We present a combined experimental and mathematical study of the proteolysis of a surface-bound rubisco protein stain. The adsorption and desorption of subtilisin A (SA) onto and from surface-bound rubisco films were found to be a strong function of the surface chemistry underlying the protein stain; the stain acted as a biosensor able to convey information about the underlying surface to the attacking protease. The apparent protease adsorption rate constants (k(a)) were 0.016 +/- 0.007, 0.014 +/- 0.004 and 0.048 +/- 0.005 min(-1) ppm(-1) while the apparent desorption rate constants (k(d)) were 1.60 +/- 0.15, 1.05 +/- 0.02 and 1.75 +/- 0.05 min(-1) for hydrophobic, neutral-hydrophilic and negatively charged hydrophilic surfaces, respectively. The apparent proteolysis rate constant of surface-bound rubisco and the enzyme deactivation rate constant were estimated to be 1.7 +/- 0.1 ml mg(-1) min(-1) and 0.03 +/- 0.01 min(-1), respectively, independent of underlying surface chemistry. The results demonstrated higher protein removal from the charged hydrophilic surface relative to the other two surfaces. Rubisco cleanability from the charged and hydrophobic surfaces increased with increasing bulk enzyme concentration (and hence surface enzyme concentration) and was better for the charged surface, perhaps reflecting the higher k(a) value. Conversely, rubisco cleanability from the neutral hydrophilic surface was surprisingly insensitive to variation in bulk enzyme concentration. Overall cleaning efficiency was also substantially lower for the neutral hydrophilic surface when compared with the hydrophobic surface, even though k(a) values for each surface were similar. These findings indicate that surface proteolysis is significantly impaired at low values of k(d), suggesting that enzyme mobility at the interface may be closely linked to cleaning performance. The model presented here is expected to be a useful tool in the detergent industries to screen and gauge the cleaning performance of detergent-enzyme formulations, and may also be able to facilitate the design of surface treatments that convey cleaning signals to attacking proteases. (C) 2009 Elsevier Ltd. All rights reserved.