Barcode sets¶

A barcode is a stretch of DNA that names one part of a library. The set has to hold three things: no two barcodes may read as one another, none may spell a site the cloning enzyme reads, and none may put a stop codon in the frame the construct is read in. One command designs a set that holds all three, and says how far apart it stands.

When to use it¶

• You are naming the parts of a pooled or combinatorial library, one barcode each.
• The parts will be cut and joined, so no barcode and no join may spell a site the enzyme reads.
• The barcode is translated, so it must not put a stop in the frame.

Golden Gate assembly joins the parts that carry these barcodes, and codon optimisation writes the parts themselves free of the same sites.

What you need¶

No input file. You give the numbers: how many barcodes, how long each one is, the cloning scar that joins one to the next, and every enzyme to keep out. The run below asks for the set a small barcoded library would use.

Run it¶

pixi run mbio barcode-design 24 --length 11 --scar AGCG --forbid BsaI --forbid BbsI --out plan/

It prints the set, then the two files it wrote:

24 barcode(s) of 11 bases, scar AGCG, 3 apart by sequence-levenshtein, free of BsaI, BbsI, seed 0, checks pass
plan/barcodes.tsv
plan/checks.tsv

Candidates are drawn from the whole space of 11-base sequences in a shuffled order and kept when they pass every rule. Walking that space in order instead opens a set on a run of one base, which no later filter repairs. Each candidate is read with the scar on both sides of it, so a BsaI or BbsI site that only shows up across a join is caught as well.

3 apart by sequence-levenshtein is the rule the set holds to. Sequence-Levenshtein is not plain Levenshtein: it counts substitutions, insertions and deletions, and charges nothing for the bases a read gains or loses past the barcode's end, which is what a lost base does to a read. checks pass means nothing failed.

What it wrote¶

Both names link to what that run wrote, published here unedited. The same count, the same rules and the same seed always give the same 24 barcodes, so you can store the command rather than the set. To change the set, change the command and run it again.

How to read the sheet¶

barcodes.tsv holds the number and the bases. The first five here are TACCATTGTCT, TCCTCCCCAGG, ACCAGTGGCCT, GACAGCAAGTT and TTGATAGTGAA. Keep that numbering, because it is what pairs a barcode with the part it names.

checks.tsv holds a row for each check: its name, its verdict, the number it measured and a line of detail. rules found nothing broken. separation is 3, which is how far apart the set really stands, against a rule of 3.

deletion_ambiguity has no verdict, on purpose. It measures the share of one-base deletions that another barcode could have left too, which is zero here, and then stops, because no sourced threshold says what share is too much. A check nothing sourced judges reports its number and leaves the decision with you.

Before you order¶

• Check the count covers the parts you have and the ones you may add.
• Check --scar is the four bases your own cloning leaves behind.
• Check --phase is the frame the construct reads the barcode in, or pass --untranslated.
• Name every enzyme you will cut with, including one you only use in a later round.
• Read deletion_ambiguity and decide whether that share is one you can live with.

Tips and troubleshooting¶

The design gives up before it has your set. The space ran out under the rules you asked for, and the message says how many it found and which rule rejected the most candidates. Add a base to --length, or drop --distance by one.

Nothing translates your barcode. Pass --untranslated, and the stop-codon rule and the whole-codon rule both go away. Otherwise --phase says how many bases of the barcode's first codon are read before it: 0, 1 or 2.

You want the same set again next year. One seed gives one set, so keep the command. Another --seed gives a different set of the same quality.

Where the numbers come from¶

The default distance of 3 is the distance a published barcode set was measured to hold. Sequence-Levenshtein is the default metric because the long reads a barcode block is read with fail mostly by insertion and deletion. Hamming counts only the positions two barcodes of one length differ in, so it cannot see an indel at any distance.

The homopolymer cap of 5 bases is on for the same reason, since a run is where an indel lands. That number is a judgement rather than a measurement. A GC band is off, and --gc-band turns one on: every band in the barcode literature traces back to one uncited sentence in Hamady et al. 2008, and Xu et al. 2009, the one study that imposed a band and then measured the result, found GC was not what separated its probes.

A set that is ligated has its junctions judged the way a Golden Gate overhang set is; see ligation fidelity. pixi run mbio barcode-design --help lists the rest.

References¶

• Buschmann, T. and Bystrykh, L.V. (2013) Levenshtein error-correcting barcodes for multiplexed DNA sequencing. BMC Bioinformatics 14: 272.
• Hamady, M., Walker, J.J., Harris, J.K., Gold, N.J. and Knight, R. (2008) Error-correcting barcoded primers for pyrosequencing hundreds of samples in multiplex. Nature Methods 5, 235–237.
• Xu, Q., Schlabach, M.R., Hannon, G.J. and Elledge, S.J. (2009) Design of 240,000 orthogonal 25mer DNA barcode probes. PNAS 106, 2289–2294.
• Hawkins, J.A. et al. (2018) Indel-correcting DNA barcodes for high-throughput sequencing. PNAS 115, E6217–E6226.