A general model for describing gradient elution separations of peptides and proteins by reversed-phase high-performance liquid chromatography (HPLC) has been presented previously. This model has now been modified so that it can be applied to any of the four HPLC methods used for separating biological macromolecules: reversed-phase, ion-exchange, hydrophobic-interaction and size-exclusion chromatography, carried out in either an isocratic or gradient elution mode. The role of sample molecule structure and the particular column used has been further studied, so that previous empirical parameters for different column/sample choices can now be estimated from three physical properties of the sample and the column: sample molecular weight, native vs. denatured sample, column packing pore diameter. This eliminates much of the empiricism of our preceding model, and minimizes the number of experimental runs now required in order to apply the model in practice. The final model has been tested for several hundred runs involving peptides and proteins in the molecular weight range 600-162,000, all four of these HPLC methods, in both isocratic and gradient elution modes, and using data from several different laboratories (including our own). The model is able to predict bandwidth in HPLC separations of proteins and peptides with an accuracy of +/- 17% (1 standard deviation), for the case of "well-behaved" separations. Separations that are not "well-behaved" will give wider bands than predicted by the model.