Target
Advanced options
Use your own target structure or add a binder complex.
Custom target structure
no file chosen
Already a full target–binder complex? Choose its target chain and tick its binder chains here.
Binder complex
Annotate the full receptor above, then add a binder to the existing structure. Paste or upload the binder + target fragment below; the shared target residues align the binder onto the full receptor.
Annotate a topology example to see a matching binder.
no file chosen
Load the full target, then paste or upload a complex to align its binder.
Add a binder to inspect contacts in the current view.
Interaction lines:
Downloads
Session
Session files include your structure and annotations. Opening a session recomputes its assessment.
Browser settings
Public UniProt, AlphaFold and structure downloads are cached locally for 7 days (64 MB maximum), with retrieval time and content hash. Uploaded structures and email addresses are not stored in this cache.
How it works
Learn how topology, membrane placement and result checks work.
Membrane protein topology: the types
The diagrams distinguish α-helical spans from transmembrane β-strands.
For helical proteins, topology is described by how many
times the chain crosses, which side the N terminus is on, and whether it was made with a
cleavable signal peptide. The classical single-pass types follow from them. The type of the
protein on screen is outlined in red. Each card links to a human example in UniProt.
“Out” includes the lumen of the ER or a vesicle; “in” means cytoplasmic.
Excluded from automatic placement and correction:
mixed α/β transmembrane annotations, such as Perforin-2 (MPEG1). They remain available for
annotation only; placement needs a topology and structure for the same membrane-inserted state.
| Type | Spans | N terminus | Signal peptide | C terminus | How this page decides |
|---|---|---|---|---|---|
| I | 1 | out | yes, cleaved | in | UniProt "Single-pass type I"; otherwise N out + curated SIGNAL |
| II | 1 | in | no (signal anchor) | out | UniProt "type II"; otherwise N in and the span not at the C terminus |
| III | 1 | out | no (reverse signal anchor) | in | UniProt "type III"; otherwise N out and no SIGNAL |
| IV | 1, at the C terminus | in | no | out (a few residues) | UniProt "type IV"; otherwise the span ends within 50 residues of the C terminus, N in |
| Multi-pass (helical) | ≥ 2 | in or out | sometimes | parity of the span count | number of curated TRANSMEM segments; N side from the flanking topological domain |
| β-barrel / pore | β-strands, counted per chain | sides from curated topology; the barrel may need several subunits | UniProt TRANSMEM features described as "Beta stranded" | ||
| GPI / lipid / peripheral | 0 | anchored without crossing | UniProt keyword, LIPID feature or subcellular-location topology | ||
What the page reports for every target (the line under the
protein's name): the type; which side the N and C termini are on (extracellular / out or
cytoplasmic / in); the number of spans and their positions; the signal peptide and its range;
re-entrant segments. Every one of these comes from curated UniProt data. UniProt's own
statement is reported first; where the page's own derivation disagrees, both are shown. None of it is
read from the structure. The signal peptide is drawn in green in the sequence, the domain bar
and the 3D view: it is cleaved during biogenesis and is not part of the mature protein.
The workflow, step by step
- Input. A UniProt accession loads the AlphaFold model; a PDB ID loads the experimental mmCIF from RCSB and its UniProt entry via SIFTS; an uploaded or pasted structure uses the accession box, if filled, as its topology source.
- Class and route. Curated UniProt data decides what kind of membrane protein this is (single-pass type I–IV, multi-pass, GPI, lipid-anchored, peripheral, soluble) and which method handles it. A protein UniProt says is not a membrane protein gets no membrane.
- Topology. The curated spans, topological domains and signal peptide are mapped onto the structure's residues by canonical sequence position.
- Membrane frame. Our own deterministic solver finds the bilayer normal, centre and thickness: from the curated spans where there are any, otherwise from the hydrophobic surface alone. Without a curated span, a slab is kept only if the residues it exposes to lipid are hydrophobic and uncharged; otherwise no membrane is drawn.
- Membrane type, thickness and tilt. The bilayer's hydrophobic thickness depends on the compartment. The page takes the membrane from UniProt's first subcellular location, or from the membrane menu in Colour & display, and uses the OPM average thickness for that membrane (Pogozheva et al. 2013): plasma membrane 33.5 Å, ER 30.2 Å, mitochondrial inner 28.6 Å, bacterial inner 30.2 Å, Gram-positive plasma 31.6 Å. Golgi (32.2 Å) is the ER value plus the ER→Golgi difference measured by Mitra et al. (2004), because OPM has no Golgi average. For a single-spanning protein, a helix of L residues is 1.5·L Å long. If that is longer than the bilayer, the helix is tilted until it fits: cos θ = D / 1.5·L. A 27-residue span tilts 42° in the ER and 34° in the plasma membrane; a 20-residue span stays upright. The direction of the tilt is whichever keeps the soluble domains out of the core. The thickness and tilt are written beside the membrane in the 3D view. Choosing fitted to the structure switches this off.
- Placed in membrane. One rigid rotation and translation of every atom puts the membrane horizontal with the outside up. Nothing is refolded.
- Corrected. For a single-pass protein whose model puts domains on the wrong side or inside the bilayer: trim the signal peptide and low-confidence termini; turn the connector torsions (articulated motion); if that is not enough, place the domain as a rigid body and close the connector (CCD); put the termini back and swing them clear of the core; relieve low-confidence regions still in the core. Bond lengths and angles are preserved throughout.
- Checks. Every model is checked for domains on the wrong side, atoms in the core, contacts across the membrane, spans that do not cross and more, before and after correction.
- Outputs. The placed and corrected structures, any selected part (a domain, the signal peptide, a span, a residue range), the annotation record (JSON), a report and the segments (TSV).
What this output does not say
Additional tools
Explore molecular context, uncertainty and external reference tools.
Method and provenance
Correction and assessment
Run an annotation to view correction and assessment details.
Topology evidence
Run an annotation to view topology evidence and its sources.
Topology sources
A topology call is stored with its source, its evidence class and a version, never as a bare interval. The confidence column is our own application-level score from topology.confidence_by_class, not a probability supplied by UniProt.
Channels
Membrane frame, sign and placement
Run an annotation to view the membrane frame, sign and placement.
Oriented-file comparison and adoption are available under Independent reference & external tools.
Biological sign
Placement
Surface patches and clearance
Run an annotation to view surface patches and clearance.
References
Each reference states how it is used: browser software, imported data, an implemented method, an optional integration, or context only. These labels describe the webpage's capabilities; the inputs and provenance determine what was used for a particular result.
Used here: software and data
These are integrated into the webpage; the selected input and display determine which resources are used.
- 3Dmol.js. Browser library Rego and Koes (2015). 3Dmol.js: molecular visualization with WebGL. Bioinformatics. — Version 2.5.5 runs in the browser to draw structures, surfaces and selections.
- UniProt. Data import The UniProt Consortium (2025). UniProt: the Universal Protein Knowledgebase in 2025. Nucleic Acids Research. — Sequences and curated topology, signal-peptide and domain annotations for accession-based inputs.
- AlphaFold. Precomputed predictions Jumper et al. (2021). Highly accurate protein structure prediction with AlphaFold. Nature. — Underlying method for imported predictions and pLDDT/PAE confidence. AlphaFold inference is not run here.
- AlphaFold DB. Data import Varadi et al. (2024). AlphaFold Protein Structure Database in 2024: providing structure coverage for over 214 million protein sequences. Nucleic Acids Research. — Downloads existing structures and PAE files when an AlphaFold input is selected.
- Protein Data Bank. Data import Berman et al. (2000). The Protein Data Bank. Nucleic Acids Research. — Experimental inputs and bundled example complexes originate from RCSB PDB. Also cite the original paper for each entry used.
- SIFTS. Mapping import Dana et al. (2019). SIFTS: updated Structure Integration with Function, Taxonomy and Sequences resource allows 40-fold increase in coverage of structure-based annotations for proteins. Nucleic Acids Research. — PDBe mapping data connect experimental-structure residue numbering to UniProt.
- Chemical Component Dictionary. Bundled display data Westbrook et al. (2015). The chemical component dictionary: complete descriptions of constituent molecules in experimentally determined 3D macromolecules in the Protein Data Bank. Bioinformatics. — Ideal component coordinates supply illustrative PTM fragments, rather than measured atoms in the input structure.
Used here: implemented methods
These references credit methods or parameter values implemented or adapted in the app. They are not all separate software packages.
- Solvent accessibility. Implemented method Shrake and Rupley (1973). Environment and exposure to solvent of protein atoms. Lysozyme and insulin. Journal of Molecular Biology. — Basis of the app's own browser calculation of solvent-accessible surface area.
- Relative accessibility. Reference values Tien et al. (2013). Maximum Allowed Solvent Accessibilites of Residues in Proteins. PLOS ONE. — Reference maxima used to convert residue surface areas into relative accessibility.
- Covalent radii. Reference values Cordero et al. (2008). Covalent radii revisited. Dalton Transactions. — Element-specific radii used to infer candidate covalent bonds in the app's geometry checks.
- Hydrophobicity. Reference values Kyte and Doolittle (1982). A simple method for displaying the hydropathic character of a protein. Journal of Molecular Biology. — Hydropathy scale used for colouring and membrane-surface assessment.
- Membrane thickness. Reference values Pogozheva et al. (2013). Structural adaptations of proteins to different biological membranes. Biochimica et Biophysica Acta — Biomembranes. — Table 1 supplies membrane-specific hydrophobic-thickness defaults.
- Golgi thickness estimate. Reference values Mitra et al. (2004). Modulation of the bilayer thickness of exocytic pathway membranes by membrane proteins rather than cholesterol. PNAS. — The measured ER-to-Golgi difference supplies an offset to the app's ER-based estimate.
- PAE graph construction. Adapted method Tristan Croll. pae_to_domains. — Source of the PAE-graph approach and defaults. The app implements its own deterministic clustering; it does not load this Python package.
- PAE domain clustering. Implemented method Clauset, Newman and Moore (2004). Finding community structure in very large networks. Physical Review E. — Greedy modularity is adapted to group residues using the input PAE matrix.
- Clustering resolution. Implemented method Reichardt and Bornholdt (2006). Statistical mechanics of community detection. Physical Review E. — Resolution term used in the app's modularity calculation.
- Domain cleanup. Adapted method / optional local Lau, Kandathil and Jones (2023). Merizo: a rapid and accurate protein domain segmentation method using invariant point attention. Nature Communications. — Browser PAE segmentation adapts Merizo's short-segment cleanup. The full Merizo network can run during a requested local cross-check if installed.
- Structure alignment. Implemented method Horn (1987). Closed-form solution of absolute orientation using unit quaternions. Journal of the Optical Society of America A. — Quaternion least-squares fitting aligns supplied complexes and reference structures.
- Connector closure. Implemented method Canutescu and Dunbrack (2003). Cyclic coordinate descent: A robotics algorithm for protein loop closure. Protein Science. — Published basis for the CCD connector-closure step in the app's correction workflow.
- Analytical connector closure. Implemented method Coutsias et al. (2004). A kinematic view of loop closure. Journal of Computational Chemistry. — Basis of the app's own kinematic-closure implementation, used after CCD when needed.
- Polynomial loop closure. Method basis Wedemeyer and Scheraga (1999). Exact analytical loop closure in proteins using polynomial equations. Journal of Computational Chemistry. — Earlier polynomial formulation underlying the analytical loop-closure method.
Membrane placement, correction and structural-test thresholds are InXBio implementations described in Method and provenance. The cited methods provide their basis; they do not independently validate the resulting coordinates.
Optional tools and reference imports
Used when the corresponding prediction, example, file import or local cross-check is requested and available.
- InterPro. Optional data import Blum et al. (2025). InterPro: the protein sequence classification resource in 2025. Nucleic Acids Research. — Supplies stored Phobius, TMbed and SignalP results when requested in TMD predictions.
- Phobius. Optional stored / remote Käll, Krogh and Sonnhammer (2004). A combined transmembrane topology and signal peptide prediction method. Journal of Molecular Biology. — Uses stored InterPro output where available; a requested live job runs at EBI.
- TMHMM. Optional browser / local Krogh et al. (2001). Predicting transmembrane protein topology with a hidden Markov model: application to complete genomes. Journal of Molecular Biology. — The browser parser and Viterbi implementation are ported from tmhmm.py, using a user-supplied TMHMM 2.0 model. A local-service backend is also available.
- TMbed. Optional stored predictions Bernhofer and Rost (2022). TMbed: transmembrane proteins predicted through language model embeddings. BMC Bioinformatics. — Imports precomputed transmembrane predictions from InterPro; no TMbed model runs in the browser.
- SignalP 6.0. Optional stored predictions Teufel et al. (2022). SignalP 6.0 predicts all five types of signal peptides using protein language models. Nature Biotechnology. — Imports precomputed signal-peptide predictions from InterPro alongside TMbed.
- DeepTMHMM. Optional remote job Hallgren et al. (2022). DeepTMHMM predicts alpha and beta transmembrane proteins using deep neural networks. bioRxiv preprint. — Requested through the local service and executed remotely by BioLib; the model is not bundled.
- Membranome 3.0. Optional data import Lomize et al. (2022). Membranome 3.0: Database of single-pass membrane proteins with AlphaFold models. Protein Science. — Queries matching single-pass database entries for TMD boundaries and orientation; no remodeling job is run.
- FreeSASA. Optional local software Mitternacht (2016). FreeSASA: An open source C library for solvent accessible surface area calculations. F1000Research. — Used by the requested local accessibility cross-check (Lee–Richards algorithm). The browser uses its own Shrake–Rupley implementation.
- UniTmp / TOPDB. Optional local data import Dobson et al. (2024). UniTmp: unified resources for transmembrane proteins. Nucleic Acids Research. — The requested local cross-check can retrieve TOPDB topology evidence for accession inputs. The browser annotation itself does not query this channel.
- Human Bindome. Optional example data Wenckstern, Diaz-Rovira et al. (2026). The Human Bindome: A Proteome-Scale Atlas of Designed Binder Candidates. bioRxiv preprint. — Source of the bundled CYB5A and CD59 computational binder examples. No binder-design model runs here; these examples are not experimentally validated.
- OPM. Optional file import Lomize et al. (2012). OPM database and PPM web server: resources for positioning of proteins in membranes. Nucleic Acids Research. — Compatible planar oriented coordinates supplied by the user can be compared or adopted as the membrane reference frame.
- PPM 3.0. Optional file import Lomize, Todd and Pogozheva (2022). Spatial arrangement of proteins in planar and curved membranes by PPM 3.0. Protein Science. — Imports a compatible planar PPM-oriented result for comparison or as the reference frame. PPM itself runs externally.
Further context only
Related resources for comparison, external analysis or downstream work. They do not generate the webpage's browser result.
- AAontology. Context only Breimann, Kamp, Steiner and Frishman (2024). AAontology: An Ontology of Amino Acid Scales for Interpretable Machine Learning. Journal of Molecular Biology. — Context for comparing and selecting physicochemical amino-acid scales.
- AlphaPPM. Context only BioMembHub, AlphaPPM server (accessed 11 October 2026). — Related public workflow offering single-pass structure remodeling; see also the Membranome 3.0 publication above.
- PDBTM. Context only Kozma, Simon and Tusnády (2013). PDBTM: Protein Data Bank of transmembrane proteins after 8 years. Nucleic Acids Research. — Structural membrane annotations and orientations for related-work comparison; the optional local TOPDB channel is listed separately.
- TmDet 4.0. Context only Tusnády and Gerdán (2025). TmDet 4.0: determining membrane orientation of transmembrane proteins from 3D structure. Nucleic Acids Research. — External orientation and fragment-analysis tool. This webpage does not execute TmDet.
- MemPrO. Context only Parrag and Stansfeld (2026). MemPrO: A Predictive Tool for Membrane Protein Orientation. Journal of Chemical Theory and Computation. — External orientation and simulation-preparation workflow. This webpage does not execute MemPrO.
- MembraneFold. Context only Gutierrez et al. (2022). MembraneFold: Visualising transmembrane protein structure and topology. bioRxiv preprint. — Combines predicted structures with topology in an interactive membrane-protein viewer.
- TmAlphaFold. Context only Dobson et al. (2023). TmAlphaFold database: membrane localization and evaluation of AlphaFold2 predicted alpha-helical transmembrane protein structures. Nucleic Acids Research. — Database of membrane-localized AlphaFold models and their membrane-compatibility assessments.
- TMVisDB. Context only Olenyi et al. (2025). TMVisDB: Annotation and 3D-visualization of Transmembrane Proteins. Journal of Molecular Biology. — Combines AlphaFold structures with TMbed topology annotations for browsing and comparison.
- iCn3D. Context only Wang et al. (2020). iCn3D, a web-based 3D viewer for sharing 1D/2D/3D representations of biomolecular structures. Bioinformatics. — General structure analysis with linked sequence, annotation and interaction views.
- Mol*. Context only Sehnal et al. (2021). Mol* Viewer: modern web app for 3D visualization and analysis of large biomolecular structures. Nucleic Acids Research. — General molecular visualization and analysis, complementary to a focused membrane workflow.
- CHARMM-GUI Membrane Builder. Context only Wu et al. (2014). CHARMM-GUI Membrane Builder toward realistic biological membrane simulations. Journal of Computational Chemistry. — Builds explicit protein–lipid systems for molecular dynamics after structural inspection.
- MemProtMD. Context only Newport, Sansom and Stansfeld (2019). The MemProtMD database: a resource for membrane-embedded protein structures and their lipid interactions. Nucleic Acids Research. — Provides structures embedded in simulated lipid bilayers and their lipid-interaction context.