Workflow
Gene to construct
A coding sequence is not a construct. This workflow handles the distance between them — host-adapted codons, chosen regulatory parts, verification primers, and an assembly record that survives being handed to someone else.
25
founding constructs
31,911
bp, founding ten
14.1 Mbp
designed DNA
SBOL
assembly record
01
The goal
Between a protein you have decided to express and a record a lab can work from sits a stack of unglamorous decisions: which codons suit this host, which promoter and terminator, what order the parts go in, which junctions need checking, and how all of that gets written down so it still makes sense next year. Done by hand, this is where projects accumulate their quiet errors.
The construct builder keeps each decision explicit and attached to the part it affects. The output is an annotated assembly, not a concatenated string.
02
Stage by stage
Bring in the sequence
Import a coding sequence from a project, the catalog, or a GenBank or FASTA upload. Biopython parses it and the annotation engine marks features, so you start from something described rather than something raw.
Adapt the codons
DNA Chisel rewrites the coding sequence for the target host under explicit constraints — usage tables, GC windows, avoided sites, repeats. Every constraint it satisfied is recorded next to the result.
Choose the regulatory parts
Promoters, terminators, selection markers and targeting signals come from the parts library, filtered by host. Each part carries its provenance, including an Addgene lookup where the part has a public deposit.
Assemble
pydna models the assembly and tracks junctions, while SBOL captures the design as structured parts with roles and relationships rather than as flat text. seqviz renders the map in the browser.
Design the checks
Primer3 proposes verification primers across junctions and inserts, with melting temperatures and amplicon sizes shown so you can judge whether the set is actually practical.
Annotate and freeze
The finished construct is re-annotated end to end and versioned. Later edits create a new version; the old one stays readable, which is what makes a construct citable.
03
Worked example
- Reference set
- Founding ten peppers
- Constructs
- 25
- Guide RNAs
- 24
- Total designed length
- 31,911 bp
- Codon adaptation
- DNA Chisel, per-host tables
- Assembly model
- pydna, exported as SBOL
- Verification
- Primer3, junction-spanning sets
The founding ten are the reference the rest of the catalog was built against: twenty-five constructs and twenty-four guide RNAs totalling 31,911 base pairs of designed DNA. Small enough to hold in your head, complete enough to expose every decision the builder has to make. The catalog has since grown to 14.1 Mbp across 1,000 constructs, and every one of them is assembled by the same path.
Working from that set is the fastest way to learn the builder. Open a founding construct, look at how its parts are annotated and why its primers sit where they do, then clone it and change one thing.
04
What you get
A versioned construct with full annotation, an SBOL representation of the assembly, a verification primer set, part provenance including Addgene references where they exist, and export to GenBank, FASTA and vendor order sheets whenever you are ready to hand it over.
Everything in this workflow — codon adaptation, primers, annotation, assembly — is a CPU engine and runs free. Credits are only spent upstream, on the structure and variant work that decided what to build.
Ready to run it
Try this in the Splice Platform.
CPU engines are free to run. GPU engines run on prepaid credits.
Related