Interlaboratory benchmarking of differential scanning fluorimetry reveals sources of variability in protein thermal stability parameters

Differential scanning fluorimetry (DSF) is widely used to assess protein thermal stability and to compare protein variants, ligands, or formulations with limited sample consumption. Yet, the reproducibility of DSF derived parameters across laboratories, instruments and analysis workflows remains poorly defined. Here, we report a two phase interlaboratory benchmark using a shared VHH sample across 26 laboratories, 28 instruments and seven DSF instrument models. In Phase 1, participants applied their routine acquisition and analysis workflows, whereas in Phase 2, acquisition conditions were harmonised. Datasets were analysed locally and then centrally reprocessed with MoltenProt. Participant reported apparent melting temperatures (Tm) showed good within instrument repeatability but moderate inter instrument dispersion. Harmonisation of acquisition conditions did not substantially reduce this dispersion. Conversely, centralised reprocessing narrowed the distribution among retained datasets, indicating contributions from analysis workflows and data acceptance criteria in addition to instrument dependent factors. Comparison with differential scanning calorimetry (DSC) further showed that apparent Tm depended strongly on the fluorescence readout. For this VHH, single wavelength fluorescence at 350 nm showed the closest agreement with the DSC transition temperature, whereas the 350/330 nm ratio and barycentric mean generally yielded higher values. Furthermore, DSF derived apparent unfolding enthalpies were more variable than apparent Tm. Together, these findings show that DSF can be highly repeatable within a defined instrument and workflow, but that quantitative comparability across instruments depends on instrument specific effects, the choice of fluorescence readout and analysis strategy. Improved comparability will require transparent reporting of acquisition and analysis conditions, explicit data quality criteria, and well characterised reference materials to monitor instrument dependent offsets.