Keep feature and language-variant meaning
Read missing, empty and repeated protein or language declarations before handing off FASTA and TSV; distinguish derived text from source evidence.
A count can conceal a different declaration
A matched CDS and an exported protein are different facts. A language-bearing TU and a uniquely paired TU are also different facts. Before handoff, compare those counts and inspect the statuses that explain the difference.
The GenBank request contains repeated NRPS_PKS values and missing qualifiers; the TMX request is an English/Spanish memory exported as UTF16LE XML. Both require structure and meaning rather than joining lines or extracting every attractive value.
Do not choose a duplicate silently
One empty translation plus MKL* is still two translation declarations. This tool labels the feature ambiguous and writes no protein for it, while retaining both declarations and their exact spans. Choosing the nonempty one would make an unrequested editorial decision.
For TMX, two ES-ES variants also remain ambiguous. The TSV row retains the TU status and counts but leaves paired text blank; JSON and original keep every alternative. A differently cased locale remains different until the user explicitly chooses it.
| Observation | Meaning | Receiving action |
|---|---|---|
| GenBank matched but unavailable | Missing, empty or repeated translation declaration | Review all qualifiers in source; do not infer protein from DNA |
| TMX ambiguous variants | More than one selected source or target alternative | Review complete alternatives; resolve in the authoring source |
| TMX derived text | Native code data excluded; hi/sub included | Keep inline-coded original when format/code identity matters |
Original bytes and normalized text serve different purposes
GenBank report spans are UTF-8 byte positions, including retained BOM and original line endings. Location strings describe declared positions but are never interpreted. A FASTA header encodes all identity fields reversibly; a readable short ID alone is insufficient when sources repeat names.
TMX JSON uses XML1.0 normalized text/attributes. original.tmx preserves UTF16LE/BE bytes and the complete source. Native bpt/ept/ph/it data is kept there and in JSON even though derived TSV excludes it. No external DTD is fetched and UDE mappings are not applied.
Keep the handoff complete within its budget
GenBank allows10MiB input and40MiB combined complete outputs; TMX allows2MiB input and10MiB outputs. Structure and field caps apply at the same time. Reduce the task explicitly after an atomic limit failure rather than treating a truncated preview as a full export.
Table cells stop at2,000 UTF-16 units and rows at200; copies above20,000 units are previews. Download every complete artifact before downstream checks. CSV/TSV formula protection adds an apostrophe to dangerous-looking text; original declarations remain available in JSON and original bytes.
- For proteins, inspect all matched unavailable statuses and decode each FASTA header.
- For pairs, compare all TU/variant ordering and inline-code identities, not just a few visible sentences.
- For either task, preserve the original alongside the full report, and cancel/retry using the same explicit controls.
Know what was never established
Neither export establishes biological validity, translation quality or full source-schema compliance. GenBank nucleotide and location interpretation are unreviewed; TMX Level2/DTD validation, locale fallback and unknown-attribute validity are unclaimed. These limits belong in the receiving workflow.
The author examples were incomplete or lacked original files. Complete synthetic starters make the workflow runnable without pretending to supply those authors’ data. Independent full-source readers test implementation behavior; actual browser interaction and central acceptance are separate evidence.
References
- GenBank author task
First-person demand for qualifier-selected existing protein declarations; complete author file unavailable.
- TMX author task
First-person UTF16LE English/Spanish TSV task; excerpt lacks final closures; starter is synthetic.
- GenBank sample record
Field-meaning correctness reference, not demand or full INSDC validation.
- TMX1.4b specification
Correctness reference for encoding and mixed inline content; this tool claims only its declared finite profile.
Tools in this category
Expand a tool to see its steps, options and supported formats, then open its workspace.
GenBank CDS declared-protein exportSelect CDS features by a literal qualifier value and export their unique declared translations, retaining all feature statuses, qualifiers and original byte spans.
Read existing protein declarations from a complete GenBank file. Missing, empty and repeated translations remain visible; the tool does not recompute proteins from DNA.
Steps
- Select a complete GenBank file or paste UTF-8 text. Set the qualifier key, literal value and contains/equals choice.
- Inspect matched and exported counts, then check each unavailable CDS status instead of assuming every match has a protein.
- Download proteins.faa with the complete report and original.gb; preserve the source when handing off declared locations.
Available options
- Qualifier key
- NRPS_PKS
- Literal value, case-sensitive
- PKS_
- Literal matching rule
- Contains literal · Equals literal
Capabilities and limits
- One UTF-8 nucleic GenBank, optional BOM and LF/CRLF, up to10MiB. Up to10,000 records,100,000 features,20,000 matched CDS,8,000,000 declared translation ASCII characters across all features and300,000 physical lines. Complete original, JSON, CSV and FASTA together must fit40MiB.
- Physical lines and declared locations each allow65,536 UTF-16 units; up to100,000 qualifiers per feature; each ordinary decoded qualifier up to1,048,576 UTF-16 units. Selection key is an ASCII identifier up to128 units; literal selection up to4,096. Limits apply together; input and output caps constrain simultaneous maxima.
- Recognizes the nucleic LOCUS prefix, exact fixed ASCII FEATURES columns and complete ORIGIN/CONTIG plus // structure. It does not validate full INSDC headers, nucleotide content/length, locations, codons, translation tables, biology or external references. Protein LOCUS, lone CR, invalid UTF-8, unsupported indentation and translation characters are rejected.
- Contains/equals selection is literal and case-sensitive, without regular expressions. Missing qualifier and no match are distinct. Only exactly one nonempty translation declaration exports; no declaration is missing, one empty declaration is empty, and any repeated declaration is ambiguous even if one is empty.
- Full JSON preserves every parsed record/feature, ordered repeated qualifiers and flags, quote form, declared location and exact UTF-8 source spans. FASTA headers reversibly include record/feature indices, LOCUS, all ID values and location. Only whitespace is removed from existing translation values; no biological validity is inferred.
- CSV includes every CDS status and protects formula-like text with a leading apostrophe; JSON keeps the underlying value. Preview shows first200 rows and first2,000 UTF-16 units per long cell. Above20,000 units, copy is a report preview; all downloads remain complete. Originals retain BOM and LF/CRLF byte for byte.
TMX language-pair and variant exportExport explicitly selected TMX language pairs to TSV while preserving all alternatives, metadata and mixed inline content in a complete report and exact original.
Inspect the full translation memory before choosing a unique pair. Exact locale selection avoids silent fallback; missing and repeated variants remain visible rather than being guessed.
Steps
- Choose a complete TMX file or paste UTF-8 XML. Set the exact source and target locale names.
- Run the export and inspect paired, ambiguous and missing unit counts; review all alternatives in the complete report.
- Download pairs.tsv, report.json and original.tmx together. Use TSV for derived text, not as a replacement for native inline codes.
Available options
- Exact source locale
- EN-GB
- Exact target locale
- ES-ES
Capabilities and limits
- One TMX1.4 file up to2MiB, up to2,000 translation units,10,000 language variants,30,000 XML element nodes and XML depth64. Explicit locale names each allow64 UTF-16 units. Complete original, JSON and TSV together must fit10MiB; all limits apply together.
- Files are recognized from their content, not extension. Fatal UTF-8 or UTF-16LE/BE with BOM and a matching XML1.0 encoding declaration are supported. Paste is UTF-8 text. A UTF-16 file is kept as raw bytes until its parser decodes it; invalid Unicode and encoding conflicts fail atomically.
- Source and target locale names are exact and case-sensitive, with no fallback or canonicalization. Only one source and one target variant yield text. Repeated locale variants are ambiguous; missing source, target or both stay visible with blank paired text. All alternatives remain in full JSON.
- Supports declared tmx/header/body/tu/tuv/seg/note/prop/ude/map containers and bpt/ept/it/ph/hi/sub inline profile. Derived text excludes native code data and includes hi/sub translatable text. Complete node order, attributes, mixed text/comments and paired bpt/ept remain in JSON. UDE mappings are retained but not applied.
- The exact literal SYSTEM tmx14.dtd is retained without loading it. Other DTDs, entity declarations, processing instructions, ut and custom elements are Unsupported. This is not DTD validation, TMX Level2 compliance, machine translation or validation of unknown attributes. XML1.0 text/attribute normalization applies to JSON; exact original bytes remain downloadable.
- TSV contains every TU status and protects formula-like cells with an apostrophe; retain JSON/original for unmodified values and code data. Table preview is first200 rows and2,000 UTF-16 units per long cell. Above20,000 units, copied text is a report preview; complete downloads retain every node, variant and pair.