Tacrolimus, an immunosuppressant drug, presents a narrow therapeutic window and a large pharmacokinetic variability with poor correlation between drug dosing regimen and blood concentration. The objective was to identify predictive factors influencing tacrolimus trough concentrations (C0) using a bottom-up approach. A physiologically based pharmacokinetic (PBPK) model of tacrolimus was proposed, taking into account the body weight, the proportion of fat (P(fat)), hematocrit, lipid fraction of organs, typical intrinsic clearance (CLi(typ)), CYP3A5 genotype of liver donor, plasma unbound fraction of tacrolimus (fu(p)), and concomitant drugs (CYP3A4 inhibitors). For the evaluation of the PBPK model, mean C0 and concentrations 2 h after oral dose of tacrolimus were compared with those from 66 liver transplant recipients included in a multicentric pharmacokinetic study and were found very close. Tacrolimus concentration profiles were simulated in a virtual population defined by a set of covariate values similar to those from the real population. The sensitivity of tacrolimus C0 with respect to each covariate has been tested to identify the most influential ones. With the range of covariate values tested, the impact of each covariate on tacrolimus C0 may be ranked as follows: fu(p), CLi(typ), bioavailability, body weight, hematocrit, CYP3A5 polymorphism, P(fat), and CYP3A4 inhibitory drug-drug interactions. Values for initial dosing regimen of tacrolimus in order to reach a C0 of 10 ng/ml at day 5 (assuming a constant dosing schedule) as a function of CYP3A5 donor genotype and patient's hematocrit and body weight are proposed.