There is considerable interest in developing microscale (i.e., high-throughput) methods that enable multiple filtration experiments to be run in parallel with smaller sample amounts and thus reduce the overall required time and associated cost to run the filtration tests. Previous studies to date have focused on simply evaluating the filtration capacity, not the separation performance. In this work, the stirred-well filtration (SWF) method was used in combination with design-of-experiment (DOE) methods to optimize the separation performance for three binary mixtures of biomolecules: protein-protein, protein-polysaccharide, and protein-DNA. Using the parallel based format of the SWF method, eight constant-flux ultrafiltration experiments were conducted at once to study the effects of stirring conditions, permeate flux, and/or solution conditions (pH, ionic strength). Four separate filtration tests were conducted for each combination of process variables; in total, over 100 separate tests were conducted. The sieving coefficient and selectivity results are presented to match the DOE design format and enable a greater understanding of the effects of the different process variables that were studied. The method described herein can be used to rapidly determine the optimal combination of process factors that give the best separation performance for a range of membrane-based separations applications and thus obviate the need to run a large number of traditional lab-scale tests. (C) 2016 Wiley Periodicals, Inc.