A total of 323 paired grain and grain dust samples (particle size <1650 μm) were collected from combines at harvest (56%), on-farm bins (28%), and experimental minibins seeded with an ochratoxin A (OTA)/Penicillium verrucosum hot spots (15%) of which >98% were soft red winter wheat. Liquid chromatography-tandem mass spectrometry (LC-MS/MS) was used to detect 21 mycotoxins, including deoxynivalenol (DON) and its plant-conjugated form, deoxynivalenol 3-β-d-glucoside (DON 3-Glc). Except for DON 3-Glc, all mycotoxin concentrations found in grain dust were higher than in grain (p < 0.0030). Pearson correlation coefficients and two-variable regression show a significant (p < 0.0001) linear relationship between the mycotoxin content in grain and that in grain dust with 19 toxins. In only five mycotoxins (DON, OTA, ochratoxin B, citrinin, and enniatin A1), more than 82% of the variation in the data is explained by the two-variable regression model. Because of its higher mean concentration and detection frequency, only DON produced a strong relationship (p < 0.0001, r2 = 0.949) with low root-mean-square error (RMSE) (293.41 ng/g). The results suggest that modeling levels in grain based upon levels in grain dust can be used to estimate DON in grain bulk.