Search effectively
The main search box scans gene symbol, UniProt entry, protein name, aliases, family name, and topology track.
Example searches: EGFR, ITGAV, DKK1, OPRM, cadherin.
Antigen constructs and supporting evidence
First-time user guide
How to use OpenAntigens reports
OpenAntigens assembles sequence, annotation, and structure evidence for mouse orthologs resolved from OpenAntigens human targets. Search for a target, review its proposed construct boundaries, and copy the sequences you choose to test.
How to cite OpenAntigens
If you use OpenAntigens in research, please cite:
Teixeira AAR, Zhu H, Kothiwal D, Cao R, Mills A. OpenAntigens: a structure-aware database for antigen construct design across the human cell-surface and secreted proteome. bioRxiv [preprint]. 2026. https://doi.org/10.64898/2026.07.30.741735
Preprint, version 1; posted 2026-08-04. Not peer reviewed.
Import the reference: BibTeX ยท RIS
Also record the portal build date, software version, report or downloaded artifact URL, and your access date. These identify the release you used; the paper describes the database.
Cite underlying databases or primary studies where you use their evidence. See Terms for data attribution and reuse requirements.
Quick start
- Search for a gene, UniProt entry, protein name, alias, family, topology track on the Browse page.
- Scan the topology, AlphaFold, PDB, construct-count, and BLAST-hit columns to identify reports with the evidence needed for review.
- Open a target report by clicking the UniProt entry name. Public snapshots show the results of a completed build. Pending or failed rows retain their recorded status and error details.
- Use the Interactive Construct Builder to adjust boundaries and inspect linked sequences, warnings, and available structure panels. You can also export sequences from the precomputed construct cards.
- Review the ordered construct sections, copy TSV or FASTA outputs, and download the release index or manifest for downstream analysis.
Filter and sort
The Browse page defaults to all rows sorted by PubTator hits from highest to lowest. Search terms narrow the table; Reset clears both search boxes and restores PubTator-hit sorting.
Click table headers to sort by entry, gene, protein name, family, topology track, PubTator hits, design region, AlphaFold availability, PDB availability, construct count, or BLAST-hit count.
Choose how many rows to show, then use the page controls to browse the remaining results.
Understand topology tracks
Secreted, GPI-anchored, and single-pass proteins usually emphasize soluble antigen constructs from the mature secreted protein or extracellular region. Multipass proteins include a membrane-expression track, and mixed cases with large extracellular regions also include soluble extracellular-region constructs.
For multipass targets, do not interpret the soluble construct list as a complete expression strategy. Review the membrane-expression suggestions and full-length context.
Static release snapshots
Public releases contain the results recorded during their build. The pages can be browsed locally or on the website.
Opening a page does not update its annotations.
Report section map
| Section | What it tells you | How to use it |
|---|---|---|
| Target / Design Region | Resolved UniProt target, organism, sequence length, topology-derived design scope, links, PubTator count, and source identifiers. | Confirm target resolution and compare the extracellular, secreted, or membrane scope with the planned design. |
| Interactive Construct Builder | Linked sequence, cysteine/furin warnings, copy-ready sequence exports, and structure/pLDDT/PAE panels when compatible local structure assets exist. | Adjust boundaries and export mouse construct name, boundaries, sequence, and equivalent regions in the originating human target and available cynomolgus monkey orthologs. |
| Homology | Mouse-to-cynomolgus monkey equivalent regions when reference mappings are available. The originating human target is listed separately as provenance. | Review conservation, boundary transfer, and cross-species reagent risk before animal screening or validation. |
| Sequence similarity and cross-reactivity context | Local BLAST hits ranked by bit score, with identity, coverage, E-value, coordinates, and alignment text. | Inspect high-scoring non-self hits as possible proteins to counter-screen. |
| Cysteines / PTMs / Furin sites | Potential expression liabilities, curated UniProt PTM or processing annotations, and optional mutation guidance. | Review warnings before ordering constructs. Optional Cys-to-Ser and furin-site edits in the builder update names and sequences in real time. |
| Family Context | Canonical family/paralog context and directional identity matrix when available. | Identify close paralogs that may create specificity risks or useful comparison antigens. |
| Interactions / Assembly | Known interaction or complex context, including possible obligatory partners when evidence supports it. | Flag proteins that may require co-expression, partner chains, or extra caution when designing isolated soluble constructs. |
| Construct Summary / Constructs | Precomputed construct classes, boundaries, homolog-equivalent sequences, images, and sequence exports. | Review construct classes in display order, then refine boundaries in the builder when needed. |
Recommended construct-review order
The summary and builder list soluble constructs in this order: full design region, PDB-backed, strict calculated, lenient calculated, then domain annotated constructs. For multipass proteins, review membrane-expression constructs separately from soluble extracellular-region constructs. The Pre-generated Constructs page explains how their boundaries are generated. Display order is not a validated ranking of expression performance.
pLDDT in one sentence
pLDDT tells you whether AlphaFold is locally confident at each residue. High pLDDT supports local fold confidence, but it does not prove that two domains have a fixed relative orientation.
PAE in one sentence
PAE reports AlphaFold's estimated uncertainty in relative placement between residues or regions. Low PAE within a block supports a cohesive structural unit; high PAE between blocks indicates uncertain relative placement.
Common workflows
| Goal | Recommended workflow |
|---|---|
| Find antibody-discovery antigen candidates | Search the gene, filter for structure availability, inspect full design-region and PDB-backed constructs, then review BLAST similarity and homolog-equivalent sequences. |
| Design a compact domain antigen | Open the report, start with domain annotated and strict calculated constructs, inspect pLDDT/PAE boundaries, and export TSV/FASTA from the builder. |
| Prioritize cross-species screening constructs | Check the mouse and available cynomolgus monkey sequences in the Live Selected Region, then review the originating human target separately as provenance. Prioritize mapped constructs with high identity and conserved boundaries. |
| Review paralog cross-reactivity risk | Use Family Context and BLAST similarity together. The family matrix compares paralogs across their sequence regions; BLAST alignments locate similar stretches. |
| Handle a multipass protein | Look for a large extracellular region if present, then separately review membrane-expression constructs, full-length AlphaFold context, full-length BLAST, and GPCR/membrane engineering suggestions. |
| Review release-state gaps | Check the index status and error details. Public pages do not rerun an analysis when opened. |
Example: EGFR
Search EGFR, open EGFR_MOUSE, compare full extracellular-region and domain constructs, inspect PDB-backed boundaries, and check ERBB family context before choosing an antigen for antibody discovery.
Example: ITGAV
Search ITGAV, review interaction and assembly context carefully, and treat isolated integrin-alpha constructs cautiously because integrins often require beta-chain context for native assembly.
Example: DKK1
Search DKK1 to see how secreted proteins are handled. Focus on mature extracellular sequence, cysteine-rich domains, disulfide context, and compact domain constructs.
Example: OPRM
Search OPRM for a multipass example. Prioritize full-length membrane-expression context and treat short loops differently from soluble extracellular-region antigens.
Troubleshooting
| Problem | Likely reason | What to do |
|---|---|---|
| Report is pending or failed | The target could not be completed for the current release because of source-data limitations, structure mismatch, or processing failure. | Inspect the release index error field. The target may be resolved in a future release. |
| AlphaFold is missing | No canonical AlphaFold model was found or the available model did not match the canonical UniProt sequence. | Review available PDB and domain evidence. A new analysis or release is needed to incorporate a later structure. |
| Homolog sequence is missing | A usable reference sequence or alignment was unavailable when the report was built, or the ortholog was withheld because it failed the acceptance gate (too little of either sequence aligned, or identity below the floor for its evidence). The table cell says which. | Use mouse-only until the ortholog table is updated, or inspect source identifiers manually. A withheld ortholog is deliberately not exported. |
| Family matrix is absent | No canonical family was identified, the family exceeded the configured size limit, or precomputed paralog data are unavailable. | Review BLAST local similarity and InterPro domain annotations for specificity context. |
| BLAST hit list is long | The query region is highly conserved, contains common domains, or the full-length track captured broad family similarity. | Inspect the list, which is ranked by bit score, and check which alignments overlap the selected construct. |
| Construct looks biologically wrong | Automated heuristics can miss ligand sites, partner requirements, topology edge cases, or literature-specific constraints. | Use the builder to adjust boundaries and check Methods for the exact evidence sources and limitations. |
Where to go next
Use the Builder guide for detailed instructions on sequence/structure selection, pLDDT, PAE, Cys-to-Ser edits, and FASTA/TSV exports.
Methods
Use the Methods page for source databases, construct heuristics, BLAST setup, family/paralog logic, update cadence, and limitations.
Construct methodology
Use the Constructs page for the dedicated explanation of pre-generated construct classes, structural diagnostics, homolog transfer, PTM and processing review, and limitations.
Downloads
Use the Downloads page for the portal index, manifest, and flat files that support reproducible downstream analysis.
Terms
Use the Terms page for software license, generated annotation license, source-data terms, and contact information.