# Single nucleotide polymorphism information estimates breed and variety composition ratio in food

## TL;DR

This paper introduces a new method using SNP data to estimate the breed or variety composition in food mixtures, showing better accuracy than previous approaches.

## Contribution

A novel SNP-based method using non-negative least squares optimization to estimate breed or variety composition ratios in food.

## Key findings

- The method achieved significantly lower average absolute error compared to a uniform probability baseline in simulated cow and cacao datasets.
- The method accurately identified the majority breed or variety in mixed samples with high accuracy compared to equal probability assumptions.

## Abstract

The quality of food products can be influenced by the breed or variety of origin, as well as the composition ratios in mixtures of breeds or varieties. We present a method to estimate the breed or variety composition ratio in food samples using single-nucleotide polymorphism (SNP) allele frequency data and a non-negative least squares (NNLS) optimization approach. To evaluate the method's performance, we simulated two datasets (cow and cacao) containing simulated samples with specified breed or variety composition ratios, then compared the predicted ratios to the actual values. Results show that the method estimates the composition ratios of breeds and varieties with significantly lower average absolute error than a uniform probability baseline (4.1 % vs 24.6 % for cows, p-value = 1.9 × 10−17; and 11.8 % vs 24.6 % for cacao, p-value = 1.1 × 10−8). Additionally, the accuracy of identifying the majority breed or variety in a sample is also significantly higher than assuming equal probability of breed mixing (92 % vs 28 % for cows and 72 % vs 28 % for cacao). The corresponding code for the breed or variety composition ratio estimation is available in the Github repository: (https://github.com/IBPA/NNLS-SNP).

Image 1

•A novel tool is developed to estimate breed or variety composition ratios in food using SNP allele frequency data.•The tool verified both the composition ratio and the majority breed or variety in simulated mixed food samples and can be applied to real datasets in the future.•The method requires only the reference SNP allele frequencies of the breed or variety to be evaluated and does not require model training.

A novel tool is developed to estimate breed or variety composition ratios in food using SNP allele frequency data.

The tool verified both the composition ratio and the majority breed or variety in simulated mixed food samples and can be applied to real datasets in the future.

The method requires only the reference SNP allele frequencies of the breed or variety to be evaluated and does not require model training.