Many biocatalytic reaction systems are biphasic with a reactive and an immiscible non-reactive phase. The reactive, mostly aqueous phase provides a natural enzyme environment and the non-reactive phase serves for delivery of dissolved substrates at high concentrations and for extraction of products. The proper choice of the non-reactive phase will have manifold influence on the catalytic parameters, such as activity, selectivity, and stability, but also on the maximum obtainable conversion or yield. Conversion or yield constitutes a concise target of practical relevance for rational solvent screening which requires thermodynamic information on coupled reaction and phase equilibria as input information. As long as the reactive phase is kept constant, only the partition equilibrium of each solute in any solvent combination has to be determined. The experimental determination of these data requires a considerable laboratory effort. Therefore, an in-silico screening of solvents for maximal conversion of alcohol dehydrogenase-catalysed oxidoreductions of prochiral ketones was evaluated. COSMO-RS was used for the prediction of solute partitioning between organic solvents and aqueous reaction medium. Although significant absolute deviations were found, COSMO-RS still predicted the correct trends for the partition coefficients of solutes in different solvents. The calculated overall reaction equilibrium using these partition coefficients again resulted in the prediction of the correct best solvent regarding conversion. (C) 2007 Elsevier B.V. All rights reserved.