Dielectrophoresis (DEP) is a versatile tool for the precise microscale manipulation of a broad range of substances. To unleash the full potential of DEP for the manipulation of complex molecular-sized particulates such as proteins requires the development of appropriate theoretical models and their comprehensive experimental verification. Here, we construct an original DEP platform and test the Holzel-Pethig empirical model for protein DEP. Three different proteins are studied: lysozyme, BSA, and lactoferrin. Their molecular Clausius-Mossotti function is obtained by detecting their trapping event via the measurement of the fluorescence intensity to identify the minimum electric field gradient required to overcome dispersive forces. We observe a significant discrepancy with published theoretical data and, after a very careful analysis to rule out experimental errors, conclude that more sophisticated theoretical models are required for the response of molecular entities in DEP fields. The developed experimental platform, which includes arrays of sawtooth metal electrode pairs with varying gaps and produces variations of the electric field gradient, provides a versatile tool that can broaden the utilization of DEP for molecular entities.