PDB vs UniProt: Map Mutations Correctly

You select a residue in a structure viewer, order a mutation, and later discover that the plasmid position refers to a different amino acid. The intended site was clear. The numbering system was not.

A mutation needs a sequence reference as well as a residue number. PDB author numbering, mmCIF sequence numbering, UniProt positions, and construct positions can all describe the same site differently. Tags, signal peptides, insertion codes, isoforms, engineered deletions, and missing coordinates make a simple offset unreliable.

Before designing primers or a synthesis order, build an explicit mapping from the structural site to the exact sequence you will express. Then verify the wild-type residue in that sequence.

Key takeaways

• A residue number needs a reference sequence, chain namespace, and any insertion code.
• Map through the deposited polymer sequence and the actual construct; missing coordinates are not necessarily a deletion.
• Verify the starting residue, local sequence, and reverse mapping before handing off a candidate.

Four coordinate systems that often get mixed together

The RCSB guide to primary sequences explains why sequence records and visible coordinates need to be considered together. A coordinate file is not always a simple numbered list of the full biological protein.

Figure 1. A biological sequence, a deposited structure and a laboratory construct can assign different identifiers to the same site. The alignment shown is conceptual, not a real residue mapping.

Start with the biological reference and the plasmid sequence

Retrieve the intended UniProt accession and explicitly identify the isoform. Record the sequence or version you used. A gene name alone is insufficient when organisms, paralogues, or isoforms are easy to confuse.

Next obtain the actual expression sequence, ideally verified against the current plasmid or synthesis record. Include every translated residue: affinity tags, cleavage sites, linkers, fusion partners, and any previously introduced mutations. Keep a second sequence representing the mature purified protein if cleavage removes part of the expression product.

These two construct sequences answer different questions. Primer design needs the DNA template and coding sequence. Structural interpretation may concern the protein remaining after processing. Label them separately so a correct mutation in one reference is not accidentally ordered against another.

For broader sequence assembly checks, see our construct-boundary design workflow.

Understand author numbering and insertion codes

In mmCIF, author and standardized identifiers can coexist. Relevant fields include author chain and residue identifiers, the polymer sequence index, and the insertion code. Viewers may expose different combinations or translate them into their own labels.

A residue labeled 100A is not necessarily residue 101. The letter may be an insertion code attached to author residue number 100. Similarly, author numbering can include zero or negative values, and two chains can contain residues with the same number.

Record a structural site with enough information to find it again: PDB entry, model when relevant, chain identifier and its namespace, author residue number, insertion code, and amino-acid identity. For automated workflows, preserve both author and standardized identifiers instead of flattening them into one ambiguous integer.

Do not remove insertion codes during spreadsheet cleanup. That can turn distinct residues into duplicates and silently direct several annotations to one site.

Read the identifier fields rather than trusting the viewer label

The wwPDB dictionary defines _atom_site.label_seq_id as a reference to the polymer sequence position. _atom_site.auth_seq_id preserves an author-assigned identifier and need not equal that position. The distinction is part of the file format, not an inconsistency for a spreadsheet to repair.

A structural residue is identified by a combination of fields. In addition to its sequence identifier, retain the chain namespace and any insertion code. A label such as “A:100” is ambiguous if one tool means the author chain and another means the standardized chain. Preserve auth_asym_id and label_asym_id when moving information between tools, and note which one the receiving tool expects. The PDBx/mmCIF user guide provides the format context for these fields.

Consider a hypothetical exported list containing author residues 100, 100A, and 101. Removing the insertion code would turn the first two into the same label. Sorting the labels as plain text can also give an order that differs from sequence order. The safe representation keeps the residue number and insertion code as distinct data and uses an explicit sequence mapping to establish order.

This matters for multiple copies of a protein as well. Two chains may have identical polymer sequences but different local structural environments. A mapping to the same UniProt position does not establish that the surrounding contacts are the same. Keep the chain used for the structural interpretation attached to the biological site rather than collapsing all copies prematurely.

For nonpolymer species, a missing polymer sequence index should not be replaced with an invented amino-acid position. Ligands, waters, and other components have their own identifiers. A workflow that accepts every row of an atom table as a protein residue can create errors before mutation selection even begins.

Use sequence-aware mappings, then verify them

SIFTS provides mappings between PDB structures and resources including UniProt. The original SIFTS publication describes the integration. This is a useful starting point for locating a structural residue in a biological sequence.

Inspect the mapping rather than treating it as a substitute for construct verification. A PDB chain may contain engineered substitutions, a fusion, a fragment, or residues that do not map neatly to the reference. SIFTS does not tell you what is currently in your laboratory plasmid.

Align the full deposited polymer sequence and the intended biological reference where possible. An alignment built only from residues with coordinates can make an unresolved loop look like a sequence deletion. Preserve gaps and annotate their cause before deriving mutation positions.

Finally align the actual construct to that reference. A fixed offset is acceptable only across a verified uninterrupted region. If the construct includes an internal deletion or insertion, the relationship changes after that point.

Figure 2. Missing coordinates do not by themselves establish a deletion in the protein sequence. Map through the deposited and construct sequences before interpreting a gap. Conceptual illustration.

A worked example: why one offset fails

Consider a hypothetical protein whose native residue 21 becomes the first residue of the target segment. The expression construct adds a 12-residue tag and linker before that segment. Native position 35 therefore maps to construct position 27: 12 added residues plus 15 residues from native 21 through 35.

Now suppose the construct also deletes native residues 51–55. Native position 70 maps to construct position 57: 12 plus 50 residues from native 21 through 70, minus the five deleted residues. The offset that worked before the deletion no longer works afterward.

This example is deliberately simple. Real constructs can contain multiple edits, processing sites, and fusion partners. An alignment-based mapping makes those changes visible. Arithmetic performed from a remembered tag length does not.

Once the position is mapped, verify the local sequence on both sides of the target. If the proposal is V70A, the intended biological reference must contain valine at 70, and the mapped construct site must contain the expected starting residue. If it does not, stop and resolve the discrepancy before ordering.

Distinguish an unresolved site, a deleted site, and a mismatch

These three outcomes require different handoffs. An unresolved site may exist in the deposited polymer sequence and in the construct while lacking observed coordinates. It can still be mapped to the construct, but an atomic contact or packing claim at that site needs another source of structural evidence. A deleted site is absent from the construct and cannot receive a simple substitution there. A mismatch means the expected starting residue differs between references and needs explanation.

A mismatch may be legitimate. The structure could contain an engineered substitution, the laboratory construct could already carry a change, or the chosen biological sequence could be a different isoform. The response should be to preserve those distinctions, not to silently overwrite one sequence until the letters agree. An engineered residue in a structure is evidence about that experimental construct, not proof that the canonical sequence is wrong.

Sequence similarity alone may not uniquely resolve every mapping. Repetitive segments, duplicated domains, and short low-complexity regions can support more than one local alignment. Check wider context and domain organization. If two candidate positions remain plausible, mark the mapping as ambiguous and withhold the design handoff until the reference is resolved.

The same caution applies to processing. A signal peptide or propeptide may belong to the translated product but be absent from the mature protein used in the assay. Preserve both references with clear names. “Position 1” in the mature chain and “position 1” in the coding sequence describe different starting points, even when both are correct in their own context.

SIFTS provides a valuable bridge between deposited structures and biological references, as described by Velankar et al. (2013). The remaining laboratory-specific bridge is the current construct. No database mapping can verify an unrecorded cloning edit or establish which plasmid version was used in the latest preparation. That evidence comes from the project's sequence records.

Create a mutation handoff table

Use one row per requested change, including unchanged context around the site. A practical record contains:

• Biological accession, isoform, and reference position.
• Structure entry, chain, author residue label, and insertion code.
• Construct identifier and version, with construct amino-acid position.
• Expected starting amino acid and intended replacement.
• Corresponding coding-sequence location and local sequence context.
• Mapping evidence and any unresolved mismatch.

For DNA design, verify the actual codon in the supplied template. Amino-acid numbering does not replace checks on reading frame, strand, primers, or synthesis constraints. If several mutations are combined, verify the final translated sequence after applying all changes.

Validate the exported result, not only the selection

A structure viewer can highlight the correct residue while a downstream export uses a different numbering convention. Inspect the mutation list after export and again after it enters the design or ordering system.

Automated checks should reject a mutation when the stated wild-type residue does not match the reference sequence. They should also flag duplicate target identifiers, unmapped positions, insertion codes that were dropped, and sites outside the intended construct. A short local sequence window is valuable for human review because it can reveal an off-by-one error quickly.

In a workflow using Orbion Design or Orbion Bench, keep the exact input sequence attached to the candidate list. Structural insight and construct preparation are most useful when they share a verified residue mapping.

Make the mapping reviewable in both directions

A useful final check starts from the intended construct position and maps back to the biological and structural references. If the reverse path does not recover the site that motivated the proposal, the mapping is incomplete or inconsistent. This catches errors that can survive a one-way export, particularly when an intermediate tool renumbers residues from one.

For the hypothetical construct above, preserve two test cases: one before the internal deletion and one after it. A single successful check in the unchanged region does not validate the offset beyond the deletion. Also retain an explicit “absent from construct” result for a site inside the removed segment. A system that returns a plausible integer for every input is less trustworthy than one that clearly rejects an unmappable position.

Use the exact starting residue as an independent check, but do not rely on it alone. Common residues can appear at several neighboring positions, allowing an off-by-one error to pass a single-letter test. The surrounding sequence and the complete alignment provide stronger evidence. The final translated sequence should also match the intended construct after all approved changes have been applied.

Store the reference sequence or a content checksum with the mapping, alongside accession, version, and retrieval date. An accession identifies a database entry; it does not necessarily freeze the exact sequence used in the design. If a reference or construct changes, generate a new mapping version and state which candidate lists depend on the old one.

The deliverable is therefore more than a mutation string. It is a traceable relationship between a biological site, a structural observation, and the material to be built. Keeping that relationship explicit prevents a technically successful experiment from testing the wrong hypothesis.

Common questions

Can I add the tag length to every residue number?

Only after verifying that the remainder of the relationship is uninterrupted and uses the same reference. Truncations, internal deletions, insertions, and processing can invalidate a single offset.

Are residues missing from a PDB absent from the protein?

Not necessarily. They may be present in the experimental sequence but unresolved in the coordinates. Check the polymer sequence and missing-coordinate annotations before treating them as a deletion.

Which numbering should appear in a publication?

State the biological reference and numbering convention clearly, and provide construct-specific mapping where it matters. Readers should be able to identify the site without guessing which chain labels or tag offsets you used.

References

1. RCSB PDB-101. Primary sequences. Official guide. Accessed October 2026.
2. Velankar S, et al. SIFTS: Structure Integration with Function, Taxonomy and Sequences resource. Nucleic Acids Research. 2013. doi:10.1093/nar/gks1258
3. EMBL-EBI PDBe. SIFTS overview. Resource documentation. Accessed October 2026.
4. wwPDB. PDBx/mmCIF dictionary: author and polymer sequence identifiers. auth_seq_id and label_seq_id. Accessed October 2026.
5. wwPDB. PDBx/mmCIF user guide. Official documentation. Accessed October 2026.