Ramachandran plot
v1.0.0Analyze protein backbone phi/psi angles and classify favored, allowed, and outlier residues.
Analyze protein backbone phi/psi angles and classify favored, allowed, and outlier residues.
A Ramachandran plot shows the backbone dihedral angles φ and ψ of every residue against reference regions observed in high-quality protein structures. The lower-left region corresponds to right-handed α-helices, the upper left to β-strands and polyproline II, and the upper right to left-handed helices, which are mostly glycine. A well-refined structure has about 98% of its residues in favored regions and very few outliers.
The allowed regions differ by residue type because of side-chain geometry. The tool classifies and plots each residue against the reference set for its own class.
| Class | Allowed region behavior |
|---|---|
| General | The standard distribution shared by most amino acids, with φ predominantly negative |
| Glycine | A single hydrogen side chain gives minimal steric hindrance, so the allowed region is large and nearly symmetric |
| Cis-proline | Proline following a cis peptide bond has its own compact reference distribution |
| Trans-proline | A cyclic side chain locks φ near -60°, confining it to a narrow strip |
| Pre-proline | Residues immediately before a proline shift due to contact with the proline ring |
| Isoleucine or valine | β-branched side chains produce a distinct distribution from the general class |
The class comes from the residue name and its neighbors, checked in a fixed order. Glycine is always glycine, even when a proline follows it. A proline takes the cis or trans class from its ω angle. Any other residue followed by a proline is pre-proline, which takes precedence over isoleucine or valine.
The remaining Ile and Val residues use the isoleucine or valine class, and every other residue name, including HETATM modified residues such as MSE, is scored as general.
The tool uses the six residue classes and favored/allowed thresholds from the cctbx rama8000 validation tables. Classification uses reference masks sampled on a 2° grid. Near classification thresholds, results can differ from cctbx, which interpolates between reference values. The smoothed contours are a visual guide to those masks; points close to an edge should be interpreted using their reported classification. Proline uses the cis-like reference for −90° < ω < 90° and the trans-like reference otherwise, including twisted peptide bonds.
| Region | Population in well-refined structures | Meaning |
|---|---|---|
| Favored | ~98% of residues | Optimal backbone geometry |
| Allowed | Up to ~2% of residues | Sterically possible but less common |
| Outlier | Under 0.2% of residues | Possible modeling error or a real strained conformation |
A residue is favored when the reference density at its φ and ψ is at least 0.02 in every class, a cutoff that encloses 98% of the quality-filtered reference residues.
It is allowed when the density is at least the class's lower cutoff: 0.0005 for the general class (the contour leaves about 1 reference residue in 2,000 outside), 0.001 for glycine, trans-proline, pre-proline and isoleucine or valine, and 0.002 for cis-proline. Below that cutoff it is an outlier (Lovell et al., 2003; Williams et al., 2018). The Ramachandran plot guide explains where the reference distributions come from.
The tool reads PDB and mmCIF files with atomic coordinates. A scored residue needs N, Cα, and C, plus the preceding residue’s C and the following residue’s N. Proline also needs the preceding Cα to determine its reference class. Coincident or collinear atoms that make a dihedral undefined are excluded. Residues are connected using peptide-bond geometry rather than requiring consecutive author residue numbers. Choose Fetch from RCSB in the upload control's three-dot menu to fetch its mmCIF structure directly from the RCSB Protein Data Bank, or upload your own file. For multi-model files such as NMR ensembles, only the first model is analyzed. Load the 1HNR example to try the tool immediately; it contains the deposited backbone coordinates needed for this analysis.
The plot opens on the General residue class, with only that class’s points and reference regions. Each class has its own distribution; glycine, cis-proline, trans-proline, pre-proline, and isoleucine/valine can be selected individually. Favored and allowed areas are shown for one class at a time, and outliers are red crosses.
All residues (overview) shows every scored residue without a reference background or secondary-structure labels. Each point retains its own class-specific classification. There is no single shared favored region for all six residue classes; use the class selector to compare points with their appropriate contours.
The view selector switches the plot to a single residue class and isolates that class's favored and allowed regions. Hover a point for a preview, or click or tap it to pin its name, number, chain, angles, residue class, and region in the residue inspector. Focus the plot and use Left/Right arrow keys to inspect residues, Home/End to jump to the first/last residue, and Escape to clear selection. Display options filter by chain, toggle the reference regions and annotations, hide outliers, and adjust point size.
The summary reports counts and percentages for the residues currently shown on the plot and is named for the selected residue class. Choose All residues (overview) in the residue-class selector to see all classes together; clear chain filters and turn off Outliers only to see whole-structure totals. The overview has no reference areas; choose General or another individual class to restore its reference areas. The whole-structure backbone assessment retains its full denominator. Export Current selection (CSV) for the visible points or All residues (CSV) for all scored residues; the data table tab also retains all scored residues.
The percentages count scored residues only, so chain ends and other unscorable residues never enter the denominator. 100% favored is not the target. The favored contour holds 98% of residues from high-quality structures, so about 2 in 100 good residues fall outside it by construction.
Ubiquitin (1UBQ) has 74 of 74 favored, crambin (3NIR, 0.48 Å) 43 of 44 (97.7%) with one allowed residue, and calmodulin (1CLL) 138 of 142 (97.2%) with no outliers. All three are well-determined crystal structures.
The goals from MolProbity are at least 98% favored and under 0.2% outliers for a well-refined structure (Chen et al., 2010). The reference data itself has about 1 residue in 2,000 (0.05%) outside the allowed contour, so 0.2% is a practical goal rather than a hard limit.
Outliers become more common as resolution worsens and as B-factors rise (Chen et al., 2010). In a short chain, count residues instead of comparing percentages: one outlier among 74 residues is already 1.4%.
With All residues (overview) selected and no chain filter, the panel adds a Backbone assessment label. It reads Excellent at 98% favored or more with under 0.5% outliers, Acceptable at 95% favored or more, and Needs review below that.
The cutoffs are fixed and looser than the MolProbity goals on outliers, and Acceptable does not look at the outlier count. Crambin reads Acceptable with nothing wrong, and 1HNR (71.1% favored, 9 allowed, 4 outliers) reads Needs review. Treat the label as a prompt to look at the residues.
The percentages in a single-class view cover that class only. In 1HNR, the Glycine view reads 0% favored and 100% allowed because all four glycines are allowed, which says little about the rest of the backbone. Use the overview for the whole-structure figure.
Use Outliers only to isolate flagged residues. Expand Inspect outliers and select a row to highlight its point and inspect its angles. All matching outliers are listed, and the data table tab lists every residue with its exact angles for sorting or export. Not every outlier is an error; check the experimental evidence and the residue’s functional role before rebuilding it.
Each row under Inspect outliers gives the residue name, number, chain, and φ and ψ in degrees. Selecting a row opens the residue inspector, which adds the class and region.
The two CSV exports have one row per residue with the columns Chain, Residue, Residue number, Insertion code, Phi (degrees), Psi (degrees), Type and Region. Type is general, glycine, cis-proline, trans-proline, pre-proline or ile-val, and Region is favored, allowed or outlier.
Read the class before the angles, because the same φ and ψ can be favored in one class and an outlier in another. The four outliers in 1HNR are Thr109 (φ 41.5°, ψ −149.5°, general), Pro115 (φ −65.4°, ψ −73.7°, trans-proline), Ile134 (φ −79.9°, ψ 48.7°, isoleucine or valine) and Lys135 (φ 64.0°, ψ −73.8°, general). Thr109 and Lys135 sit at positive φ, where only glycine is common.
Pro115 and Ile134 have ordinary φ values, and it is their ψ that the class reference does not support.
Glycine and proline need separate care. Glycine angles are more permissive and less accurately determined than those of other residues (Lovell et al., 2003), so a glycine at positive φ is routine.
About 5% of prolines are cis, so a cis-proline row is expected (Williams et al., 2018). The tool does not report ω for other residues and does not flag cis non-proline or twisted peptides. Only about 0.03% of non-proline residues are genuinely cis, and twisted peptides should almost always be treated as modeling errors (Williams et al., 2018), so check those in MolProbity.
A note below the plot summary lists residues excluded from scoring, with their identifiers and reasons. Chain ends, missing atoms, chain breaks, undefined dihedrals, and proline residues without a usable ω angle are excluded from the percentage denominator. Coverage counts include residues with at least one parsed backbone atom in the first model; they do not count residues absent from the coordinate file.
Every chain loses its first and last residue, because φ needs the preceding C and ψ needs the following N. Ubiquitin 1UBQ has 76 residues and 74 are scored (Met1 and Gly76 are listed as chain end). A gap in the chain removes the residue on each side (chain break), and a residue missing N, CA or C is skipped (missing atoms).
Skipped residues leave the denominator, so a model with large unbuilt loops reports percentages for the residues that were built.
Calcium ions written as HETATM CA are read as backbone atoms. They appear in the coverage list and add to the warning that some residues could not be scored, as in calmodulin 1CLL, but they do not change the statistics. A residue with alternate locations is scored in one conformer, the one with the most complete backbone and then the highest occupancy, so a strained minor conformer is not shown.
Only the first model of a multi-model file is scored, so an NMR entry reports one conformer. When each of the 16 models of the H-NS DNA-binding domain (1HNS) is run as its own file, favored ranges from 51.1% to 73.3% and outliers from 0 to 8 per model, and no residue is an outlier in more than 4 models. The screen shows one draw from that spread. To judge an ensemble, save individual models as separate files and compare them.
At 2.5 to 4 Å and in cryo-EM models, a clean plot says less. Refinement programs can restrain φ and ψ toward the favored regions, which reduces the plot's value as an independent check (Sobolev et al., 2020), and large distortions in series can mislead the analysis.
MolProbity's CaBLAM is more reliable in that range (Williams et al., 2018). Rama-Z, which tests whether the whole distribution resembles real structures, reads normal below 2, unlikely but possible from 2 to 3 and improbable above 3 in absolute value (Sobolev et al., 2020). This tool does not compute it; MolProbity does.
For predicted models, a clean plot shows that the backbone geometry is physical, not that the fold is right. Read it with pLDDT and the structure prediction confidence guide.
Run MolProbity for clashes, rotamers, Cβ deviations and Rama-Z, and see the MolProbity score guide for those goals. PDB Fixer can add missing atoms and residues so that more residues can be scored, but it does not move existing backbone atoms, so outliers carry through. Correcting a real error means refitting the residue against the experimental map in model-building software.
| Application | Purpose |
|---|---|
| Model validation | Assess crystal structures, NMR ensembles, and homology models |
| Structure refinement | Find residues that need manual adjustment during model building |
| Comparative analysis | Compare conformational states or mutants of the same protein |
| Teaching | Show how backbone geometry relates to secondary structure |
Ramachandran plot generation is free and runs entirely in your browser. No credits required.
Validate protein structures with clashscore, Ramachandran, rotamer, and geometry checks.
Generate structural summary reports and diagrams for a single protein PDB structure.
Model accuracy estimation and sequence ranking with AlphaFold
Score every protein interface with ESM-2 and a graph neural network
Assess docking quality between model and native structures
Score interprotein interactions in AlphaFold and Boltz predictions
Validate generated and docked ligand poses with structure, chemistry, and reference-pose checks.
Plot net charge vs pH to visualize protein charge behavior and identify the isoelectric point.
Generate hydropathy plots to visualize hydrophobic/hydrophilic regions along protein sequences using sliding window analysis.
Explore local protein hydrophobicity with 24 amino acid scales, including Eisenberg and Hopp-Woods, and an adjustable sliding window.
