To predict the isomer shifts of Fe complexes in different oxidation and spin states more accurately, we have performed linear regression between the measured isomer shifts (delta(exp)) and DFT (PW91 potential with all-electron triple-zeta plus polarization basis sets) calculated electron densities at Fe nuclei [rho(0)] for the Fe(2+,2.5+) and Fe(2.5+,3+,3.5+,4+) complexes separately. The geometries and electronic structures of all complexes in the training sets are optimized within the conductor like screening (COSMO) solvation model. Based on the linear correlation equation delta(exp) = alpha[rho(0) - 11884.0] + C, the best fitting for 17 Fe(2+,2.5+) complexes (totally 31 Fe sites) yields alpha = -0.405 +/- 0.042 and C = 0.735 +/- 0.047 mm s(-1). The correlation coefficient is r = -0.876 with a standard deviation of SD = 0.075 mm s(-1). In contrast, the linear fitting for 19 Fe(2.5+,3+,3.5+,4+) complexes (totally 30 Fe sites) yields alpha = -0.393 +/- 0.030 and C = 0.435 +/- 0.014 mm s(-1), with the correlation coefficient r = -0.929 and a standard deviation SD = 0.077 mm s(-1). We provide a physical rationale for separating the Fe(2+,2.5+) fit from the Fe(2.5+,3+,3.5+,4+) fit, which also is clearly justified on a statistical empirical basis. Quadrupole splittings have also been calculated for these systems. The correlation between the calculated (DeltaE(Q(cal))) and experimental (DeltaE(Q(exp))) quadrupole splittings based on |DeltaE(Q(exp))| = A |DeltaE(Q(cal))| + B yields slope A, which is almost the ideal value 1.0 (A = 1.002 +/- 0.030) and intercept B almost zero (B = 0.033 +/- 0.068 mm s(-1)). Further calculations on the reduced diferrous and oxidized diferric active sites of class-I ribonucleotide reductase (RNR) and the hydroxylase component of methane monooxygenase (MMOH), and on a mixed-valent [(tpb)Fe3+(mu-O)(mu-CH3CO2)Fe4+(Me3aneN3)]2+ (S = 3/2) complex and its corresponding diferric state have been performed. Calculated results are in very good agreement with the experimental data.