Finding the exact location of a tyrosyl radical in an enzyme is an important step in elucidating the enzyme's mechanism. Although the electron paramagnetic resonance (EPR) spectrum of a tyrosyl radical strongly depends on the radical's conformation and environment, no methodology is available for locating the tyrosine residue responsible for the spectrum solely on the basis of the crystal structure of an enzyme. A method of predicting tyrosyl radical EPR spectra from first principles should encompass a computational methodology that yields a close correlation between the calculated and empirically obtained EPR parameters. An approach is developed that allows a clear comparison of these two groups of parameters on an array of different tyrosyl radicals. The B3LYP density functional method is used to calculate the EPR parameters (spin densities and g- and A-tensors) for tyrosyl radicals of four different enzymes. The calculated EPR parameters are compared to those derived from empirical data by an algorithm. The results of the DFT calculations are found to be in good agreement with the principal dependences employed by the algorithm. However, the calculated EPR parameters do not correlate with empirical data on the whole set of the four radicals studied. The correlation is significantly improved when hydrogen bonds to the radical are formed, both in the case of the protein radicals and in a radical model, when hydrogen bonds to the radical are formed. Therefore, it is suggested that a full correlation between the calculated and the algorithm-derived empirical EPR parameters might be achieved when water molecules are included in the calculation model.