Abstract: The application describes a method for predicting chemical or biological properties, e.g. toxicity, mutagenicity, etc., of complex multicomponent mixtures from 2D separation date, e.g. GC-MS. The data are resolved into peaks (C) and spectra (S) for individual components by an automated curve resolution procedure (GENTLE). The resolved peaks are then integrated and the characteristic area, separation parameter and associated spectrum combined to yield a predictor matrix (X), which is used as input to a multivariate regression model. Partial least squares (PLS) are used to correlate the 2D separation date for a training set to the measured property. The regression model can then be used to predict the property for other samples.