Extrapolation of intrinsic clearance from in vitro systems such as liver microsomes or hepatocytes is an established approach to predict clearance in preclinical species and in humans. A common discussion in the literature is whether the predictive accuracy of such extrapolations is influenced by the chemotype and whether these methods are also applicable to compounds studied in early drug discovery programs. Compounds in such programs are frequently lipophilic and show low solubility and low free fraction in plasma, which may pose challenges to the extrapolation of clearance different from those of the final clinical candidates. A similar discussion has been raised about compounds residing beyond the traditional small-molecule property space, such as PROTACs© and other molecules incompatible with Lipinski's rule-of-five. To further enlighten the field on these matters, we present a study comparing the predictive accuracy between mouse hepatocytes and microsomes for a set of molecules (N = 211) from the Merck Healthcare drug discovery pipeline. This set was dominated by compounds belonging to class 2 and 4 of the extended clearance classification systems (ECCS). It contained a similar proportion of molecules compliant with the Lipinski rule-of-five (N = 127) and molecules lacking such compliance (N = 84). This study showed no or little differences in predictive accuracy nor bias between the two groups, with an average fold error close to 1, an absolute average fold error of just over 2, and around 50% being within 2-fold and >90% being within 5-fold of the predicted unbound clearance in both in vitro systems. Furthermore, no significant differences in accuracy were observed for compounds with an extremely low free fraction (down to 0.05%) in plasma. The accuracy of in vitro-in vivo extrapolation of female CD-1 mouse clearance was not affected by the physicochemical properties.