect the biological origin of oils, they are inherently resistant to forgery and external interference," says Dr. Zhang. Additionally, the incorporation of machine learning enhances the flexibility of this approach. Local health and market regulatory authorities can retrain the model using their own sample data, thereby expanding the training dataset or incorporating additional microRNAs (increasing data dimensionality). This adaptability holds promise for developing a more robust and precise system for assessing oil quality.

Li J, Cong L, Liu Y, Li L, Zhang Y. MicroRNA profiling as novel biomarkers for detecting gutter oil using machine learning. ExRNA 2025(1):0002, https://doi.org/10.55092/exrna20250002.