2007 Annual Report 1a.Objectives (from AD-416) Objective 1 - Develop sampling plans that minimize buyers' risk (bad lots accepted) and sellers' risk (good lots rejected) when detecting quality attributes in agricultural commodities. Sampling plans will be developed to detect the following attributes in foods: aflatoxin in almonds, genetically modified (GM) proteins in grains, peanut proteins (allergens) in foods, foreign material (FM) in shelled peanuts, and TCK spores in wheat. Objective 2 - Determine the percentage of total aflatoxin in farmers' stock peanuts that is partitioned into each milled peanut category (jumbo, medium, number one, splits, oil stock, loose shelled kernels, and discolored or damaged kernels) during the shelling process. Objective 3 - Develop sampling plans for biosecurity purposes that maximize the chance of detecting various food matrices intentionally adulterated with biological and/or chemical agents. 1b.Approach (from AD-416) Objective 1 - Experiments are designed to obtain variability and distributional data to construct operating characteristic (OC) curves which can be used to predict the performance of sampling plan designs to detect specific quality attributes in a food matrix. From the OC curve, the false positives (sellers' risk or good lots rejected) and false negatives (buyers' risk or bad lots accepted) for specific sampling plan can be determined. Objective 2 - Forty 50-kg samples of farmers' stock (FS) peanuts, contaminated with varying levels of aflatoxin, will each be processed in an USDA, ARS pilot shelling plant. Each 50-kg sample will be divided into loose-shelled kernels (LSK) and in-shell peanuts. The in-shell peanuts will be shelled and the shelled kernels will be sized into five commercial peanut grades, jumbo, medium, number ones, splits, and oil stock. Shelled kernels in each grade will be color sorted into accepts and rejects components for a total of 10 categories (5 size grades x 2 accept/reject components). The LSK will also be sized in a like manner into the same five commercial grades and color sorted into accept/reject components or 10 categories. After the peanuts in each of the 20 categories are weighed, the aflatoxin in each of the 20 categories will be measured. From the weights and aflatoxin values in each category, a mass balance can be used to compute aflatoxin in all kernels before sorting by size and color. Then, the percentage of total aflatoxin in the FS peanuts before shelling, sizing, and color sorting that is partitioned into each of the 20 categories will be determined. Objective 3 -The development of mathematical models requires the measurement of chemical or biological agents among replicate samples taken from contaminated lots, from which variability and distributional characteristics provide the basis for development of statistical models and prediction of validity, reliability, and feasibility (false positives and false negatives) associated with various sample plan designs. 3.Progress Report The variability and distribution among aflatoxin sample test results was determined for pistachios. The variance associated with the aflatoxin test procedure was partitioned into sampling, sample preparation, and analytical variances. Regression equations were developed that demonstrated that each variance component was a function of aflatoxin concentration. The observed aflatoxin distribution among sample test results was compared to the negative binomial, compound gamma, and lognormal distributions. Using goodness of fit methods, the negative binomial was chosen to simulate the observed aflatoxin distribution among sample test results. A model was developed, using the variance equations and the negative binomial distribution to predict the effect of various samples size and maximum limits on the performance of sampling plan designs. The Electronic Working Group (EWG) created by the Codex Committee on Contaminants in Foods to recommend a harmonized aflatoxin maximum limit and sampling plan design requested that USDA/ARS predict the performance of aflatoxin sampling plan designs for all 3 treenuts using 1, 2, and 3 samples, a 5, 10, 20, 30, and 40 kg sample size, and a maximum limits of 4, 8, 10, 15, and 20 ng/g total aflatoxin. A website was constructed to display the 75 OC curves along with documents explaining sampling theory for each member of the EWG to review. If a U.S. almond export handler is an active participant of the Voluntary Aflatoxin Sampling Program (VASP), then the EU will test 5% of lots imported from the US almond industry, else the EU will test 100%. To demonstrate the value of a US handler to participate in the VASP program, the US and EU aflatoxin testing system was simulated to show the number of US lots rejected in Europe when a US handler did and did not participating in the VASP program. As of August 2007, 99% of all US almond handlers had signed up for the VASP program. The simulation is also being used to demonstrate to the EU that even when the VASP and EU aflatoxin sampling plans are used perfectly, about 6% of the imported lots will be rejected at destination and zero rejects among imported lots at destination is next to impossible to achieve. Past studies have demonstrated that the binomial distribution can accurately predict the uncertainty and distribution among sample test results when measuring various quality attributes such as foreign material, spotted peanuts, and various grade factors. A generalized model was developed to predict the performance of attribute sampling plans. A website was constructed with a Excel spreadsheet model that would allow handlers to design sampling plans to meet their stated performance standard. Studies to measure Cry9c protein (StarLink) in corn are continuing. The uncertainty associated with sampling, sample preparation, and analysis is being conducted. Previous survey data describing the extent of Cry9c in the market has been shared with EPA who is studying the need to continue mandatory testing of the corn supply for Cry9c protein. USDA/GIPSA, ARS, and FDA are participating in the development of an EPA White paper. 4.Accomplishments Accurate detection of aflatoxin in almonds, hazelnuts, and pistachios –Because of the large variability associated with the test procedure, lots may be misclassified when handlers, exporters, importers, and regulatory agencies test treenuts for aflatoxin, which may cause an economic loss to the industry and increase the health dangers to the consumer. The sampling, sample preparation, and analytical variances were determined for all three treenuts. The variances were compared for all three treenuts and it was determined that a single model based upon the almond data can be developed to evaluate the performance of sampling plan designs. USDA/ARS, as part of the U.S. delegation to the Codex Committee on Contaminants in Foods (CCCF), has asked to use the data to recommend a harmonized sampling plan to detect aflatoxin in treenuts traded in international markets for CCCF approval. The U.S. almond industry, which provides 70% of the world’s demands for almonds, has used the model to develop an industry wide aflatoxin-testing program for almonds marketed in the export trade that would reduce lots rejected upon retesting in the EU. This research addresses National Program 306 problem area 1b (Methods to evaluate and predict quality.) Accurate detection of percent foreign material in bulk lots of shelled peanuts – Because of the variability among replicate sample test results taken from the same lot, it is difficult to get an accurate estimate of the true percent foreign in a bulk lot. The variability among 80 samples from each of four lots was measured. The binomial function was compared to the observed variability and distribution among sample test results and was found to accurately describe the experimental data. A model, based upon the binomial distribution, was developed to predict the effect of sample size on the performance (risk of misclassifying lots) of sampling plan designs for the US peanut industry. This research addresses National Program 306 problem area 1b (Methods to evaluate and predict quality.) 5.Significant Activities that Support Special Target Populations What goes here? 6.Technology Transfer Number of web sites managed 1 Number of non-peer reviewed presentations and proceedings 2 Number of newspaper articles and other presentations for non-science audiences 3 Review Publications Ozay, G., Seyhan, F., Yilmaz, A., Whitaker, T.B., Slate, A. 2006. Sampling hazelnuts for aflatoxin, part i: uncertainty associated with sampling, sample prepatation, and alalysis. Journal of Association of Official Analytical Chemists International 90:1028-1035. Vargas, E.A., Whitaker, T.B., Santos, E.A., Slate, A.B., Lima, F.B., Framca, R.C. 2005. Design of sampling plans to detict ochratoxin a in green coffee. Journal of Food Additives & Contaminants. Whitaker, T.B., Slate, A.B., Jacobs, M., Hurley, M.J., Adams, J.G., Giesbrecht, F.G. 2006. Sampling almonds for aflatoxins; part i: estimation of uncertainty associated with sampling, sample preparation, and analysis. Journal of Association of Official Analytical Chemists International 89:1027-1034. Whitaker, T.B., Slate, A.B., Hurley, M.J., Giesbrecht, F.G. 2006. Sampling Almonds for Aflatoxin, Part II: Estimating Risks Associated with Various Sampling Plans Designs. Journal Association. Official Analytical Chemists, International 90 (3):778-785.