Summary
A bioRxiv preprint benchmarking differential scanning fluorimetry across 26 laboratories found good repeatability within instruments but moderate differences between them. The fluorescence readout and analysis workflow also affected apparent protein melting temperatures.
A bioRxiv preprint benchmarking differential scanning fluorimetry (DSF) across 26 laboratories found that apparent protein melting temperatures were repeatable within a given instrument but showed moderate dispersion between instruments. The study used the same VHH protein sample across 28 instruments representing seven DSF models, and found that both the fluorescence readout and data-analysis approach affected the results.
DSF estimates a protein’s apparent melting temperature, or Tm, by tracking fluorescence as the sample is heated. The method is used to assess protein thermal stability and compare protein variants, ligands or formulations. The benchmark’s findings show why numerical results from different instruments or analysis workflows may not be directly interchangeable.
A two-phase comparison across laboratories
In the first phase, participating laboratories used their routine acquisition and analysis workflows. In the second, acquisition conditions were harmonised. Researchers analysed the data locally and then centrally reprocessed them using MoltenProt.
Participant-reported Tm values showed good repeatability within instruments, alongside moderate dispersion between instruments. Harmonising acquisition conditions did not substantially reduce that between-instrument spread. Central reprocessing narrowed the distribution among the datasets retained for analysis, pointing to analysis workflows and data-acceptance criteria as contributors to variability as well as instrument effects.
Fluorescence readout changed the apparent temperature
The team also compared DSF results with differential scanning calorimetry (DSC), a method that measures heat changes during a transition. For the VHH tested, single-wavelength fluorescence at 350 nm gave values closest to the DSC transition temperature. The 350/330 nm fluorescence ratio and barycentric mean generally produced higher values. Apparent unfolding enthalpies—estimates related to the heat involved in unfolding—varied more than apparent Tm.
The authors say improving comparability will require clear reporting of acquisition and analysis conditions, explicit criteria for accepting data, and well-characterised reference materials to monitor instrument-dependent offsets. The benchmark tested one shared VHH, so its results describe this interlaboratory comparison rather than establishing the size of such differences for every protein. The work is a preprint; its authors also disclose that several coauthors are employees of instrument companies.