Although lipases are known to catalyze alkoxycarbonylation reactions in organic solvents, the existing knowledge base on their substrate specificity in alkoxycarbonylation reaction is sparse. Moreover, models to predict substrate specificity have not been reported. Here, we report the experimentally measured rate constants for 180 acyl donor-alcohol pairs and demonstrate the two-step synthesis of over 70 disubstituted carbonate products from simple precursors such as diphenyl carbonate and alcohols. The efficiency of synthesis was found to be dependent on the order of alcohol addition. This motivated the need to develop a model to predict lipase specificity in alkoxycarbonylation reactions. A partial least square model has been developed to correlate the reaction rate with (i) descriptors of alcohol for a fixed acyl donor, (ii) descriptors of acyl donor for a fixed alcohol, (iii) descriptors of both the acyl donor and the alcohol. The number of descriptors being far greater than the number of observations was a potential limitation in the model development. This was addressed by selecting a subset of descriptors using a systematic procedure based on (a) correlation among the descriptors and step-wise regression methodology, and (b) variable influence on projection methodology. The model was able to accurately predict the reaction rate and the optimal order of addition of alcohols in the two-step synthesis of disubstituted carbonates using the enzyme mixture. The descriptor subset and the relevant model would benefit the users of lipases in synthetic applications while the modeling strategy presented here can have applications in predicting specificity of other enzymes. (c) 2006 Wiley Periodicals, Inc.