How to Read a Helical Wheel and a Hydropathy Plot (Amphipathic Helices and Transmembrane Segments)

By Dr. Zubair Khalid, DVM, MS, PhD · 2026-10-02

A hydropathy plot turns a protein sequence into a graph of local hydrophobicity, and a helical wheel turns a short stretch of that sequence into a view down the axis of an alpha helix. Together they answer two questions that come up constantly in sequence analysis: does this protein contain a membrane-spanning segment, and does this helix have one hydrophobic face and one hydrophilic face?

You will meet these tools when you annotate a new sequence, when you check whether a predicted transmembrane domain makes sense, when you study antimicrobial or cell-penetrating peptides, and when you read a paper that claims a helix is amphipathic. Both are simple enough to compute by hand and easy to misread, so the goal here is to make the numbers mean what they actually mean.

Quick Answer

A hydropathy plot slides a window of fixed length along the sequence, averages the hydrophobicity of the residues inside the window, and plots that average at the middle residue [1].
The classic Kyte-Doolittle scale runs from +4.5 (isoleucine, most hydrophobic) to -4.5 (arginine, most hydrophilic) [1].
For membrane-spanning segments, use a 19-residue window; an average above +1.6 indicates a high probability of a transmembrane sequence [1].
A helical wheel places residue n at an angle of (n - 1) x 100 degrees, because an alpha helix has 3.6 residues per turn [4].
An amphipathic helix has hydrophilic side chains on one face and hydrophobic side chains on the opposite face; the hydrophobic moment measures how strongly that separation occurs [2].
High hydrophobicity with a low hydrophobic moment suggests a transmembrane anchor; high moment with lower hydrophobicity suggests a surface-seeking or membrane-active helix [3].

What a Hydropathy Plot Actually Shows

Kyte and Doolittle built their scale in 1982 by combining experimental observations about the 20 side chains into a single number per residue [1]. Isoleucine sits at the top with +4.5, valine at 4.2, leucine at 3.8, phenylalanine at 2.8, cysteine at 2.5 and alanine at 1.8. Glycine is slightly negative at -0.4, tryptophan at -0.9, tyrosine at -1.3, lysine at -3.9 and arginine at the bottom with -4.5 [1].

The plot itself is a moving average. Pick a window length, sum the hydropathy values of the residues in that window, divide by the window length, and plot the result at the middle residue. With a 7-residue window, the first plotted value is the average of residues 1 through 7, placed at position 4. That is why odd window lengths are used: there is always a true middle residue to plot against [1].

The plot includes a horizontal line at the grand average hydropathy of amino acid compositions across most sequenced proteins. Regions above that line are hydrophobic, regions below it are hydrophilic [1]. For soluble globular proteins, the hydrophobic regions correspond well to interior segments and the hydrophilic regions to exterior segments, as confirmed against crystal structures [1].

Window length matters more than most people expect. Windows shorter than 7 residues gave noisy, unsatisfactory profiles, and long windows missed small features. Information content for distinguishing globular protein interiors from exteriors was greatest at 7 to 11 residues. For membrane-spanning segments, discrimination from the most hydrophobic parts of soluble proteins was clearest with a 19-residue window [1]. A 19-residue helix spans 19 x 1.5 = 28.5 Angstrom along its axis [4].

How to Read a Hydropathy Plot

Start by checking the window length and the scale. A peak means something different at window 7 than at window 19, and averages are sensitive to both choices. Report both with any plot you publish.

For transmembrane prediction, look for a sustained run of windows above +1.6 using the 19-residue window [1]. A single window above threshold is weak evidence; eight consecutive windows above threshold covering roughly 26 residues is strong evidence. Membrane-spanning sequences appear as large uninterrupted areas on the hydrophobic side of the midpoint line, and they are more hydrophobic than segments passing through the interior of soluble proteins [1].

Then check the N-terminus. Signal peptides are hydrophobic and will produce a peak that looks like a transmembrane segment. A hydropathy plot alone cannot tell them apart, and this is a common source of false positives in manual annotation.

Finally, confirm with a dedicated topology predictor. TMHMM, a hidden Markov model for membrane protein topology, correctly predicted 97 to 98% of transmembrane helices and distinguished soluble from membrane proteins with specificity and sensitivity better than 99%, though accuracy drops when signal peptides are present [5]. TMHMM-based estimates suggest that 20 to 30% of genes in most genomes encode membrane proteins [5]. DeepTMHMM, a deep learning method based on protein language models, predicts topology for both alpha-helical and beta-barrel transmembrane proteins [6].

What a Helical Wheel Shows

An alpha helix has 3.6 residues per turn. Each residue is related to the next by a rise of 1.5 Angstrom along the axis and a rotation of 100 degrees, giving a pitch of 5.4 Angstrom per turn [4]. A helical wheel projects that helix down its axis, placing residue n at an angle of:

$$\theta_n = (n - 1) \times 100^\circ \pmod{360^\circ}$$

where n is the residue number starting at 1 and the result is taken modulo 360 to keep it within one turn. Because 100 degrees is exactly 360/3.6, residues i, i+3, i+4 and i+7 land on the same face of the wheel. That clustering is the whole point: it lets you see at a glance whether one side of the helix is greasy and the other is charged.

An amphiphilic (also written amphipathic) helix has one surface projecting mainly hydrophilic side chains and the opposite surface projecting mainly hydrophobic side chains. Most helices of myoglobin and hemoglobin are amphiphilic [2], which is a useful reminder that this geometry is not exclusive to membrane proteins.

The Hydrophobic Moment

Eisenberg, Weiss and Terwilliger defined the mean helical hydrophobic moment in 1982 as the mean vector sum of the side-chain hydrophobicities of an N-residue helix. Each residue contributes a vector whose length is its hydrophobicity and whose direction is set by its angle around the helix axis [2]. In formula form:

$$\mu_H = \left| \sum_{n=1}^{N} H_n \left( \cos(n\delta), \sin(n\delta) \right) \right|$$

where H_n is the hydrophobicity of residue n on whatever scale you chose, delta is the rotation per residue (100 degrees for an alpha helix), and the vertical bars denote the magnitude of the resulting vector. The mean moment is mu_H divided by N. A large mean moment means the helix is amphiphilic perpendicular to its axis; a small one means the hydrophobic residues are spread around the wheel instead of segregated [2].

The moment depends on the hydrophobicity scale. Values computed with different scales cannot be compared directly. Magainin 2 residues 1 to 18 give a mean moment of 1.535 with Kyte-Doolittle values and 0.506 with the Eisenberg consensus values. Same peptide, same geometry, different numbers.

Plotting mean hydrophobic moment against mean hydrophobicity separates helix types into different regions [2]. Eisenberg and colleagues found in 1984 that monomeric transmembrane anchors lie at highest hydrophobicity and smallest hydrophobic moment, paired or bundled membrane helices have higher moment and lower hydrophobicity, surface-seeking helices such as melittin have still higher moment, and globular-protein helices have lower hydrophobicity and moment [3]. That two-axis view is more informative than either number alone.

Worked Example

Take human glycophorin A (UniProt P02724), a 150-residue precursor with a signal peptide at 1 to 19, an extracellular domain at 20 to 91, a single helical transmembrane segment at 92 to 114 and a cytoplasmic domain at 115 to 150 [8]. Kyte and Doolittle used erythrocyte glycophorin as a model membrane-spanning protein with an easily recognized membrane-spanning segment [1].

Run a Kyte-Doolittle profile with a 19-residue window. There are 132 windows. The maximum is 2.668 for the window covering residues 92 to 110 (ITLIIFGVMAGVIGTILLI), plotted at its center, residue 101. The next highest values are 2.389 (93-111), 2.358 (94-112), 2.247 (91-109), and 2.137 for both 96-114 and 95-113. Windows above +1.6 start at residues 90 through 97, with scores 1.963, 2.247, 2.668, 2.389, 2.358, 2.137, 2.137 and 1.663. Together they cover residues 90 to 115, which matches the UniProt transmembrane annotation of 92 to 114.

Now the false positive. The signal peptide at 1 to 19 also exceeds the threshold: windows starting at residues 1 through 5 score 1.726, 1.826, 1.853, 1.837 and 2.005. The next N-terminal window (start 6) scores 1.584, just under the threshold. A hydropathy plot alone would call a transmembrane segment at the N-terminus that is not one.

Everything else behaves. Windows starting at residue 20 or later that do not overlap the transmembrane region (starts 20-73 and 115-132) never exceed -0.142, and the profile minimum is -1.953 for the window starting at residue 43. GRAVY, the grand average of hydropathy, is -0.080 for the precursor and -0.342 for the mature chain (residues 20 to 150). A single transmembrane helix does not make a whole protein's GRAVY positive. You can compute GRAVY for your own sequence with the Protein Properties Calculator.

For the helical wheel, use the first 18 residues of magainin 2 (GIGKFLHSAKKFGKAFVG, from UniProt P11006) [9]. Applying the angle formula gives G1 at 0 degrees, I2 at 100, G3 at 200, K4 at 300, F5 at 40, L6 at 140, H7 at 240, S8 at 340, A9 at 80, K10 at 180, K11 at 280, F12 at 20, G13 at 120, K14 at 220, A15 at 320, F16 at 60, V17 at 160 and G18 at 260. Ordered around the wheel: G1 0, F12 20, F5 40, F16 60, A9 80, I2 100, G13 120, L6 140, V17 160, K10 180, G3 200, K14 220, H7 240, G18 260, K11 280, K4 300, A15 320, S8 340.

With delta = 100 degrees on the Kyte-Doolittle scale, the total moment vector has magnitude 27.63, giving a mean moment of 1.535 and a mean hydrophobicity of 0.183, with the moment pointing at 78 degrees. On the Eisenberg consensus scale the mean moment is 0.506, mean hydrophobicity 0.204, direction 74 degrees. Residues within 60 degrees of the moment direction are I2, F5, A9, F12, G13 and F16, which form the hydrophobic face. The four lysines sit at 180 to 300 degrees, on the opposite face.

Two controls make the interpretation concrete. Recomputing with delta = 180 degrees, which is beta strand geometry instead of helix, drops the mean moment to 0.072 on the Kyte-Doolittle scale. The amphipathicity is specific to helical periodicity. And for the glycophorin A transmembrane segment (residues 92 to 109, ITLIIFGVMAGVIGTILL), mean Kyte-Doolittle hydropathy is 2.567 but mean moment is only 0.558 on Kyte-Doolittle or 0.088 on Eisenberg. High hydrophobicity with low moment is the signature of a monomeric transmembrane anchor [3].

Common Mistakes

Using the wrong window length. A 7-residue window is fine for globular protein interiors but too short for transmembrane calls. Use 19 residues for membrane-spanning prediction and state the window in any figure legend [1].
Treating a single peak as proof. One window above +1.6 is weak. Look for a sustained run of consecutive windows above threshold covering roughly 20 or more residues [1].
Ignoring the signal peptide. The glycophorin A example shows an N-terminal signal peptide scoring above the transmembrane threshold. Check the N-terminus against a signal peptide predictor before calling a transmembrane segment there.
Comparing hydrophobic moments across scales. A mean moment of 1.535 on Kyte-Doolittle and 0.506 on Eisenberg describe the same helix. Always name the scale.
Reading GRAVY as a membrane predictor. GRAVY is the sum of hydropathy values divided by the number of residues; positive values indicate an overall hydrophobic sequence [7]. Glycophorin A is -0.080 despite having a clean transmembrane helix. GRAVY describes the whole chain, not any local segment.
Assuming a high moment means transmembrane. Surface-seeking helices have the highest moments, not the lowest. Transmembrane anchors sit at high hydrophobicity and low moment [3].

Limitations

The +1.6 threshold and the 19-residue window come from a small 1982 training set of nine known membrane-spanning segments. Modern predictors such as TMHMM and DeepTMHMM are more accurate and handle signal peptides better [5][6]. Treat a manual hydropathy plot as a first pass, not a final annotation.

The hydrophobic moment formula is written here from the verbal definition in the 1982 abstract. Some tools normalize differently, using a per-residue mean or an 11-residue window as in Eisenberg's original moment plots. Check the documentation of whatever tool you use before comparing your numbers to published values.

The Eisenberg consensus scale values used in the worked example come from Biopython's ProtParamData table, which attributes them to Eisenberg et al. 1984. The 1984 paper's own table was not consulted directly, so verify against the original if the exact values matter for your analysis.

Biopython's protein_scale with edge = 1.0 gives a plain average. Other tools weight window edges or use different scales, so peak values can differ slightly from the numbers here. Report your tool, scale and window length.

DeepTMHMM accuracy figures are not quoted here because only the 2022 preprint abstract was available. Check the current documentation and any peer-reviewed version for performance claims.

Frequently Asked Questions

How do I read a hydropathy plot for the first time?

Check the window length and scale in the axis labels, then look for sustained regions above the midpoint line. For transmembrane prediction, use a 19-residue window and a +1.6 threshold [1]. A single peak is not enough; look for a run of consecutive windows above threshold.

What is the difference between a helical wheel and a hydropathy plot?

A hydropathy plot is one-dimensional: it shows local hydrophobicity along the sequence. A helical wheel is a two-dimensional projection that shows how residues are arranged around the axis of a helix. The plot tells you where hydrophobic stretches are; the wheel tells you whether those stretches form one face of a helix.

What does a high hydrophobic moment mean?

A high mean hydrophobic moment means the helix has a strongly segregated hydrophobic face and hydrophilic face, perpendicular to its axis [2]. Surface-seeking helices such as melittin have the highest moments. Transmembrane anchors typically have high hydrophobicity but low moment [3].

Can I use Kyte-Doolittle values for anything besides transmembrane prediction?

Yes. The scale is used for GRAVY scores [7] and, with 7- to 11-residue windows, for predicting interior and exterior segments of globular proteins [1]. Just remember that the +1.6 threshold applies specifically to a 19-residue window for transmembrane prediction, not to other uses.

Why does my transmembrane helix prediction disagree with TMHMM?

Manual hydropathy plots cannot distinguish signal peptides from transmembrane segments, and they do not model topology. TMHMM handles both, with 97 to 98% accuracy on transmembrane helices and better than 99% specificity and sensitivity for soluble versus membrane protein classification, though accuracy drops when signal peptides are present [5]. When the two disagree, trust the dedicated predictor and check whether a signal peptide explains the discrepancy.