The fusion of therapeutic proteins to the Fc domain of monoclonal antibodies (mAbs) generally improves the proteins’ pharmacokinetic (PK) characteristics, extending in vivo half-lives due to the binding of the Fc domain to the FcRn receptor. Yet, several of these Fc-fusion biologics have been observed to have unexpected rapid clearance associated with non-specific off-target binding. Variability in non-specific clearance is often challenging to predict, not well understood, and ultimately can delay the drug development process. In this investigation, we present a computational approach leveraging in silico protein structural properties to extend a physiologically based pharmacokinetic (PBPK) model of mAbs validated on in vivo plasma PK profiles in mice. Selected model parameters affecting protein half-life have been scaled by analytical functions of a panel of calculated in silico protein properties identified by a novel and ad hoc symbolic regression procedure. The resulting extended model has been successfully validated against an independent set of protein plasma PKs, indicating that it can generalize to novel biologics of the same class. Moreover, the extended PBPK model has a median absolute average fold error (AAFE) of 1.18 (min = 1.09; max = 1.51), where values less than 2 typically indicate a good fit. The results enable the de-risking of aberrant PK behaviors, ultimately leading to the selection of Fc-fusion proteins with increased therapeutic value for patients.