A variety of biophysical methods used to study proteins requires protein modification using conjugated molecular probes. Cysteine is the main residue that can be modified without the risk of altering other residues in the protein chain. It is possible to label several cysteines in a protein using highly selective labeling reactions, if the cysteines react at very different rates. The reactivity of a cysteine residue introduced into an exposed surface site depends on the fraction of cysteine in the deprotonated state. Here, it is shown that cysteine reactivity differences can be effectively predicted by an electrostatic model that yields site-specifically the fractions of cysteinate. The model accounts for electrostatic interactions between the cysteinyl anion and side chains, the local protein backbone, and water. The energies of interaction with side chains and the main chain are calculated by using the two different dielectric constants, 40 and 22, respectively. Twenty-six mutants of Escherichia coli adenylate kinase were produced, each containing a single cysteine at the protein surface, and the rates of the reaction with 5,5'-dithiobis(2-nitrobenzoic acid) (Ellman's reagent) were measured. Cysteine residues were chosen on the basis of locations that were expected to allow modification of the protein with minimal risk of perturbing its structure. The reaction rates spanned a range of 6 orders of magnitude. The correlation between predicted fractions of cysteinate and measured reaction rates was strong (R = 92%) and especially high (R = 97%) for cysteines at the helix termini. The approach developed here allows reasonably fast, automated screening of protein surfaces to identify sites that permit efficient preparations of double- or triple-labeled protein.