“…Experimental attributes such as the chemical nature of the columns through McReynolds coefficients (x', y', z', u', and s'), column geometry (length [L, m], internal diameter [ID, mm] and film thickness [PE, μm]) and other operational aspects of the chromatographic methods, such as the modulation period (PM, s), the carrier gas flow rate (mL min −1 ), carrier gas viscosities at the oven temperatures (GV, Po) and heating rates (A, • C min −1 ) were combined to predict retention times in the first and second dimensions (t1D, min and t2D, s). For this, two DNN-based models were developed: one specialized in 1 The architecture of the networks is presented in Figure 1. Briefly, the input data and outputs were range scaled, and the constructed neural networks had 18 neurons in the input layer (representing the variables under investigation), four or six neurons in the first hidden layer, three or four neurons in the second hidden layer (representing the processing units), and one in the output layer (representing the results predicted by the models).…”