Validation of binary non-targeted methods: mathematical framework and experimental designs

Abstract

Non-targeted approaches (NTAs) are being increasingly developed and adopted to detect food fraud and identify the authenticity of the food. Method validation is a critical step before bringing the NTAs to the routine, to be assured of method performance and trust that its outcomes are reliable. However, the paucity of well structured and harmonized validation strategies has been one of the hurdles, withholding the exploitation of NTAs. This report aims to describe a validation framework for methods that involve binary classification, which are prevalent in non-targeted workflows. We foresee this work to contribute to filling the current gap in the provisions for NTA method validation; perhaps also push the dialogue further to collectively resolve existing challenges.

Essential Terminology and Considerations for Validation of Non-targeted Methods

“…Assuming the decision limit is 2, about 10% of Italian olive oil samples are misidentified as originating from Turkey (FPR of 10%). Examples of validation procedures fitting to a general statistical model for this approach and corresponding validation parameters have been recently described by Uhlig and colleagues [51]. Such a construct for validation approach B was also one of the central themes of the work by Alewijn et al [25].…”

Towards a Mathematical Framework for Method Validation and Performance Characteristics of Non-targeted Methods for Food Authenticity

“…Assuming the decision limit is 2, about 10% of Italian olive oil samples are misidentified as originating from Turkey (FPR of 10%). Examples of validation procedures fitting to a general statistical model for this approach and corresponding validation parameters have been recently described by Uhlig and colleagues [49]. Such a construct for validation approach B was also one of the central themes of the work by Alewijn et al [24].…”

Untargeted metabolomic approaches in food authenticity: A review that showcases biomarkers

“…The number of samples for each modality (e.g., origin, species, quality…) should be enough to reflect natural variations like seasoning, terroirs, processes variations, etc. Recommendations for sample size range from 60 to 200 samples for each modality (McGrath et al, 2018, Uhlig et al, 2021. Once this design step is completed, the first step in analysis can begin with sampling.…”