Workflows

"I want to do X" — the route through the tools, end to end.

A tool page tells you what one tool does. These pages tell you which tools to use in what order for a real piece of bench work, and what to check between the steps. They exist because the tools are one workbench with shared context, not a drawer of utilities: a PCR product goes straight into an assembly, an assembly product goes straight into a digest, and a Sanger read comes back and is graded against the construct you designed three steps earlier.

Every number on these pages was produced by running the tool on a named file in data/, so you can reproduce them. Where a tool cannot answer something, the page says so rather than filling the gap in.

• Clone and verify — clone an insert into a vector and prove it worked: linearize → assemble → check by digest → verify by Sanger. The workflow the rest of the cloning tools hang off. (pBR322 and the Gibson pair.)

• Primer → PCR → gel — design primers, simulate the reaction, predict the band: region → ranked pairs → binding sites on the whole molecule → cycling program → the band at the agarose percentage you actually own. (pBR322 tet and ROP.)

• CRISPR guide → validation — design a guide, check where else it binds, measure the edit you got: 3,857 candidates on EMX1, three on-target models that disagree, an off-target census that says what it is specificity of, then ICE over real edited chromatograms. (EMX1 and the cane-toad tyrosinase traces.)

• Optimize a CDS — recode a gene for expression in another host: what yeast ADH1 looks like to E. coli, the constraints that must hold, and what the optimization costs you in the original host. (SC_ADH1.gb.)

• Read a VCF by locus — open a call set and jump to a region: indexed region queries, the refusal that catches a mismatched index, and a bcftools-shaped census. (data/samples/VCF/.)

• MSA → tree — align a set of homologues and build a tree from it: read the alignment before trusting it, four methods that answer different questions, and bootstrap support read as resolved-versus-not. (Eight cricket COI barcodes.)

• Protein → peptides — will mass spectrometry see the region you care about? Read the protein your construct encodes, digest it, apply the window the instrument can actually detect, and watch coverage fall from 100 % to 73.2 % — then get it back with a missed cleavage or a second enzyme. (eGFP in pEGFP-N1.)

• Check read quality — read a sequencing run's per-cycle quality curve, then settle the one thing a FASTQ never states: which quality encoding it is written in. The same 25 reads read as mean Q19.2 or Q50.2 depending on that choice. (A public Drosophila RNA-seq subset, the FASTQ specification's own example, and an older Illumina run.)

---

More routes are worth writing — Golden Gate / MoClo hierarchies end to end, a Sanger verification run from raw traces, and a protein-first path from UniProt to a back-translated, optimized construct — and they will be added as the tool pages behind them are written. (The protein-first path is now half-covered: Protein → peptides takes a construct's own CDS through analysis and digestion. What is still unwritten is the leg that starts at a UniProt entry.)