Source: FLORIDA INTERNATIONAL UNIVERSITY submitted to
MACHINE LEARNING APPROACH TO IDENTIFY PARAMETERS AFFECTING MYCOTOXIN RISK IN US WINES
Sponsoring Institution
National Institute of Food and Agriculture
Project Status
COMPLETE
Funding Source
Reporting Frequency
Annual
Accession No.
1012019
Grant No.
2017-67018-26229
Cumulative Award Amt.
$258,253.00
Proposal No.
2016-10392
Multistate No.
(N/A)
Project Start Date
May 1, 2017
Project End Date
Apr 30, 2022
Grant Year
2017
Program Code
[A1331]- Improving Food Safety
Recipient Organization
FLORIDA INTERNATIONAL UNIVERSITY
(N/A)
MIAMI,FL 33199
Performing Department
Sch of Hosp and Tourism Mgmt
Non Technical Summary
This project aims to develop, design, prototype and test a machine learning system that can be used to predict the likelihood of mycotoxin risk in US wines based on various climactic, pre-harvest, and post-harvest factors. Mycotoxins are secondary metabolites produced by filamentous fungi that grow on many fruits, cereals, and nuts. They can cause illness and disease in both humans and animals, in very low concentrations (e.g., parts-per-billion). Economic losses in the agricultural sector due to mycotoxins in the US are approximately $1 billion per year resulting from reduced crop value and yield due to mycotoxin contamination, loss in animal production, and human health costs. Recognizing the threat posed by mycotoxins in wines, we seek to assess a multitude of mycotoxins never before assayed in US wines using complementary HPLC-MS/MS chemical analyses. Combining this novel data set with information on the climate, pre-harvest, and production factors affecting each individual wine we will design, implement, and validate a machine learning system to predict the likelihood of mycotoxin risk in US based wines. This innovative research will be the first of its kind to create a machine learning system to accurately predict mycotoxin likelihood in wine that can aid farmers and wine producers in curbing mycotoxin presence in their wines. Moreover, because much of this data will be based on climactic factors, which strongly mediate mycotoxigenic fungal growth, we predict that this machine learning system may be potentially adaptable to predicting mycotoxin contamination of a wide range of other valuable agricultural commodities.
Animal Health Component
(N/A)
Research Effort Categories
Basic
50%
Applied
(N/A)
Developmental
50%
Classification

Knowledge Area (KA)Subject of Investigation (SOI)Field of Science (FOS)Percent
3141131200050%
3141131208050%
Goals / Objectives
The proposed research aims to develop, design, prototype and test a Machine Learning (ML)system that can be used to predict the likelihood of mycotoxin risk in US based wines. We proposethis applied research in agricultural engineering with the primary goals of 1) developing a riskbased approach to control foodborne hazards aimed at reducing the incidences during production,harvest, post-harvest storage, or processing and 2) identifying and characterizing emerging orunder-researched hazards that are known or expected to cause foodborne disease. Specifically, thefoodborne hazards we will examine are mycotoxins affecting wines and grapes, including severalnot previously assessed in the US wine industry.
Project Methods
Approximately 500 wines from distinct geographical locations, with defined compositions will be aquired. Additionally, the conditions of their production, including climatic, pre-harvest, and production parameters will be collected. These wines will next be chemically analyzed for the presence of at least 14 different mycotoxins. Following chemical analysis, all the data concerning wines including their climatic, pre-harvest, production, and mycotoxin contents, will be entered into a database. This database will be analyzed using a new Machine Learning system to determine significant correlations between independent variables climatic, pre-harvest, and production factors with the dependent variables of mycotoxin amounts. Through this analysis, we anticipate being able to determine climatic, pre-harvest, and production parameters of wine making that significantly contribute to the accumulation of mycotoxins in wine.

Progress 05/01/21 to 04/30/22

Outputs
Target Audience: Nothing Reported Changes/Problems:The project encountered obstacles due to the breakdown of LC-MS instrumentation intended for mycotoxin analyses. To overcome this challenge, analyses utilized equivalent instrumentation available through the FIU Advanced Mass Spectrometry Facility (AMSF), and directed funds for materials and supplies to the charges associated with analyses by this re-charge center. What opportunities for training and professional development has the project provided?The project provided funding for two graduate Ph.D. students (Ariel Lawson and Mark Annunziato) in Summer 2019 and 2021 to travel to the University of Leipzig (Leipzig, Germany) to receive training and conduct research with instrumentation and techniques related to high-resolution magic angle spin nuclear magnetic resonance (HRMAS NMR) for metabolic profiling, specifically in relation to toxicity of mycotoxins, and identification of biomarkers for exposure and effect. In addition, one of the graduate students (Mark Annunziato) led mass spectrometry analysis of mycotoxins, and received as a result, extensive training with LC-MS and associated techniques. How have the results been disseminated to communities of interest?Dissemination was through academic publications and conferences. What do you plan to do during the next reporting period to accomplish the goals?The grant period is finished. There is data analysis to be completed and manuscripts remain to be written. These will be tackled in the coming months.

Impacts
What was accomplished under these goals? Due to the delays caused by the failure of the LC-MS instrument, data generation was delayed. After an alternative instrument was identified, all the LC-MS data has been generated for the wines under study. The analysis of the data remains to be performed. The two aforementioned aims can be achieved after the analyses are completed.

Publications

  • Type: Journal Articles Status: Published Year Published: 2022 Citation: Eeza, M.N.H., Bashirova, N., Zuberi, Z., Matysik, J., Berry, J.P. and Alia, A. (2022). An integrated systems-level model of ochratoxin A toxicity in the zebrafish (Danio rerio) embryo based on NMR metabolic profiling. Sci. Rep. 12: 6341.
  • Type: Other Status: Other Year Published: 2022 Citation: Annunziato, M., Stieglitz, J., Benetti, D., Grosell, M., Eeza, M.N.H, Bashirova, N., Matysik, J., Alia, A. and Berry, J.P. Metabolic alterations by the mycotoxin, zearalenone, as measured by high-resolution magic angle spin NMR in early life stages of zebrafish (Danio rerio) and Olive Flounder (Paralichthys olivaceus). Toxicology, in preparation.
  • Type: Other Status: Other Year Published: 2022 Citation: Lawson, A., Eeza, M.N.H., Bashirova, N., Matysik, J., Alia, A. and Berry, J.P. Comparative toxicometabolomics of trichothecene mycotoxins in the zebrafish embryo model by high-resolution magic-angle spin NMR. Manuscript in Preparation for Toxins
  • Type: Other Status: Other Year Published: 2022 Citation: Bashirova, N., Eeza, M.N.H., Annunziato, M., Lawon, A., Matysik, J., Alia, A. and Berry, J.P. Identification of biomarkers, and an integrated model, of fumonisin B1 toxicity in the zebrafish embryo model based on high-resolution magic-angle spin (HRMAS) NMR metabolomics. Manuscript in Preparation for World Mycotox J
  • Type: Other Status: Other Year Published: 2022 Citation: Annunziato, M., Chozo, J., Cozart, J., Garcia, G., Spoor, V., Hernandez, Sommers, A., Welch, A., Alonso, M., Narasimhan, G. and Berry, J.P. Assessment of potential drivers and exposure risks of mycotoxins in wines from the U.S. Manuscript in Preparation for J. Agric. Food Chem.
  • Type: Conference Papers and Presentations Status: Published Year Published: 2022 Citation: Shi, Jain, Narasimhan. Time Series Forecasting (TSF) Using Various Deep Learning Models, International Journal of Computer and Systems Engineering Vol:16, No:04, 2022


Progress 05/01/17 to 04/30/22

Outputs
Target Audience:Academics in toxicology of wines and agricultural engineering. Changes/Problems: The project encountered obstacles due to the breakdown of LC-MS instrumentation intended for mycotoxin analyses. To overcome this challenge, analyses utilized equivalent instrumentation available through the FIU Advanced Mass Spectrometry Facility (AMSF), and directed funds for materials and supplies to the charges associated with analyses by this re-charge center. What opportunities for training and professional development has the project provided? The project provided funding for two graduate Ph.D. students (Ariel Lawson and Mark Annunziato) during Summer semester of 2019. In Summer 2021, they traveled to the University of Leipzig (Leipzig, Germany) to receive training and conduct research with instrumentation and techniques related to high-resolution magic angle spin nuclear magnetic resonance (HRMAS NMR) for metabolic profiling, specifically in relation to toxicity of mycotoxins, and identification of biomarkers for exposure and effect. In addition, one of the graduate students (Mark Annunziato) led mass spectrometry analysis of mycotoxins and received extensive training with LC-MS and associated techniques. How have the results been disseminated to communities of interest? Dissemination was through academic publications and conferences. What do you plan to do during the next reporting period to accomplish the goals?Since the grant period is finished, the data analysis and tools will be completed and manuscripts will be written in the coming months.

Impacts
What was accomplished under these goals? The research aimedto develop, design, prototype, and test a Machine Learning (ML) system that can be used to predict the likelihood of mycotoxin risk in US-based wines. In order to examine the mycotoxins in wines and grapes, LC-MS spectra from over 400 wines were collected. Due to delays caused by COVID-19 pandemic and the breakdown of the LC-MS instrument, data were generated only in early 2022. Data analysis and risk-based hazard prediction tools will be achieved in the coming months.

Publications


    Progress 05/01/20 to 04/30/21

    Outputs
    Target Audience: Nothing Reported Changes/Problems: Nothing Reported What opportunities for training and professional development has the project provided?Students could not be trained due to COVID-19 issues and instrumentation problems. How have the results been disseminated to communities of interest? Nothing Reported What do you plan to do during the next reporting period to accomplish the goals? Plans to achieve the stated aims in the next year are laid out.

    Impacts
    What was accomplished under these goals? Very little progress was made due to COVID-19 issues and instrumentation problems.

    Publications


      Progress 05/01/19 to 04/30/20

      Outputs
      Target Audience: Nothing Reported Changes/Problems: Nothing Reported What opportunities for training and professional development has the project provided? Nothing Reported How have the results been disseminated to communities of interest?Yes. There have been four publications. What do you plan to do during the next reporting period to accomplish the goals?Due to COVID-19, much of the work slowed down during the final quarter of the third year of the project. We have requested and were subsequently granted a one-year no-cost extension. Therefore, during the final reporting period, the mycotoxin analysis will be completed. The remaining time will be spent analyzing the pre-harvest data, finalizing the analysis of the mycotoxin data, and creating the dataset for exploring various Machine Learning techniques to develop a predictive model of toxin presence in U.S. wines based on the gathered features and data.

      Impacts
      What was accomplished under these goals? Toward detection and quantitation of mycotoxins in wines - in support of the machine learning approach - a "multi toxin" analytical technique, specifically based on liquid chromatography/mass spectrometry (LC-MS), was developed and validated during the most recent funding period. The LC-MS method was previously developed (see 2018-19 annual report), and this year focused on validating the effectiveness and efficiency of the extraction technique for mycotoxins toward application to the analyses of wine samples. Progress toward analysis was unfortunately hampered by instrumental issues that needed to be diagnosed and repaired. These have now been resolved. However, alongside this (and to supplement LC-MS analyses), a method based on high-performance liquid chromatography with fluorescence detection (HPLC-FD) was developed as an alternative, to quantify a subset of the mycotoxins (i.e., ochratoxin [OTA], zearalenone [ZEN]). Delays due to the COVID-19 pandemic have further hindered progress on the analyses of wine samples, but the samples have been prioritized for this Summer (with approval of the investigators as "essential personnel" to complete these studies). The research was additionally continued, as part of the modified aims of the project, on the use of high-resolution magic angle spin (HRMAS) NMR, coupled to intact zebrafish embryo as a toxicological model, for metabolic profiling to identify and characterize biomarkers of exposure, and toxicological pathways, associated with mycotoxins. This research builds on a technique previously developed with, and specifically carried-out in, the laboratories of Dr. Berry's collaborator, Dr. A. Alia, at the University of Leipzig (Germany) and Leiden University (Netherlands). The grant supported travel for the Co-PI (Berry), and two Chemistry Ph.D. students (Mark Annunziato and Ariel Lawson), to the laboratories of Dr. Alia to perform toxin exposures and NMR analyses in Summer 2019. This research previously generated metabolomics data for aflatoxin B1 (AFB1; Zuberi et al., 2019) and OTA (Eeza et al., in preparation); the manuscript for the latter will be completed and submitted (to PLoS One) in Summer 2020. In Summer 2019, NMR data were additionally collected for several other mycotoxins including fumonisin B1 (FB1), T-2 and nivalenol (NIV). Data are currently being analyzed with manuscripts anticipated for each of these latter mycotoxins in the coming year. This research represents a growing focus of the Co-PI's research group, and the data accumulated for mycotoxins represents one of its kind data sets relevant to the toxicology of these mycotoxins.Additionally, the pre-harvest data was collected from wineries that produced all the wine samples.

      Publications

      • Type: Journal Articles Status: Published Year Published: 2019 Citation: Zuberi, Z., Eeza, M.N.H., Matyski, J., Berry, J.P. and Alia, A. (2019) NMR-based metabolic profiles of intact zebrafish embryos exposed to aflatoxin B1 recapitulates hepatotoxicity and supports possible neurotoxicity. Toxins, 11: 258.
      • Type: Journal Articles Status: Other Year Published: 2020 Citation: Eeza, M.N.H., Zuberi, Z., Matysik, J., Berry, J.P. and Alia, A. (2020) High-resolution magic angle spin (HRMAS) NMR of intact zebrafish reveals toxicological pathways, and potential metabolomic biomarkers, of ochratoxin A. PLoS One, in preparation (to be submitted Summer 2020)
      • Type: Conference Papers and Presentations Status: Accepted Year Published: 2020 Citation: Yudong Tao, Erik Coltey, Tianyi Wang, Miguel Alonso Jr., Mei-Ling Shyu, Shu-Ching Chen, Hadi Al-haffar, Albert Elias, Biayna Bogosian, Shahin Vassigh, Confidence Estimation Using Machine Learn-ing in Immersive Learning Environments, accepted for publication, IEEE 3rd International Conferenceon Multimedia Information Processing and Retrieval, August 6-8, 2020, Shenzhen, Guangdong, China.
      • Type: Conference Papers and Presentations Status: Submitted Year Published: 2020 Citation: E. Coltey, Y. Tao, T. Wang, M.P. Reyes, M. Alonso Jr., S. Vassigh, S. Chen, M. Shyu, AI-Enabled Im-mersive Learning Environment Based on Dynamic Multi-modal Content Adaptation and Monitoring,ACM Multimedia, 2020 (In Review).


      Progress 05/01/18 to 04/30/19

      Outputs
      Target Audience: Nothing Reported Changes/Problems: Nothing Reported What opportunities for training and professional development has the project provided? Nothing Reported How have the results been disseminated to communities of interest?Yes. There has been one publication accepted into the journal Toxins in 2019. Zain Zuberi, Muhamed Eeza, Joerg Matysik, John Berry and A Alia (2019). NMR-Based Metabolic Profiles of Intact Zebrafish Embryos Exposed to Aflatoxin B1 Recapitulates Hepatotoxicity, and Supports Possible Neurotoxicity, Toxins, May 2019. What do you plan to do during the next reporting period to accomplish the goals?During the next reporting period, the mycotoxin analysis will be complete. The remaining time will be spent gathering the pre-harvest conditions of these wines. Machine Learning techniques such as Natural Language Processing will be explored to perform the data gathering. Once the dataset has been completed, several machine learning models will be trained on the data to predict toxin levels based on the gathered featres and data.

      Impacts
      What was accomplished under these goals? Toward detection and quantitation of mycotoxins in wines - in support of the machine learning approach - a "multi toxin" analytical technique, specifically based on liquid chromatography/mass spectrometry (LC-MS), was developed and validated during the most recent funding period.The method developed was shown to effectively resolve 10 major mycotoxins found in wines including ochratoxin A (OTA), patulin (PAT), aflatoxin B1 and G1 (AFB1 and AFG1), zearalenone (ZEA), and the "Fusarium toxins" including trichothecenes (T-2 and deoxynivalenol [DON]) and fumonisins B1 and B2 (FB1 and FB2, respectively), as well as "Alternaria toxins" including alternariol (ALT).Quantitation of these toxins have demonstrated limits of detection (LOD) and limits of quantitation (LOQ) which are, in most cases, sufficiently sensitive for detection at typical levels in wines as follows: LOD (ppb) LOQ (ppb) OTA 0.4 0.9 PAT 3.5 11.2 AFB1 0.5 1.3 AFG1 1.2 2.1 ZEA 2.9 3.7 T-2 0.3 0.6 DON 7.2 13.3 ALT 2.5 5.2 Currently, we are validating a requisite method for efficient extraction of mycotoxins from wine samples, specifically based on liquid/liquid extraction (LLE), and the method is being applied (beginning in Summer 2019) to the >600 collected wine samples. In addition, as part of a modified aim of the project, research was conducted on the use of high-resolution magic angle spin (HRMAS) NMR, coupled to intact zebrafish embryo as a toxicological model, for metabolic profiling to identify and characterize biomarkers of exposure, and toxicological pathways, associated with mycotoxins. This research builds on a technique previously developed with, and specifically carried-out in, the laboratories of Dr. Berry's collaborator, Dr. A. Alia, at the University of Leipzig (Germany) and Leiden University (Netherlands). The grant supported travel for the Co-PI (Berry), and two Chemistry Ph.D. students (Mark Annunziato and Ariel Lawson), to the laboratories of Dr. Alia to perform toxin exposures and NMR analyses. Research in the current funding period continued (in May/June 2019) work initiated in the previous period (June 2018). In 2018, the initial research collected HRMAS NMR data for OTA and AFB1. Results from the latter toxin (i.e., AFB1) were recently published (Zuberi et al., 2019) in the open-access journal Toxins; briefly, these findings demonstrated a clear role of hepatoxicity of AFB1, aligned with previous ex vivo studies in other systems, and identified several interrelated pathways of toxicity and possible exposure biomarkers. A manuscript on the finding with respect to OTA is currently in preparation. The most recent work (in May/June 2019) expanded metabolic profiling data to four additional mycotoxins, i.e., ZEA, NIV, T-2 and FB1. NMR analyses have been performed, and data analysis is currently toward generation of manuscripts (three in total planned) in the coming funding period. This research represents a growing focus of the Co-PI's research group, and the data accumulated for mycotoxins represents a one of its kind data set relevant to the toxicology of these mycotoxins.

      Publications

      • Type: Journal Articles Status: Accepted Year Published: 2019 Citation: Zain Zuberi, Muhamed Eeza, Joerg Matysik, John Berry and A Alia (2019). NMR-Based Metabolic Profiles of Intact Zebrafish Embryos Exposed to Aflatoxin B1 Recapitulates Hepatotoxicity, and Supports Possible Neurotoxicity, Toxins, in press.


      Progress 05/01/17 to 04/30/18

      Outputs
      Target Audience: Nothing Reported Changes/Problems: Nothing Reported What opportunities for training and professional development has the project provided? Nothing Reported How have the results been disseminated to communities of interest? Nothing Reported What do you plan to do during the next reporting period to accomplish the goals?During the next reporting period, we will begin the analysis of these wines for various mycotoxins. We will also complete the curation of the wine samples database.

      Impacts
      What was accomplished under these goals? We have collected over 400 wines that will be the subject of analyses for mycotoxins. We are in the process of obtaining information on the pre-harvest conditions of these wines.

      Publications