Protein Hydrophobicity Calculator
Calculate Grand Average of Hydropathicity (GRAVY) score and predict protein solubility.
What GRAVY Measures
The grand average of hydropathy is the mean Kyte–Doolittle value across every residue. Positive means overall hydrophobic, negative means hydrophilic.
| Residue | Kyte–Doolittle value |
|---|---|
| Isoleucine | +4.5 — most hydrophobic |
| Valine | +4.2 |
| Leucine | +3.8 |
| Alanine | +1.8 |
| Glycine | −0.4 |
| Lysine | −3.9 |
| Arginine | −4.5 — most hydrophilic |
| GRAVY | Interpretation |
|---|---|
| > +0.5 | Likely membrane protein |
| −0.5 to +0.5 | Ambiguous |
| < −0.5 | Soluble, cytoplasmic |
| < −1.0 | Highly soluble, often good expression |
Why Whole-Protein GRAVY Misleads
Averaging across an entire sequence hides local structure. A single-pass membrane protein with one 20-residue transmembrane helix and 400 soluble residues has a negative GRAVY despite genuinely being a membrane protein.
This is why hydropathy is normally plotted as a sliding window rather than reported as one number. A window of 19–21 residues matches the length needed to span a bilayer as an alpha helix, so peaks above roughly +1.6 sustained across that width indicate a transmembrane segment.
| Window size | Detects |
|---|---|
| 7–9 residues | Surface-exposed loops, antigenic sites |
| 19–21 residues | Transmembrane helices |
| Longer | Broad domain-level trends |
What Hydropathy Cannot Predict
Beta-barrel membrane proteins are the major blind spot. Their membrane-spanning strands alternate inward and outward-facing residues, so no sustained hydrophobic stretch appears — outer membrane porins are missed entirely by hydropathy plots and require dedicated beta-barrel predictors.
Signal peptides also produce N-terminal hydrophobic peaks that resemble transmembrane helices, so predictions should be run alongside signal peptide detection. And GRAVY correlates only loosely with expression success — solubility depends on surface charge distribution and folding kinetics as much as average hydropathy.
Worked Examples
Common Mistakes
A single transmembrane helix among hundreds of soluble residues gives a negative average. Use a sliding window plot instead.
Nineteen to twenty-one residues matches a bilayer-spanning helix. Shorter windows detect surface loops instead.
Alternating residues in beta strands produce no sustained hydrophobic stretch. Porins need dedicated beta-barrel prediction methods.
Both give N-terminal hydrophobic peaks. Run signal peptide prediction alongside hydropathy analysis.