A Machine Learning pipeline for the automated DNA-based classification of Italian olive cultivars: from leaves to olive oil

The origin and identity of our food is becoming an essential question for the modern consumer. Therefore, it becomes imperative to adopt and employ technologies that enhance consumers’ trust. Among different applications towards this matter, the utilization of molecular markers as a tool for genetic characterization of olive oil and/or table olives are gaining ground. These DNA-based tools not only enhance product transparency, but bring a significant economic, environmental and societal positive impact to the industry. In the herein study we present a machine learning (ML) pipeline targeting the automated classification of olive cultivars by post-processing real-time PCR High-Resolution Melting (HRM) outputs using Single Nucleotide Polymorphism (SNP). As a case study we focus on four economically important Italian olive cultivars, namely Leccino, Moraiolo, Dolce d’Agogia and Rajo, with overall classification accuracy of the ML model reaching 0.92 as trained using raw data of 682 HRM experiments. Additionally, for the first time olive oil samples are classified using the herein proposed ML pipeline, adding promising results for future work targeting the varietal composition of olive oil. Concluding, the ML pipeline accelerate time to results, increases consistency as it minimizes human bias, and eliminates the necessity of certified reference samples.