Refining single-kernel maize sorting to reduce mycotoxin contamination
By Ruben Chavez, PHL Scholar
My research focuses on single-kernel spectral identification of mycotoxins in corn. I started working with commercial corn from the U.S. as a master’s student at the University of Illinois Department of Food Science and Human Nutrition. My principal advisor, Dr. Matthew Stasiewicz, helped me to apply to the ADMI postharvest loss fellowship to test my research in other countries. In 2019, I was granted the fellowship, and I was able to continue my research in mycotoxin identification.
The PHL Scholar award allowed me to transition into a doctoral degree program and find new ways to reduce mycotoxin contamination through spectral identification and with the goal of preventing postharvest losses in corn.
In 2019, my first research objective was to test a small-scale kernel sorting system as an avenue to remove mycotoxin-contaminated corn from the food supply. To plan the experiment required to achieve the goal, Dr. Stasiewicz traveled to Ghana to meet with the team and collaborators from the Feed the Future Innovation Lab for the Reduction of Post-Harvest Loss (PHLIL). The trip gave us insights into current best practices in the Dormaa region of Ghana.
We observed that there was room for improvement and a potential application for kernel sorting on poultry farms. We also noticed that most poultry farmers had a specific maize cleaning machine based on size separation. After the site visit, I proposed a detailed experimental design for evaluating the effect of kernel sorting technologies in conjunction with local grain cleaning technology, both on grain that is well-dried and stored, or poorly stored.
However, COVID-19 affected our research activities. Our first goal plan was to perform an on-site sample collection in summer of 2020. Due to pandemic travel constraints, we had to adapt our plans and redirect our strategy. Ultimately, the team determined the best path forward was to work directly with a local PHLIL Ghana project collaborator to have them collect samples and prepare them for shipment to the U.S. This way, the first phase of in-country kernel sorting work could progress without international travel.
Once that decision was made in late summer 2020, we were able to develop a detailed protocol for collecting maize samples from farms with and without local cleaning capacity, stratifying by well- and poorly stored conditions as defined by storage moisture content, and characterizing the efficacy of local sorting ability to reduce mycotoxins.
Due to the delays in sample collection, I spent significant research time refining the risk-based sampling of individual kernels to recover maize kernels with mycotoxin contamination. Previous literature has shown that visual characteristics associated with mycotoxin could be used to identify mycotoxin-contaminated corn. These features include black and moldy kernels, broken kernels, and bright green-yellow fluorescence under ultra-violet light.
With this idea in mind, I decided to test it in some preliminary experiments by comparing random selection and risk-based selection of contaminated corn kernels. The risk-based sampling recovered contaminated kernels at a rate of 10% for aflatoxin contamination (29/300 kernels > 20 ppb) and 20% for fumonisin contamination (59/300 > 2 ppm). Those results showed significant improvement over random sampling, with a recovery rate of 1% (7/864) and 5% (47/864) for aflatoxin and fumonisin, respectively. These results helped develop a new hypothesis that will later be implemented in my research. The new research hypothesis focused on using high-risk visual features associated with mycotoxin contamination to develop spectral classification algorithms for mycotoxin sorting.
In spring 2021, with the help of a local PHLIL Ghana project collaborator, Dr. Stasiewicz and I conducted a remote sample collection from poultry farms. We imported 76 samples of maize from poultry farms in the Dormaa-Ahenkro area in Ghana. With these new samples, our updated goal was to train our spectral sorter based on high-risk features associated with mycotoxin contamination and test with the imported samples to assess a reduction in mycotoxin contamination.
Previous studies used wet chemistry coupled with spectral analysis to classify kernels. Our approach focused on a method that avoids wet chemistry and relies on visual features for the classification of mycotoxin. We believed this method could offer a cost-effective solution for mycotoxin classification.
Overall, sorting based on high-risk features associated with mycotoxin contamination showed a significant aflatoxin reduction (p < 0.001, 73/76 samples reduced, mean reduction 31 ppb, range -9.7 – 67 ppb). We observed similar results in fumonisin contamination. All models showed a significant fumonisin reduction (p < 0.001, mean reduction 1.9 ppm, range 9.3×10-2 –6.1 ppm).
From the accepted stream, 61/76 samples tested <15 ppb aflatoxin, significantly more than the 40/76 before sorting (p < 0.001). For fumonisin, all accepted samples tested <2 ppm concentration compared to only 2/76 before sorting. In terms of rejected mass, we could observe that only a small fraction was rejected after sorting. The results showed that the average mass rejected was 12% (range 1.2% – 36%), and that the rejected mass contained an average of 46% of the total aflatoxin (range 4.3% – 97%) and 88% of the total fumonisin (range 10% – 84%).
With these results, we were able to prove that visual characteristics associated with mycotoxin contamination can inform classification models, which can enable sorting contaminated corn to reduce aflatoxin and fumonisin contamination.
With my advisor’s guidance, I am planning an in-country validation experiment. The goal is to repeat the sorting experiment in Ghana with samples collected from poultry farmers based on the connections of the PHLIL Ghana team. We were able to ship the spectral sorter to Ghana, and I created standard procedures and video-supporting materials to train the Ghana team for spectral sorting.
Soon, we will receive an update with the results of the experiments and assess if the sorting machine can be used in a different sample set while maintaining the same spectral training data from our U.S.-based experiment. We expect to find positive results that could help us develop this technology further and find a scalable solution for improved mycotoxin risk management and post-harvest loss reduction in corn.
I am highly grateful for this fellowship. It has allowed me to conduct applied research in food safety and security. I have developed my research skills further and improved skills in critical thinking, networking, and communication skills. I am excited to see the results of the latest on-site experiments.
I hope that in the future, I can continue to develop new scientific discoveries that will help improve food safety and security for consumers and reduce postharvest loss associated with mycotoxin contamination. My short-term goal is to become a better researcher and continue to publish more journal articles related to food safety, and in the future, become an expert in my field driving change in regulations and industry standards.