Paste DNA, a protein, or a GenBank file. Choose a codon table and the rules. Run design. The job keeps running in the background, and the finished gene can be copied or saved as FASTA, GenBank, a CSV of the scores, or SBOL 3.
epistasis.context_naturalness absent
Checking whether ViennaRNA is installed.
Checking installed models.
Codon design keeps the protein the same and changes which codons spell it, so the gene suits the host you name. Later steps already have a working default.
DNA is the coding sequence only. A protein is reverse-translated with the codon table you load, then designed. A GenBank file uses one CDS feature. If the file has more than one, type that feature’s name.
For DNA, paste the coding sequence (the part that is translated) or a FASTA record. It should use A, C, G and T, and its length should divide by 3. Switch the menu above for a protein (M, L, K…) or a GenBank record.
rpsL is the E. coli K-12 MG1655 gene, 125 codons. BBa_E0040 is the untrimmed iGEM GFP part.
Off unless you tick it. Standard-code stops are TAA, TAG and TGA. An iGEM part that ends TAATAA, such as untrimmed BBa_E0040 (720 bp), keeps the first stop and drops the second (717 bp). With the box off, that part is still refused as CF-SEQ-003.
Use the gene, label, or locus tag of the CDS you want. Leave this blank when the file has one CDS.
No sequence yet. The rpsL demo is the E. coli K-12 MG1655 gene rpsL, 125 codons.
Leave blank to start from the coding sequence above. If you paste another coding sequence, it must translate to the same protein. A difference is CF-SEQ-004 and names the residue positions. The search then starts from this sequence.
The menu is the codon-usage files installed with this page, from GET /v1/hosts. It is not a catalogue of every organism. The name is a label on the job and has to look like genus_species:taxid/strain. The filled-in name is the rpsL demo host. The table is what the optimizer uses. The bundled table is examples/mg1655_ribosomal_usage.json, counted from E. coli K-12 MG1655 ribosomal proteins and labelled user-supplied.
CAI (codon adaptation index) compares your gene with this table. 1.0 means every counted codon is a top synonym of the table that was loaded. Stops and single-codon families are excluded. A codon whose weight is 0 makes the whole CAI 0. Kazusa (taxid 316407) is genome-wide frequencies converted to relative adaptiveness. Loading it replaces the table the search uses, so the result is a different number, not a second column on the ribosomal score. Upload accepts a codon_usage JSON, a Kazusa or CoCoPUTs-style CSV or text file, or a FASTA of coding sequences. The server counts or normalizes it and labels the result user-supplied unless you already marked the file. There is still no downloaded catalogue of species.
No codon table loaded yet.
Paste Kazusa text (triplet and a count in parentheses), a codon and count table, or FASTA coding sequences. Relative adaptiveness is each count divided by the maximum in its synonymous family. Unobserved families stay at 0. No pseudocount is added. The table is stored as user_supplied, not as a highly expressed Sharp and Li set, unless you say so. GET /v1/hosts remains the example files, not a catalogue of organisms.
Constraints are hard limits, separate from scores. Leave this on “No extra rules” for a first run. The rpsL example file, examples/rpsL_constraints.json, is named as an empty demonstration set: its list is empty, so it checks nothing.
Each button loads that preset into the JSON box and sends it. BioBrick RFC[10] forbids EcoRI, XbaI, SpeI, PstI and NotI. Golden Gate forbids BsaI and BsmBI. Synthesis forbids a 50 nt GC window outside 40–60%, a homopolymer longer than 6, and a repeated 12-mer. Gibson forbids that same repeated 12-mer on the 5′ flank, the coding sequence, and the 3′ flank as one string. Paste the promoter and the terminator into the flanks under Advanced. A clear Gibson result does not guarantee that the fragments will assemble. A rule this version does not search is reported as CF-FEAS-008 rather than treated as satisfied.
Paste a constraint-set document, or load the example and edit it. A file this page does not understand is reported back with a CF-FEAS code rather than ignored.
A preset is expanded by the server. The finished job lists the term ids it actually used. CAI only is the right first design. Harmonize needs a second table: the organism the DNA came from. Match frequency is ranked by the population loop, so use the standard budget if you want the search to move it; a fast budget reports the score and says the seeds were chosen without it. Structure needs ViennaRNA on this computer. The context model is used only when it is installed. Choose “Scores I tick below” to send the boxes instead of a preset.
The budget chooses which search stages run. Fast runs two stages (a first pass, then a polish) and is the right first choice. Standard adds a population of alternative designs, evolved for the number of generations below. Exact also asks for a final ranking stage this version does not run; the job says so with CF-ORCH-006 and still returns the earlier stages. Tiers are those stages. The progress panel shows one row per tier, not a single percentage.
How many rounds the population stage runs. Fast does not run that stage, so this number waits until you choose Standard or Exact. 4 generations and seed 7 are the rpsL example.
A number that makes the random part of the search repeatable. The same sequence, table, options, and seed produce the same design.
NCBI translation table. 11 is the bacterial code, which is what E. coli uses. Change it only if your host uses a different code.
Bases before and after the coding sequence that must not change, such as a promoter and a terminator. The Gibson repeat rule counts these together with the coding sequence. A 12-mer that occurs once in the promoter and once in the gene is a repeat the search can try to remove from the gene. A repeat that sits only in the flank is reported and left in place. Leave both blank if the coding sequence is the whole design. Sequence you do not paste is not scanned.
Used by Gibson report, not by the codon search. Each arm is this many bases copied from the end of the 5′ flank and the start of the 3′ flank. 40 is the default. 20 to 150 is the accepted range. Melting temperature is not computed, and bases are not invented when a flank is shorter than this number.
The local server asks for a token shaped like t_ and then 8 to 32 letters or digits. The filled-in value is fine when you are the only person using this computer. It is not a password for an account.
Built from the steps above. Tick the box only if you need to change a field the form does not show. While it is ticked, Run design sends this text as-is.
Run design posts the job and returns immediately. Fast, on the rpsL demo, usually finishes within a minute. Check these rules asks whether the DNA you pasted can satisfy the constraints. It does not start a design. Vendor-class screen flags GC windows, homopolymers, repeats and cloning sites using this suite’s presets. It is not a manufacturer score and not an expression assay. Global fold reports a ViennaRNA minimum free energy when that program is installed, or a labelled Nussinov base-pair count when it is not. That fold is not the windowed accessibility objective.
Rules not checked yet. Check these rules uses the DNA in the box. A protein or a GenBank file is checked on the finished design.
No job yet. A running job fills the tier tracks from GET /v1/jobs/{id}.
No jobs yet. Run design starts one in the background. This list is filled from GET /v1/jobs.
The designs the search kept. When there is one objective, this is that one design. The recommended candidate is marked. A column is a score the search used. UNKNOWN means that score was not measured. FASTA, GenBank, and CSV download from this result.
No front yet. A finished job fills this panel. A failure names its code and the reason in words.
Record a number you measured. It is stored and can be downloaded as CSV. The server may fit a straight line of that number against GC fraction. That line does not change codon choices, and no sequence model is trained.
A request this build cannot run answers CF-CLI-003. Nothing is silently skipped.
Primer design picks a forward and a reverse primer around a region you select. NCBI BLAST is not part of this tool. Specificity against a genome is not checked.
Paste the DNA you will amplify. A FASTA header (a line starting with > or ;) is dropped before numbering, the same way the API drops it, so the header does not shift the region. Lowercase is read as uppercase. A letter that is not A, C, G or T is named with its position and the search does not run. The sequence is drawn in lines of 60 bases. The number at each end of a line is the first or last base on that line, counting from 1. The numbers along the top are the position within the line. Drag across the letters, including from one line to the next, or type a start and end. The readout above the sequence is the start, the end, and the length. On a phone, Select the whole sequence is the sure path.
No region selected.
Bases are numbered from 1, and both ends are included. Typing a start and an end moves the highlight. Dragging the highlight fills these boxes. Keyboard: tab to these boxes. You do not have to drag.
Tm, the temperature where the primer is half bound. 60 °C suits a typical PCR.
How long the amplified piece may be, in base pairs. If your region is longer than the maximum, raise the maximum or select a shorter region.
How many bases on each side of your region the primers may sit in.
18 to 25 bases is a usual primer length. The minimum must be less than or equal to the maximum.
No primer pairs yet. Paste a template, choose a region, then Find primer pairs. You can also paste a pair you already have and check it below.
Paste two primers. This posts to /v1/primers/analyze and returns immediately. It does not search.