Human metabolic profiles for substances such as toxic food-derived compounds are usually allometrically extrapolated from traditionally determined in vivo rat concentration profiles. To evaluate internal exposures in humans without any reference to experimental data, physiologically based pharmacokinetic (PBPK) modeling could be used if the model input parameters could be estimated in silico. This approach would simplify the use of PBPK models for forward dosimetry after oral doses. In this study, the in silico estimation of input parameters for PBPK models (i.e., fraction absorbed × intestinal availability, absorption rate constants, and volumes of the systemic circulation) was updated for an panel of 355 chemicals (212 previously analyzed and 143 additional substances) using a light gradient boosting machine learning algorithms (LightGBM) based on between 11 and 29 in silico-calculated chemical descriptors. Simplified human PBPK models were then used to calculate virtual maximum plasma concentrations (C_max) and areas under the concentration–time curve (AUC) based on two sets of input parameters, i.e., traditionally derived values from in vivo data and those calculated in silico using the current updated systems. Both sets of C_max and AUC data were well correlated (r = 0.87 and r = 0.73, respectively; p < 0.01, n = 355). Therefore, input parameters for human PBPK models for a diverse range of compounds could be successfully estimated using chemical descriptors and in silico tools. This approach to pharmacokinetic modeling has potential for application in computational toxicology and in the clinical setting for assessing the potential risk of general chemicals.