Drybench Sequence.

In the workspace, Find a protein sequence accepts a gene, protein name or accession. Typing offers local suggestions; Search UniProt explicitly sends the query for broader matches. Selecting a match retrieves its full canonical record and opens it directly in an unchanged empty editor. Existing input requires confirmation with the accession, organism and residue count, and any named work is saved first. If inputs or ownership change during retrieval, nothing is replaced; review the source preview and choose Use sequence when ready. The imported FASTA and saved inputs retain source metadata and the full sequence even when it is too long for a share link. Editing the import does not recheck it against UniProt. DNA and RNA remain paste/import inputs; this is not a nucleotide database search.

What the app shows, how it works and how it was built, in the words of its own write-up, followed by the parameter defaults and the data sources with their licences.

What it does

One paste, every everyday operation: reverse complement, translation in any frame with any NCBI genetic code, all six frames, open reading frames, GC percent, the molecular weight of a DNA or RNA strand, motif search with IUPAC codes, and for a protein its molecular weight, isoelectric point and extinction coefficient at 280 nm, with the affinity tags and protease sites it carries found and the numbers given before and after cleavage. For oligos, the nearest-neighbour melting temperature with named salt and magnesium corrections, the pair's ΔTm with an annealing temperature stated as the rule of thumb it is, hairpin and dimer ΔG at 37 °C, the sites where a primer binds a template, and candidate primer pairs for a target region of a template, enumerated within it under stated constraints and each checked for one exact product.

Each operation has its own page that answers above the fold for a worked example and recomputes as you type, with a copy button and the method, constants and verification badge beneath; the workspace holds the paste and shows every operation on it. Raw letters, FASTA, multi-FASTA and GenBank records are read, the alphabet is detected and shown, and ambiguity codes are kept where the operation allows them and refused by name where it does not.

How it works

The numbers come from one pure module the suite shares (seq): the same molecular weight and extinction coefficient the protein and calculator apps show, the genetic code tables parsed from NCBI's published file, SantaLucia's 1998 nearest-neighbour parameters with the Owczarzy 2004 and 2008 salt corrections, and the SantaLucia and Hicks 2004 loop, mismatch and dangling-end increments for hairpins and dimers, every parameter transcribed from the paper it cites. The scientific methods run locally. In the workspace, optional UniProt lookup sends only submitted search text and selected accessions; pasted input is not sent. Scientific use needs no account; optional Bench sign-in uses session cookies. Inputs under 100 kb are computed on the page as you type; longer ones, to a limit of 1 Mb, run in a Web Worker that never fetches.

The link keeps your inputs after the # of the address, never a result: a nucleotide sequence is packed two bits a base, a protein deflated, and a paste too long for a link stays in this browser and travels to another app as a sequence object. A paste you pin, send or add to a study becomes a sequence object; a tagged protein you keep becomes a construct. Versioned example links (insulin@1 and the others) pin the method versions, so a teaching link gives the same numbers next year.

How it is checked

Each method is compared with a named external oracle on a published fixture set, never with its own output: 22 proteins and peptides against Expasy ProtParam's saved responses for molecular weight, isoelectric point and ε280; 45 oligos against Biopython's MeltingTemp under both concentration models with every salt regime; 20 hairpin and dimer cases and 11 salt probes against Primer3's ntthal; 108 frame translations and 5 annotated coding sequences against Biopython; 18 ORF cases against an independent scan cross-checked with NCBI ORFfinder; 5 DNA molecular weights by hand; and 101 peptides against RDKit's MolFromSequence for the SMILES sent to the molecule app. The fixture files, their hashes and the scripts that wrote them are published on the verification page, and each badge names the oracle, its version, the count, the tolerance and the date.

Equations and defaults

Reading a paste

Letters outside A to Z, * and - are dropped; the alphabet is detected as nucleotide when at least 90 percent of the letters are A, C, G, T, U or N, with the confidence shown, and a FASTA header's alphabet= field or a GenBank LOCUS line overrides the detection. Accessions in a header (UniProt, RefSeq, GenBank) become aliases of the sequence object.

Reverse complement

Each base to its Watson-Crick complement and each IUPAC code to the code of its complementary set (R to Y, K to M, B to V, D to H; S, W and N to themselves), the result read 3′ to 5′; U pairs with A for RNA.

Translation

NCBI genetic code tables version 4.6 (gc.prt, last modified 10 June 2020), table 1 by default; a frame starts at its first complete codon; * marks a stop; an ambiguity code gives X unless every base it stands for gives one amino acid; alternative initiation codons are off unless turned on, and then only the first codon is read as Met.

Open reading frames

From ATG (or the table's initiation codons when turned on) to the next in-frame stop on both strands; the coding length counts to the codon before the stop and must reach 75 nt by default; an ORF whose span lies inside another's is nested and hidden by default; ORFs without a stop before the sequence ends are listed apart. These are NCBI ORFfinder's conventions, adopted on 6 October 2026.

Protein molecular weight

The sum of ExPASy ProtParam's average residue masses plus water, the table the protein app has always used (id protparam); a table derived from the IUPAC CIAAW 2021 atomic weights sits behind the id ciaaw-2021 and reads higher by 0.03 Da per 100 residues on the fixtures.

Isoelectric point

Bisection on the net charge over pH 0 to 14 with the Bjellqvist pK values as ProtParam applies them, including a C-terminal pK of 3.55 whatever the last residue (confirmed on the fixtures); never a bound of the search.

ε280 and A280

Pace et al. 1995: 5,500 per Trp, 1,490 per Tyr and 125 per cystine, in both cysteine conventions; the absorbance of a 1 g/L solution is ε280 divided by the molecular weight.

Tags and protease sites

Nine peptide tags (His6 to His10, FLAG, HA, c-Myc, Strep-tag II, V5, S, Avi, T7) and five protease motifs (TEV ENLYFQ|G/S, thrombin LVPR|GS, Factor Xa I(E/D)GR|, HRV 3C LEVLFQ|GP, enterokinase DDDDK|) found by their published motifs; a cut is proposed at the site nearest an end and applied only when you confirm it; the SUMO protease has no motif and is not detected.

Oligo Tm

ΔH° and ΔS° from SantaLucia 1998 Table 2 (ten stacks, initiation with a terminal G·C or A·T, the symmetry term for a self-complementary oligo); Tm = ΔH°/(ΔS° + R ln(C/x)) − 273.15 with R = 1.987 cal K⁻¹ mol⁻¹; primer in excess (x = 1 with the primer's concentration) by default, equal strands (x = 4, or 1 with the symmetry term) as the named alternative; only A, C, G and T are read.

Salt corrections

Owczarzy 2004 equation 22 for monovalent ions by default (Na⁺ + K⁺ + Tris⁺/2); Owczarzy 2008 equations 16 and 18 to 20 whenever Mg²⁺ is present, with free Mg²⁺ after dNTP binding (Ka = 3 × 10⁴ M⁻¹) and the regime chosen by R = √[Mg²⁺]/[Mon⁺] exactly as published (R < 0.22 the monovalent equation; 0.22 ≤ R < 6 adjusted constants; R ≥ 6 the original constants), never smoothed, with a note when R is within 10 percent of a boundary; SantaLucia 1998 equation 7 as entropy as the alternative without Mg²⁺.

Fitted ranges and refusals

Owczarzy 2004 was fitted at 69 mM to 1.02 M Na⁺; the 50 mM default sits below that measured range. It is accepted, with 20 to 69 mM, as an extrapolation whose error has not been established here: Owczarzy 2008 places typical PCR buffers (20 to 100 mM monovalent) in the regime where this equation applies, though neither paper reports a fit below 69 mM. Outside 20 to 1,020 mM the Tm carries a fitted-range note. The Owczarzy 2008 free Mg²⁺ range, recalled as 0.5 to 600 mM, is marked unverified until the paper is read; SantaLucia 1998's range has not been read. Conditions no supported buffer reaches are refused: monovalent above 2,000 mM, Mg²⁺ above 1,000 mM, dNTPs above 100 mM, the oligo outside 1 pM to 1 mM, or monovalent below 1 mM with free Mg²⁺ below 0.1 mM.

Primer pair

ΔTm is the difference of the two Tm values; the annealing temperature is the lower Tm less 5 °C, a rule of thumb stated as such; with Mg²⁺ = 0 the Tm values use the monovalent correction, so a PCR buffer's Mg²⁺ should be set.

Hairpin and dimer ΔG°37: the structure class

Two-state secondary structures of one or two DNA oligos: Watson-Crick stacks, bulges, internal loops of at most 30 bases, hairpin loops of at least 3, dangling ends and terminal mismatches at helix ends, no multibranch loops, oligos of at most 100 bases. The structure reported is the one with the lowest ΔG°37 over every structure, scored at 1 M NaCl and corrected for salt as entropy.

Hairpin and dimer ΔG°37: sources

SantaLucia and Hicks 2004, Annu Rev Biophys Biomol Struct 33:415-440, Tables 1 to 4 and equations 7 to 13, read 6 October 2026 (initiation +0.2 kcal/mol and −5.7 cal K⁻¹ mol⁻¹, terminal A·T +2.2 and +6.9, internal single mismatches, loop initiation by length with the Jacobson-Stockmayer extrapolation 2.44 RT ln(n/x) between listed lengths and the value for 30 beyond, 0.3 kcal/mol per base of internal-loop asymmetry); the Watson-Crick stacks from SantaLucia 1998 Table 2 (read 6 October 2026); the 32 dangling ends from Bommarito, Peyret and SantaLucia 2000 as Table 3 of the 2004 review reprints them (read 6 October 2026); the hairpin terminal-mismatch increments from the Nearest Neighbor Database's DNA table (rna.urmc.rochester.edu/NNDB, read and saved 6 October 2026), the only public source of the values the review attributes to Varma and SantaLucia (unpublished).

Hairpin and dimer ΔG°37: conventions, as Primer3's thal, inferred from its output

Each of these fills a gap the paper leaves and was inferred from ntthal's output on the fixtures and its man page, not from its code: a terminal mismatch at a duplex end or inside an internal loop is the sum of the two dangling-end increments; a destabilising dangling end at a helix end is left off; a first mismatch in a 4-base hairpin loop that would be Watson-Crick (and so cannot pair) stacks as that pair would, and in a longer loop its bases count as two dangling ends; a 3-base loop closed by A·T takes the terminal A·T penalty; a bulge of 1 keeps the stack across it and longer bulges take the A·T penalties of both closing pairs; no symmetry correction for a self-dimer; salt by the sodium equivalent Na⁺ + 120 √(Mg²⁺ − dNTP) mM (von Ahsen 2001, as the Primer3 manual states), as 0.368 ln[Na⁺]eq cal K⁻¹ mol⁻¹ per pair beyond the first.

Hairpin and dimer ΔG°37: departures from ntthal, with their size

Selection: ntthal reports the structure with the highest melting temperature (its man page); this app reports the lowest ΔG°37, so the two can show different structures and ntthal a less negative ΔG; on the fixtures the structures coincide on 19 of 27 cases with a structure and differ by up to 3.6 kcal/mol where they do not (the H08 hairpin). Salt: ntthal counts one pair fewer for a hairpin whose stem includes the oligo's last base; this app counts every pair, so ntthal reads lower by one salt step, up to 0.37 kcal/mol on H01, H02, H04, H06 and H07. Flanks: two complementary bases flanking a duplex end or a hairpin's open end earn no credit here (they pair, or they count nothing), where ntthal keeps a single dangling end (0.23 kcal/mol on H08). The gate: ntthal's own structures rescored under these rules within 0.4 kcal/mol on all 15 hairpin fixtures and 0.1 on all 16 dimer fixtures, this app's structure never less stable than ntthal's, and the four no-structure cases held; the per-case table is below under Verification.

Hairpin and dimer ΔG°37: unchecked classes

The special 3- and 4-base loop bonuses of SantaLucia and Hicks 2004 (their supplementary table was not available) are not applied, so hairpins with GNA or GNRA loops come out less stable than they are; on the two probe cases ntthal is more stable by 1.5 (GNRA) and 2.3 kcal/mol (GNA). Bulges agree with ntthal within 0.01 kcal/mol on the two probe cases but are not gated. Asymmetric internal loops follow equation 12 of the review; on the one 2×3 probe this app is 1.7 kcal/mol more stable than ntthal, so the rule is marked unchecked until more cases are measured.

Hairpin and dimer ΔG°37: the search check

On 200 seeded random oligos per engine (hairpins of 10 to 14 bases, dimers of 5 to 8 bases each) every structure is enumerated outright and scored by the same rules; the dynamic programme's minimum equals that minimum on every one (src/lib/seq/structure.test.ts).

DNA and RNA molecular weight

The residue formulas of the nucleoside 5′-monophosphates less water (dA C10H12N5O5P, dC C9H12N3O6P, dG C10H12N5O6P, dT C10H13N2O7P; the ribonucleotides with one more oxygen), summed with the IUPAC CIAAW 2021 atomic weights; a linear strand gains water and loses a phosphate group for its 5′-OH end unless the 5′-phosphate is chosen; double-stranded adds the complementary strand; circular adds nothing. The result is a neutral free acid.

GC percent

(G + C + S) over (A + C + G + T + U + S + W); the other IUPAC codes are left out of both counts, and the result is undefined when nothing can be counted.

Motif search and binding sites

Each IUPAC code matches its set of bases, on both strands of DNA when asked; a protein motif matches letter for letter. A primer's binding sites are the places on either strand where its 3′ six bases match exactly and the rest has at most the allowed mismatches; a product runs from a forward site to the next reverse site.

Predicted PCR product (exact-pair-v1)

Each primer's annealing part, or its reverse complement, is matched base for base on an A, C, G and T template of at most 50,000 bases, on both strands; a circular template is read once around, so a site or a product may cross the origin. A site's start is the primer's 5′ end and its end the 3′ end, 1-based on the top strand. A product runs from the forward primer's 5′ end to the reverse primer's 5′ end, read 5′ to 3′ from the forward primer, with the forward tail before it and the reverse tail, reverse complemented, after it; tails never count towards the annealing length. Pairs whose footprints overlap are counted and build no product. Every product is listed, never the first chosen: none, unique or ambiguous. More than 100 sites for one primer or more than 100 products refuse, since the list would be incomplete. Annealing parts of 8 to 200 bases and tails of at most 200. Mismatches, single-primer products and added bases are not modelled; the product is a sequence construction, not evidence of amplification or of specificity in a genome.

Primer design (bounded-flanks-v1)

A bounded enumeration within the pasted A, C, G and T template (at most 50,000 bases, linear or circular): candidate windows of the chosen lengths whose 3′ ends sit 0 to the chosen flank (at most 200 bases) before the target's first base and after its last, read around the circle when circular; a window is rejected, under the first reason that applies and counted, when it falls off a linear template, touches the target or an excluded region, or has a GC or a nearest-neighbour Tm (annealing part alone, the primer in excess, the panel's conditions) outside the bounds. Each side is ranked by |Tm − preferred Tm| + 0.2 × |length − preferred length| and cut to 24; the pairs are ranked by the two Tm departures, 0.2 times the two length departures and the pair's ΔTm, after the product length (tails included) and ΔTm bounds; the best 24 pairs are checked by the exact-pair method for one exact product at the designed sites and the first 5 that pass are returned, each with its exact site and product counts. A search cut at any cap is reported as capped. No Tm of a tailed oligo, no annealing temperature, no genome specificity, no mismatched binding and no amplification outcome is derived.

Fragment search in saved sequences

A query of 4 to 10,000 A, C, G and T bases (spaces and line breaks removed, nothing else) and its reverse complement are matched base for base on the saved, presented strand of each saved DNA sequence. An IUPAC code in a saved sequence is a literal letter no query base matches, never a wildcard, and the query itself is never a pattern. Overlapping hits are all kept; a palindromic query's hit is reported once, as both strands. Coordinates are 1-based and inclusive on the saved strand whatever the strand of the hit; on a circular molecule each start is read once and a hit across the origin is marked.

Peptides as SMILES

A linear peptide of the 20 standard L-amino acids with free amine and acid ends, neutral, cysteines as free thiols, drawn as RDKit's MolFromSequence draws it, for the molecule app; up to 2,000 residues; a trailing stop is dropped and any other letter refuses the peptide by name.

Links

A nucleotide sequence travels packed two bits a base (the b: payload), a protein deflated (z:) or plain, and a sequence too long for a link stays in this browser; the operation, frame, genetic code, motif and conditions travel as plain keys after the # of the address.

Data sources, licences and versions

Times on screen are local time with the UTC offset; times in exports are UTC. Dates with no time are shown at the precision the source gave them.

The Export menu offers the results as CSV (the ORF, primer, Tm and binding-site tables with units in the headers, the method and constants table per row, with the provenance sidecar), the pasted sequence as FASTA with its alphabet in the header (the sequence payload another app reads back), and the paste and operation on screen as a printed handout with the method block. Copy link puts the sequence and the settings after the # of the address; the method block copies as text.

Release gate: before the app leaves the preview, its exports and printed reports will carry the data versions, parameters, limits and a timestamp needed to reproduce a result, as the regulatory position says. This section will describe each format as it ships.

Each module below is a set of automated tests in the app's source code, with the expected results written into the tests. They check the app's logic against recorded inputs and stored reference records; none of them calls a live service.

src/lib/seq/translate.test.ts and orf.test.ts

108 frame translations and 5 annotated coding sequences against Biopython exactly, and 18 ORF cases field by field against the independent scan cross-checked with Biopython and NCBI ORFfinder, from pipeline/sequence-toolkit/fixtures.

src/lib/seq/protein.test.ts

22 proteins and peptides against Expasy ProtParam's saved responses: molecular weight within 0.01 Da, pI within 0.01 and never a bound, ε280 exact; the CIAAW table's difference measured beside them.

src/lib/seq/tm.test.ts

45 oligos against Biopython MeltingTemp.Tm_NN within 0.1 °C under both concentration models, the Owczarzy 2008 regime exact, the seven refusals and three acceptance cases exact.

src/lib/seq/structure.test.ts

The hairpin and dimer gate against ntthal: its own structures rescored within the tolerance on every case, this app's structure never less stable, the no-structure cases held, and the search check on 200 random oligos per engine; the five loop-class probes reported, not gated.

src/lib/seq/dna-mw.test.ts and peptide.test.ts

5 DNA molecular weights against hand arithmetic within 0.01 g/mol; 89 peptides as SMILES against RDKit's canonical form and 12 refusals exact.

src/lib/seq/primer-design-fixtures.test.ts

9 primer-design cases and 4 refusals against the independent derivation (every melting temperature from Biopython, the windows, counts, ranking and caps derived in the script, exact binding by the exact-match derivation, one case worked by hand): every returned pair, coordinate, product, count and cap exact, Tm and ranking terms within 1e-6 °C.

pipeline/sequence-toolkit/fixtures.test.ts

Every fixture file matches the sha256 the oracle scripts recorded in the manifest, with their case counts and the rulings that shaped them.

pipeline/sequence-toolkit/reference.py and peptides.py

The oracle scripts (Python 3.11 with the pinned requirements): Biopython 1.88, primer3-py 2.3.1, NCBI ORFfinder 0.4.3, the saved ProtParam responses, exact rational arithmetic and RDKit 2026.03.6; they write the fixtures and their hashes.

Release gate: before the app leaves the preview, each release will publish a verification dataset with expected outputs on this page, so you can run your own checks.

Scientific computation runs in your browser, and individual apps need no account. Optional Bench accounts and backups use the site server and configured providers. The privacy statement, the hosts it contacts and the headers its workspace is served with are generated from its registry entry, the same record that sets the browser's policy.

What leaves your browser

Hosts the app contacts

• rest.uniprot.org

  Operator

  UniProt Consortium (EMBL-EBI, SIB and PIR)

  What for

  On Search UniProt, find protein records; on result selection, retrieve the full canonical sequence and source metadata.

  What it receives

  the search text submitted to UniProt and the selected accession; never the separate pasted sequence

  When

  on lookup

Headers the workspace is served with