What a hydropathy plot shows
Each residue is assigned a hydropathy value from a published scale, and a sliding window averages those values along the sequence. The result is a profile: peaks where a run of hydrophobic residues sits, troughs where the chain is polar or charged.
The reason to plot rather than to average is that the average of a strongly hydrophobic block and a strongly hydrophilic one is close to zero, which describes neither. A peptide like that behaves like a detergent, not like a neutral one.
Choosing a window size
The window is a smoothing parameter. A window of one shows the raw per-residue values, which is noisy. A window of nineteen, the classic setting for finding membrane-spanning helices, smooths so heavily that short features disappear.
For peptides of ten to fifty residues, a window of five to nine is usually right: long enough to suppress single-residue noise, short enough to keep a five-residue hydrophobic patch visible.
The three scales and when each is appropriate
Kyte-Doolittle is the default and the most widely cited. It was derived to identify buried and membrane-spanning regions, so it emphasises the interior-versus-exterior distinction and gives the largest spread between the extremes.
Hopp-Woods is a hydrophilicity scale, built to predict antigenic sites, so its sign convention is inverted relative to the other two and its peaks mark surface-exposed regions. Eisenberg's consensus scale averages several earlier scales and is the usual choice for helical wheel and amphipathicity analysis.
- •Kyte-Doolittle: general purpose, membrane and buried-region prediction.
- •Hopp-Woods: surface exposure and antigenicity. Higher means more hydrophilic.
- •Eisenberg: consensus values, best suited to amphipathicity and helical analysis.
Reading the profile for practical decisions
A profile that stays below zero throughout belongs to a peptide that will dissolve in water. A profile with a sustained peak above about plus two over five or more residues belongs to one that probably will not, and the peak marks the region responsible.
An alternating profile, where hydrophobic and hydrophilic regions swap every three or four residues, indicates amphipathicity: a helix with one face hydrophobic and the other polar. Those peptides tend to be surface-active, prone to aggregating at interfaces, and are common among antimicrobial sequences.
How the hydropathy profile is calculated
A per-residue lookup followed by a centred moving average. The profile is shorter than the sequence by one window, because the window cannot be centred on residues at either end.
score(i) = mean of scale values over positions i-h .. i+h
where h = floor(window / 2)
GRAVY = mean scale value over the whole sequence- Assign a value to each residue. Straight lookup from the selected scale. Non-standard characters are excluded before the lookup so they cannot contribute a zero and drag the average toward neutral.
- Slide a centred window. Each plotted point is the mean of the values in a window centred on that residue. Odd window sizes centre exactly; the tool uses odd sizes for that reason.
- Trim the ends. The first and last few residues have no full window around them and are omitted rather than being averaged over a partial window, which would bias them toward whichever end they are near.
- Identify the extremes. The most hydrophobic and most hydrophilic windows are reported with their residue ranges, so the regions can be read off directly rather than eyeballed from the chart.
- Report GRAVY separately. The whole-sequence mean, unwindowed. It is the number to quote; the profile is the number to reason with.
What this method cannot tell you
- •Hydropathy scales are empirical and disagree with each other. A feature that appears on one scale and not another is a property of the scales, not necessarily of the peptide.
- •The window trims the sequence ends, so a hydrophobic patch at the very N or C terminus is under-represented.
- •It assumes a linear chain. Cyclisation, disulfide constraints and non-standard residues are not modelled.
- •The plot predicts bulk behaviour, not structure. It does not tell you whether a hydrophobic stretch is actually buried in a real conformation.
Where the numbers come from
Hydrophobicity plotter: frequently asked questions
A graph of hydrophobicity along a sequence, produced by assigning each residue a value from a published scale and averaging over a sliding window.
Peaks mark hydrophobic regions and troughs mark hydrophilic ones, so the plot shows where along the chain the character changes.
Because an average destroys the distribution. A peptide with a hydrophobic block at one end and a charged block at the other averages near zero, which describes neither region.
That kind of peptide behaves like a surfactant, and only the plot reveals it.
It depends on the question:
- •Kyte-Doolittle for general purposes and membrane region prediction
- •Hopp-Woods for surface exposure and antigenic sites
- •Eisenberg for amphipathicity and helical wheel analysis
If two scales disagree about a feature, that disagreement is information: the feature is scale-dependent and not robust.
For peptides of ten to fifty residues, five to nine works well. Smaller windows are noisy; larger ones smooth away short features.
The classic window of nineteen was chosen to find membrane-spanning helices in proteins, which are about that long. It is far too wide for a short peptide.
Because a centred window cannot be placed on residues within half a window of either end. Those positions are omitted rather than averaged over a partial window.
With a window of nine, the first and last four residues have no plotted point. Reducing the window recovers them at the cost of a noisier profile.
On Kyte-Doolittle and Eisenberg, hydrophobic. On Hopp-Woods, hydrophilic, because that scale was built with the opposite sign convention.
Checking the direction of the scale before reading a plot avoids the most common misinterpretation.
A profile that stays below zero on Kyte-Doolittle throughout will generally dissolve. A sustained peak above plus two across five or more residues suggests it will not without help.
The solubility predictor combines this with the charged-residue proportion and returns a recommendation rather than a plot.
One with hydrophobic and hydrophilic residues arranged so that a folded structure has a hydrophobic face and a polar face.
In a helix this means an alternation with a period of about 3.6 residues, which shows up in the plot as a regular oscillation. Amphipathic peptides are surface-active and often membrane-disrupting.
No. It shows chemical character along the chain, which constrains structure without determining it.
Dedicated secondary structure prediction uses different methods entirely, and short peptides frequently have no stable structure in water regardless of what any method predicts.
Because they were derived from different experiments: partition coefficients between solvents, buried surface area in known structures, or statistical propensities in protein databases.
They agree on the extremes, that isoleucine is hydrophobic and arginine is not, and disagree in the middle, where residues such as glycine, tyrosine and tryptophan are genuinely ambiguous.
Directly. Reverse-phase columns separate by hydrophobicity, so a more hydrophobic peptide is retained longer and elutes at a higher organic concentration.
This is why oxidation, which makes a peptide more polar, produces an impurity peak eluting earlier than the parent. The HPLC purity interpreter uses that relationship to classify impurity peaks.
Yes, and this is what hydropathy plotting was designed for. Use a window of nineteen for membrane-spanning region prediction in a full-length protein.
Because their sequences were not real. Two entries carried invented single-letter sequences for peptides that contain residues with no single-letter code at all.
A hydropathy plot of a made-up sequence looks exactly as authoritative as a plot of a real one, which is why the entries were removed rather than annotated.
No. A D residue has the same side chain and therefore the same hydropathy value as its L form, so the plot is unchanged.
What changes is the structure the chain can adopt, which the plot does not model in either case.
GRAVY is the mean Kyte-Doolittle value across the whole sequence, unwindowed. It is the single number that summarises the plot.
It is worth quoting and a poor basis for decisions on its own, for the same reason any average is. The plot is what you reason from.
Exposed hydrophobic surface is the main driver of aggregation in water, because burying it away from the solvent is thermodynamically favourable and the easiest way to bury it is against another molecule.
A peptide with a sustained hydrophobic peak is an aggregation risk, particularly at high concentration, near its pI, or after freeze-thaw cycling.
Related Products
Related Research News
Medicare Pays Billions For Obesity’s Consequences. Its GLP-1 Coverage Gap
Medicare spends billions treating obesity-related diseases but does not cover GLP-1 receptor agonists for weight loss alone. This analysis explores the financial and clinical implications, including recent developments in triple peptide agonists like ASC37 from Ascletis.
Why 99% Purity Means Something Different on a Tripeptide and a 44-mer
At 99% per coupling a tripeptide comes off the resin 98% correct and a 44-residue chain comes off at 65%. What had to be removed, why deletion sequences are the hardest impurity to separate, and the trifluoroacetate counterion that is in the vial and not on the paperwork.
Identity Is the Test Almost Nobody Runs
Rearrange TB-500's seven residues and there are 1,260 arrangements that all weigh exactly the same. Purity gets reported constantly and identity almost never, so here is what each method can and cannot establish about whether a vial holds the sequence on the label.