# Thermal shift assay

A thermal shift assay (TSA), also called [differential scanning fluorimetry](https://www.edgepedia.org/differential-scanning-fluorimetry) (DSF) or ThermoFluor, measures the thermal stability of a purified protein by heating it and detecting unfolding, most often through the fluorescence of a dye that binds hydrophobic surfaces exposed as the protein unfolds.<sup>[1](https://doi.org/10.1038/nprot.2007.321)</sup> The assay reports an apparent melting temperature \( (T_{\mathrm{m}}) \) and the shift \( \Delta T_{\mathrm{m}} \) caused by a ligand or buffer condition, and it is used to detect ligand binding, screen for stabilizing compounds, and optimize protein formulations and buffer conditions.<sup>[2](https://pmc.ncbi.nlm.nih.gov/articles/PMC10663957/)</sup> Because it runs on a conventional real-time PCR instrument and delivers results in a single day, it is a standard inexpensive screen in early drug discovery and in quality control of proteins for structural studies.<sup>[1](https://doi.org/10.1038/nprot.2007.321)</sup>

| Key fact | Detail |
|---|---|
| Other names | DSF, ThermoFluor, TSA; ThermoFluor refers to the dye-based form of these assays, while DSF and TSA more broadly also cover label-free and other variants<sup>[2](https://pmc.ncbi.nlm.nih.gov/articles/PMC10663957/)</sup> |
| Standard conditions | 5 µM protein with 5× SYPRO Orange (10 µM), heated 25–95 °C at 1 °C/min; a Current Protocols unit instead recommends 2× dye with 5 µM protein<sup>[2](https://pmc.ncbi.nlm.nih.gov/articles/PMC10663957/)</sup><sup> • </sup><sup>[3](https://pmc.ncbi.nlm.nih.gov/articles/PMC4332540/)</sup> |
| Reported quantity | \( \Delta T_{\mathrm{m}} \), the tested condition's \( T_{\mathrm{m}} \) minus the control \( T_{\mathrm{m}} \)<sup>[2](https://pmc.ncbi.nlm.nih.gov/articles/PMC10663957/)</sup> |
| Typical shift sizes | Ligands at or near their \( K_{\mathrm{D}} \) often give \( \Delta T_{\mathrm{m}} \) of 1–12 °C; a \( \Delta T_{\mathrm{m}} \) above 2.0 °C over solvent control is a common first-pass hit threshold<sup>[2](https://pmc.ncbi.nlm.nih.gov/articles/PMC10663957/)</sup><sup> • </sup><sup>[3](https://pmc.ncbi.nlm.nih.gov/articles/PMC4332540/)</sup> |
| Failure rate | An estimated 15–25% of recombinant proteins give non-ideal denaturation curves that preclude analysis<sup>[3](https://pmc.ncbi.nlm.nih.gov/articles/PMC4332540/)</sup> |
| Throughput | Dye-based DSF runs up to 384 or 1536 samples simultaneously; nanoDSF measures 48 capillaries in parallel<sup>[4](https://pmc.ncbi.nlm.nih.gov/articles/PMC10903397/)</sup><sup> • </sup><sup>[5](https://onlinelibrary.wiley.com/doi/10.1002/pro.3622)</sup> |
| Affinity caveat | No general correlation exists between ΔTm and binding affinity<sup>[6](https://link.springer.com/article/10.1007/s12551-025-01280-3)</sup> |

## How it works

Heating drives the protein from its folded state to an unfolded state, exposing hydrophobic regions that bind an environment-sensitive dye such as SYPRO Orange; the resulting fluorescence increase traces the unfolding transition.<sup>[1](https://doi.org/10.1038/nprot.2007.321)</sup> The Tm is the temperature at which 50% of the sample is folded and 50% unfolded.<sup>[7](https://doi.org/10.1007/s12551-020-00619-2)</sup> A ligand that preferentially binds and stabilizes the folded state raises the \( T_{\mathrm{m}} \), and \( \Delta T_{\mathrm{m}} \) is the difference between the ligand-containing and control \( T_{\mathrm{m}} \) values.<sup>[2](https://pmc.ncbi.nlm.nih.gov/articles/PMC10663957/)</sup> The shift has a simple thermodynamic meaning only when the receptor is saturated with ligand; at lower concentrations, melting can appear multiphasic because ligand redistributes as the protein unfolds.<sup>[8](https://pmc.ncbi.nlm.nih.gov/articles/PMC4906806/)</sup> \( T_{\mathrm{m}} \) is extracted either by fitting the sigmoidal fluorescence curve to a [Boltzmann equation](https://www.edgepedia.org/boltzmann-equation) or as the minimum of the first derivative \( (-dF/dT) \) plot.<sup>[3](https://pmc.ncbi.nlm.nih.gov/articles/PMC4332540/)</sup>

Several readouts report the same transition. Label-free variants monitor intrinsic tryptophan fluorescence at 330 nm and 350 nm.<sup>[5](https://onlinelibrary.wiley.com/doi/10.1002/pro.3622)</sup> The thiol-reactive dye CPM is used for membrane proteins,<sup>[9](https://liebertpub.com/doi/10.1089/adt.2011.0390)</sup> the detergent-tolerant GloMelt dye (excitation 468 nm, emission 507 nm) offers an alternative to SYPRO Orange, and [Thioflavin T](https://www.edgepedia.org/thioflavin-t), ProteoStat, and molecular rotor dyes such as DCVJ and CCVJ report an aggregation temperature \( (T_{\mathrm{agg}}) \) rather than a \( T_{\mathrm{m}} \).<sup>[10](https://pmc.ncbi.nlm.nih.gov/articles/PMC10855643/)</sup>

## How it is done

A typical assay uses 20 µL in 96- or 384-well plates with 5–10 µM protein and takes 20–120 min per run.<sup>[10](https://pmc.ncbi.nlm.nih.gov/articles/PMC10855643/)</sup> Protein is combined with SYPRO Orange, commonly 5× from a 5000× stock, and dispensed with ligand or buffer additive into the plate; DMSO should stay below 2% in the final reaction.<sup>[11](https://cmi.hms.harvard.edu/sites/g/files/omnuum4016/files/2025-03/CMI%20QuantStudio%20DSF%20Getting%20Started%20Guide.pdf)</sup><sup> • </sup><sup>[12](https://www.ncbi.nlm.nih.gov/books/NBK586874/)</sup> The plate is heated from 25 °C to 95 °C (or 99 °C) at about 1 °C/min while fluorescence is recorded.<sup>[2](https://pmc.ncbi.nlm.nih.gov/articles/PMC10663957/)</sup> SYPRO Orange cannot detect protein below roughly 500 nM, and a tenfold higher concentration is usually used for reasonable signal-to-background.<sup>[10](https://pmc.ncbi.nlm.nih.gov/articles/PMC10855643/)</sup> Recommended controls include a no-protein control and a ligand-only control, since ligands that bind the dye can mask protein–dye interactions, with at least four replicates per condition.<sup>[13](https://genomics2.lsu.edu/documentation/vendor/qPCR_Protein_Thermal_Shift_Studies%28software-vB.0%29.pdf)</sup> BSA must be kept out of the assay buffer because it binds the dye and gives very high background.<sup>[11](https://cmi.hms.harvard.edu/sites/g/files/omnuum4016/files/2025-03/CMI%20QuantStudio%20DSF%20Getting%20Started%20Guide.pdf)</sup> Hit calling uses a ΔTm threshold over solvent control, with 2.0 °C suggested for a first-pass small-molecule screen<sup>[3](https://pmc.ncbi.nlm.nih.gov/articles/PMC4332540/)</sup> and shifts above 1 °C counted as significant by other authors.<sup>[10](https://pmc.ncbi.nlm.nih.gov/articles/PMC10855643/)</sup>

## Origin

Fluorescence-monitored thermal denaturation of proteins has been used since at least the 1960s.<sup>[6](https://link.springer.com/article/10.1007/s12551-025-01280-3)</sup> The high-density miniaturized thermal shift assay for drug discovery, under the name ThermoFluor, was reported by Michael W. Pantoliano and colleagues in 2001.<sup>[14](https://doi.org/10.1177/108705710100600609)</sup> Early versions used dyes such as ANS or dapoxyl sulfonic acid before SYPRO Orange proved the most favorable,<sup>[12](https://www.ncbi.nlm.nih.gov/books/NBK586874/)</sup> and SYPRO Orange itself had been introduced as a protein gel stain by Thomas H. Steinberg and colleagues in 1996.<sup>[15](https://doi.org/10.1006/abio.1996.0319)</sup> Mei-Chu Lo and colleagues evaluated the assay for hit identification in 2004,<sup>[16](https://doi.org/10.1016/j.ab.2004.04.031)</sup> and Daumantas Matulis and colleagues showed in 2005 that ThermoFluor could yield binding affinity and stoichiometry for carbonic anhydrase.<sup>[17](https://doi.org/10.1021/bi048135v)</sup> Ulrika B. Ericsson and colleagues applied Thermofluor to high-throughput stability optimization of proteins for structural studies in 2006,<sup>[18](https://doi.org/10.1016/j.ab.2006.07.027)</sup> and A protocol published in Nature Protocols popularized DSF on real-time PCR instruments.<sup>[1](https://doi.org/10.1038/nprot.2007.321)</sup>

## Variants

[The 1](https://www.edgepedia.org/the-1) °C/min continuous-ramp protocol runs on a wide range of real-time thermocyclers and temperature-controlled plate readers, including the Stratagene M3005P, Opticon2, iCycler/iQ, Rotorgene, LightCycler, and several [Applied Biosystems](https://www.edgepedia.org/applied-biosystems) models.<sup>[8](https://pmc.ncbi.nlm.nih.gov/articles/PMC4906806/)</sup> nanoDSF removes the extrinsic dye, relying on intrinsic tryptophan fluorescence at 330 nm and 350 nm.<sup>[5](https://onlinelibrary.wiley.com/doi/10.1002/pro.3622)</sup><sup> • </sup><sup>[7](https://doi.org/10.1007/s12551-020-00619-2)</sup> The Prometheus NT.48 measured 48 capillaries of 15 µL in parallel, heated 20–90 °C at 1 °C/min, with a minimum reliable protein concentration of 15 µg/mL, and worked in detergent-containing solutions, but NanoTemper discontinued the NT.48 and NT.Plex instruments as of November 6, 2023; the current model is the Prometheus Panta, which handles up to 48 capillaries (or 24 in a capillary chip) with a 15–110 °C temperature range.<sup>[5](https://onlinelibrary.wiley.com/doi/10.1002/pro.3622)</sup><sup> • </sup><sup>[7](https://doi.org/10.1007/s12551-020-00619-2)</sup><sup> • </sup><sup>[19](https://nanotempertech.com/prometheus/nt48-software/)</sup> DSF-GFP uses a GFP tag as the unfolding reporter; GFP only starts losing fluorescence around 75 °C, so a GFP-only control is needed to distinguish compounds binding GFP from compounds binding the target.<sup>[7](https://doi.org/10.1007/s12551-020-00619-2)</sup> The cellular thermal shift assay (CETSA) extends thermal unfolding to compound-target engagement in cell lysates and intact cells without protein purification.<sup>[4](https://pmc.ncbi.nlm.nih.gov/articles/PMC10903397/)</sup> High-throughput thermal scanning (HTTS) inverts the logic, screening libraries of protein variants for relative thermal stability, and calculates an apparent \( T_{\mathrm{m}} \) with a variation on the Clarke and Fersht (1993) equation.<sup>[20](https://pmc.ncbi.nlm.nih.gov/articles/PMC2701314/)</sup><sup> • </sup><sup>[21](https://doi.org/10.1021/bi00067a022)</sup> DSLS on the StarGazer-384 is a label-free aggregation-based alternative measuring \( T_{\mathrm{agg}} \) that is insensitive to detergents and usable for membrane proteins.<sup>[9](https://liebertpub.com/doi/10.1089/adt.2011.0390)</sup> Protein-adaptive DSF (paDSF), combining a 312-dye Aurora library with a protein–dye matching procedure, was applied successfully to 94% (66 of 70) of proteins, tripling previous dye compatibility, and Tm values from Aurora dyes and SYPRO Orange correlate strongly (Pearson \( R = 0.97 \) across 2,211 dye–protein pairings).<sup>[22](https://www.nature.com/articles/s41587-024-02158-7)</sup>

## Applications

A protein can typically be screened in a 384-well plate against up to 384 compounds or buffer conditions.<sup>[9](https://liebertpub.com/doi/10.1089/adt.2011.0390)</sup> Screening MurF, a bacterial cell wall biosynthesis enzyme, against 200,000 compounds by ThermoFluor yielded 17 hits, one inhibiting MurF with an \( IC_{50} \) of 25 µM.<sup>[9](https://liebertpub.com/doi/10.1089/adt.2011.0390)</sup> Beyond ligand screening, the assay optimizes buffer, pH, and additive conditions for protein formulation and for structural studies.<sup>[18](https://doi.org/10.1016/j.ab.2006.07.027)</sup> In fragment workflows, cross hits are confirmed with orthogonal methods such as [X-ray crystallography](https://www.edgepedia.org/x-ray-crystallography), SPR, NMR, or DSF itself.<sup>[23](https://link.springer.com/article/10.1007/s12551-025-01359-x)</sup>

## Limitations and alternatives

An estimated 15–25% of recombinant proteins give non-ideal curves with high baseline fluorescence or no sigmoidal transition.<sup>[3](https://pmc.ncbi.nlm.nih.gov/articles/PMC4332540/)</sup> A protein must unfold within roughly 15–85 °C to be accessible on a qPCR instrument,<sup>[2](https://pmc.ncbi.nlm.nih.gov/articles/PMC10663957/)</sup> and proteins with \( T_{\mathrm{m}} \) above 98 °C produce melt curves with no distinct melt phase; multi-domain and oligomeric proteins show multi-state unfolding with multiple melt phases.<sup>[13](https://genomics2.lsu.edu/documentation/vendor/qPCR_Protein_Thermal_Shift_Studies%28software-vB.0%29.pdf)</sup>

Dye-specific artifacts are a distinct failure mode. SYPRO Orange is quenched at pH below 5 or above 10, shows high background with micelles, membranes, and viscosities above about 10%, and reducing agents, EDTA, Triton X-100, peptides, compounds, or metal ions can activate the dye without any protein unfolding.<sup>[2](https://pmc.ncbi.nlm.nih.gov/articles/PMC10663957/)</sup> Dye binding itself can destabilize the protein and underestimate \( T_{\mathrm{m}} \), because dye molecules selectively associate with aromatic residues and bias the equilibrium toward the unfolded state, an effect more pronounced in proteins rich in Arg, Glu, Trp, and Phe.<sup>[4](https://pmc.ncbi.nlm.nih.gov/articles/PMC10903397/)</sup>

Compared with alternatives, differential scanning calorimetry (DSC) is the label-free reference method that directly measures thermodynamic parameters, but it runs sequentially at roughly 90 min per sample and offers low screening throughput.<sup>[3](https://pmc.ncbi.nlm.nih.gov/articles/PMC4332540/)</sup><sup> • </sup><sup>[10](https://pmc.ncbi.nlm.nih.gov/articles/PMC10855643/)</sup> Ligand \( K_{\mathrm{D}} \) values calculated from DSF may be less reliable than those from isothermal titration calorimetry (ITC), time-resolved FRET, or fluorescence polarization.<sup>[2](https://pmc.ncbi.nlm.nih.gov/articles/PMC10663957/)</sup>

\( \Delta T_{\mathrm{m}} \) reflects binding affinity at the melting temperature, which can differ substantially from affinity at a 25 °C reference temperature because of temperature-dependent entropy–enthalpy balance and heat capacity effects; no general correlation exists between ligand-induced \( \Delta T_{\mathrm{m}} \) and binding affinity.<sup>[6](https://link.springer.com/article/10.1007/s12551-025-01280-3)</sup> Covalent and other high-affinity binders can cause \( \Delta T_{\mathrm{m}} \) of 10–30 °C or more, as with streptavidin–biotin,<sup>[10](https://pmc.ncbi.nlm.nih.gov/articles/PMC10855643/)</sup> while weak binding fragments below around 1 µM often do not stabilize proteins detectably, giving frequent false negatives.<sup>[23](https://link.springer.com/article/10.1007/s12551-025-01359-x)</sup> DSF therefore produces false negatives for enthalpy-driven binding with large unfavorable entropic or heat-capacity contributions, and false positives for entropy-driven binding nearly balanced by unfavorable enthalpy, so it is best restricted to screens identifying strong binders without affinity ranking.<sup>[6](https://link.springer.com/article/10.1007/s12551-025-01280-3)</sup> A remedy for quantification is isothermal analysis: at a single temperature near the transition, the fraction of folded protein at each ligand concentration follows coupled unfolding \( (K_{\mathrm{U}}) \) and unbinding \( (K_{\mathrm{d}}) \) equilibria, allowing \( K_{\mathrm{d}} \) determination with only two free parameters.<sup>[24](https://www.nature.com/articles/s41598-018-37072-x.pdf)</sup>

## References

1. [Frank H Niesen, Helena Berglund, Masoud Vedadi (2007). The use of differential scanning fluorimetry to detect ligand interactions that promote protein stability. Nature Protocols.](https://doi.org/10.1038/nprot.2007.321)
2. [Protocol for performing and optimizing differential scanning fluorimetry experiments (STAR Protocols, 2023)](https://pmc.ncbi.nlm.nih.gov/articles/PMC10663957/)
3. [Analysis of protein stability and ligand interactions by thermal shift assay (Current Protocols in Protein Science)](https://pmc.ncbi.nlm.nih.gov/articles/PMC4332540/)
4. [Thermal unfolding methods in drug discovery](https://pmc.ncbi.nlm.nih.gov/articles/PMC10903397/)
5. [Facile measurement of protein stability and folding kinetics using a nano differential scanning fluorimeter (Protein Science)](https://onlinelibrary.wiley.com/doi/10.1002/pro.3622)
6. [Differential scanning fluorimetry followed by microscale thermophoresis and/or isothermal titration calorimetry as an efficient tool for ligand screening (Biophysical Reviews, 2025)](https://link.springer.com/article/10.1007/s12551-025-01280-3)
7. [Kai Gao, Rick Oerlemans, Matthew R. Groves (2020). Theory and applications of differential scanning fluorimetry in early-stage drug discovery. Biophysical Reviews.](https://doi.org/10.1007/s12551-020-00619-2)
8. [Recent Developments in the Use of Differential Scanning Fluorometry in Protein and Small Molecule Discovery and Characterization](https://pmc.ncbi.nlm.nih.gov/articles/PMC4906806/)
9. [Thermal Denaturation Assays in Chemical Biology (Assay and Drug Development Technologies)](https://liebertpub.com/doi/10.1089/adt.2011.0390)
10. [Fluorescence-Based Protein Stability Monitoring, A Review (2024)](https://pmc.ncbi.nlm.nih.gov/articles/PMC10855643/)
11. [Harvard CMI QuantStudio DSF Getting Started Guide](https://cmi.hms.harvard.edu/sites/g/files/omnuum4016/files/2025-03/CMI%20QuantStudio%20DSF%20Getting%20Started%20Guide.pdf)
12. [Chapter 15 Thermofluor-Based Analysis of Protein Integrity and Ligand Interactions (Methods in Molecular Biology)](https://www.ncbi.nlm.nih.gov/books/NBK586874/)
13. [qPCR Protein Thermal Shift Studies(software vB.0) (genomics2.lsu.edu)](https://genomics2.lsu.edu/documentation/vendor/qPCR_Protein_Thermal_Shift_Studies%28software-vB.0%29.pdf)
14. [Michael W. Pantoliano and colleagues (2001). High-Density Miniaturized Thermal Shift Assays as a General Strategy for Drug Discovery. SLAS DISCOVERY.](https://doi.org/10.1177/108705710100600609)
15. [Thomas H. Steinberg and colleagues (1996). SYPRO Orange and SYPRO Red Protein Gel Stains: One-Step Fluorescent Staining of Denaturing Gels for Detection of Nanogram Levels of Protein. Analytical Biochemistry.](https://doi.org/10.1006/abio.1996.0319)
16. [Mei-Chu Lo and colleagues (2004). Evaluation of fluorescence-based thermal shift assays for hit identification in drug discovery. Analytical Biochemistry.](https://doi.org/10.1016/j.ab.2004.04.031)
17. [Daumantas Matulis and colleagues (2005). Thermodynamic Stability of Carbonic Anhydrase: Measurements of Binding Affinity and Stoichiometry Using ThermoFluor. Biochemistry.](https://doi.org/10.1021/bi048135v)
18. [Ulrika B. Ericsson and colleagues (2006). Thermofluor-based high-throughput stability optimization of proteins for structural studies. Analytical Biochemistry.](https://doi.org/10.1016/j.ab.2006.07.027)
19. [Prometheus NT.48 Software - NanoTemper Technologies](https://nanotempertech.com/prometheus/nt48-software/)
20. [High-throughput thermal scanning: a general, rapid dye-binding thermal shift screen for protein engineering](https://pmc.ncbi.nlm.nih.gov/articles/PMC2701314/)
21. [Jane Clarke, Alan R. Fersht (1993). Engineered disulfide bonds as probes of the folding pathway of barnase: Increasing the stability of proteins against the rate of denaturation. Biochemistry.](https://doi.org/10.1021/bi00067a022)
22. [Protein-adaptive differential scanning fluorimetry using conformationally responsive dyes (Nature Biotechnology, 2024)](https://www.nature.com/articles/s41587-024-02158-7)
23. [High-throughput investigation of macromolecular interactions for drug development using spectral shift technology (Biophysical Reviews, 2025)](https://link.springer.com/article/10.1007/s12551-025-01359-x)
24. [Isothermal Analysis of ThermoFluor Data can readily provide Quantitative Binding Affinities (Scientific Reports)](https://www.nature.com/articles/s41598-018-37072-x.pdf)

---
*Topic: Encyclopedia › Life and health › Biological foundations › Biochemistry and metabolism › Biochemistry field and methods › Biochemical methods and techniques › Assay techniques*

*Initially written Sep 29, 2026 · Edited: Oct 15, 2026*

*Copyright 2026 EdgeChat AI, a subsidiary of Biostate AI.*

License: Edgepedia Community License 1.0, https://www.edgepedia.org/edgepedia/license