The Yeast Toolkit (YTK) and MoClo-yeast workflows use BsmBI
(recognition sequence CGTCTC) as their Golden Gate enzyme:
each cassette is flanked by outward-facing BsmBI sites that expose 4-nt
overhangs for modular ligation. If an internal BsmBI site exists
anywhere in the assembled construct, the enzyme will cut it during the
Golden Gate reaction and shatter the cassette, hence it’s important to
use parts that have been ‘domesticated’ so they are free from
restriction sites.
One issue that is harder to prevent is the junction-emergent site:
neither part carries BsmBI on its own, but when two specific parts are
placed in sequence the recognition site appears at their boundary. One
way to prevent this is by generating all possible combinations to screen
one at a time, which can quickly go exponential. We do not have to do
that, however, and can use search() directly on the
sequence graph and find matches across any combination of junctions.
Scenario: Expressing the bacterial geranylgeranyl pyrophosphate (GGPP) synthase crtE, the committed step in carotenoid biosynthesis, in Saccharomyces cerevisiae as part of a lycopene production strain. Three constitutive promoters and two terminators drawn from the YTK part collection are combined with two crtE variants (native and codon-optimised) in a 3 × 2 × 2 factorial design of 12 expression cassettes. During cloning, you notice that some combinations appear to assemble much more efficiently than others.
library(genr)
library(Biostrings)
ws <- file.path(tempdir(), "yeast_carotenoid")
dir.create(ws, showWarnings = FALSE)
setwd(ws)
repo <- Repository(NULL)Parts are DNAStringSet objects, one column per slot. Gen
stores each part’s name as its annotation label; when plotting we map
that name to a role (promoter, cds, terminator, or scar) and colour by
role.
Backbone scars are single-part columns so every path through the graph is a complete, self-contained expression cassette, ready for Golden Gate ligation into the pYTK001 destination vector.
# Helper: build a named DNAStringSet for one library column.
make_parts <- function(nms, seqs) {
DNAStringSet(setNames(seqs, nms))
}
# Upstream backbone scar, the left BsmBI assembly site.
left_scar <- make_parts(
"left_scar",
"AACTAGTATGGCGCGCCAACAAGGATCTGCGG"
)
# Three constitutive promoters from the YTK collection.
# TEF1p (strong) ends with ...GCGTCT, the first six bases of the BsmBI
# recognition sequence. When followed by a CDS that begins with C, the
# junction spells CGTCTC and the cassette is destroyed during assembly.
promoters <- make_parts(
c("TEF1p", "PGK1p", "HXT7p"),
c(
# TEF1p: strong, ~150 bp abbreviated. Ends with GCGTCT.
paste0(
"AGATCTGTTGACAAATAATCATCCGGCTCGTAATGTGTGGAATTGTGAGCGGATAACAAT",
"TTCACCAGTTCCCAACCTTACTCGGCACAAATCACCTTAATTTGATATTTATATCAAAGA",
"ATTTCTTTTGTATGGCGTCT"
),
# PGK1p: medium strength constitutive promoter.
paste0(
"AGTACTTTTTTTTTTATTTTTTTTTTTTTTTGAATTTTTTTTATTTTTTTCAATTCAATT",
"CATCATTTTTTTTTTTTTTTTTTTTTTTCTTCTTCTTTCTTTCTTTTTTTTTATTTTTTT",
"CATTTTCTTTTTTTTTTCAATT"
),
# HXT7p: strong under glucose limitation (useful for fed-batch).
paste0(
"ATCAGTGAGCAGTATGATGGTTATATTATATAGAGCATATATAGACATATATTATATTAA",
"ATAGATATAATTATATAATATAAGAAATAATAATATAAGAAATAATAATATAAGAAATAAT",
"ATCAGTGAGCAACTAGCAC"
)
)
)
# Two crtE variants (GGPP synthase from Erwinia uredovora).
# crtE_native: unoptimised bacterial codons, ORF starts with CATG.
# TEF1p + crtE_native junction: ...GCGTCT | CATG... = CGTCTC = BsmBI site!
# crtE_coopt: S. cerevisiae codon-optimised, ORF starts with ATGG.
# TEF1p + crtE_coopt junction: ...GCGTCT | ATGG... = CGTCTA = no BsmBI site.
cds_variants <- make_parts(
c("crtE_native", "crtE_coopt"),
c(
paste0(
"CATGGTTACTATCCCAGAAAACGTTGGTGCGAACATGGGCATCGTCGAAGCTGATGACGA",
"CGCTGCGGAATCCCAGTCTCAAGCTGCAATTTCCCAAATCAGACAAGCAATGATTTCTGA",
"AGGTATTCACCAATGCAAGAAATGA"
),
paste0(
"ATGGTTACCTCTCCAGAAAACGTTGGTGCTAACATGGGTATCGTCGAAGCTGACGACGAC",
"GCAGCTGAATCCCAATCTCAAGCAGCGATTTCCCAAATCAGACAAGCGATGATTTCTGAA",
"GGTATCCATCAATGTAAGAAATGA"
)
)
)
# Two terminators from the YTK collection.
terminators <- make_parts(
c("CYC1t", "ADH1t"),
c(
paste0(
"CATGTAATTAGTTATGTCACGCTTACATTCACGCCCTCCCCCACATCCGCTCTAACCAAA",
"AATGGCCCAAAACTCGCCCCATTTCACGTTTATTATTGATTTTTTTTTTTTTTTTTCTTT",
"TTTTTTTTTTTTTTTTTTTTTTT"
),
paste0(
"GCGAATTTCTTATGATTTATGATTTTTATTATTTATTATTTTTTATTATCTTTTTATTATT",
"TTATTATTTATAATTTAATATATAAATAATATATTTAAATATATATTTTTTTATTATTTTAT",
"TATTTATAATTTAATATATAAATAAT"
)
)
)
# Downstream backbone scar, the right BsmBI assembly site.
right_scar <- make_parts(
"right_scar",
"GCGGCCGCAAGCTTAGATCTCGCGGCCATACT"
)
parts_list <- list(left_scar, promoters, cds_variants, terminators, right_scar)
n_combos <- prod(lengths(parts_list))
cat(paste(lengths(parts_list), collapse = " × "), "=", n_combos, "combinations\n")## 1 × 3 × 2 × 2 × 1 = 12 combinationsimport_library() accepts a list of
DNAStringSet columns and builds a compact sequence graph
where parts are shared nodes. Search traverses the graph directly,
screening every junction simultaneously without enumerating individual
constructs.
repo$import_library("lycopene-cassette", parts_list)## <pointer: 0x600002f15100>
## attr(,"class")
## [1] "SequenceGraph"sg <- Filter(function(x) x$name() == "lycopene-cassette", repo$get_sequence_graphs())[[1]]
cat("Sequence graph:", sg$name(), "\n")## Sequence graph: lycopene-cassettecat("Sample: ", sg$sample_name(), "\n")## Sample: referenceThe annotation group added automatically on import labels every part.
Because GenPlot calls
auto_load_annotation_groups() at construction time, the
track appears without an explicit add_track_group()
call.
The colors parameter accepts a function that receives
each annotation and returns a hex color string. Gen stores each part’s
name as its annotation label, so we colour by name:
ROLE_COLORS holds the palette and NAME_ROLE
records which role each part name belongs to.
ROLE_COLORS <- c(
scar = "#aaaaaa",
promoter = "#4393c3",
cds = "#1a9850",
terminator = "#e08214"
)
# Gen does not persist per-part metadata, so drive colour from the part
# name (stored as the annotation label) rather than a stored role field.
NAME_ROLE <- c(
setNames(rep("scar", length(left_scar)), names(left_scar)),
setNames(rep("promoter", length(promoters)), names(promoters)),
setNames(rep("cds", length(cds_variants)), names(cds_variants)),
setNames(rep("terminator", length(terminators)), names(terminators)),
setNames(rep("scar", length(right_scar)), names(right_scar))
)p <- plot(sg, rows = 20L,
colors = function(ann) ROLE_COLORS[[NAME_ROLE[[ann$name]]]])
pThe fork-and-rejoin topology of the 12-path graph is immediately visible: three promoter branches merge at the CDS column, split again into two CDS variants, reconverge at the terminator column, and so on.
design <- expand.grid(
promoter = names(promoters),
cds = names(cds_variants),
terminator = names(terminators),
stringsAsFactors = FALSE
)
design$construct_id <- seq_len(nrow(design))
design## promoter cds terminator construct_id
## 1 TEF1p crtE_native CYC1t 1
## 2 PGK1p crtE_native CYC1t 2
## 3 HXT7p crtE_native CYC1t 3
## 4 TEF1p crtE_coopt CYC1t 4
## 5 PGK1p crtE_coopt CYC1t 5
## 6 HXT7p crtE_coopt CYC1t 6
## 7 TEF1p crtE_native ADH1t 7
## 8 PGK1p crtE_native ADH1t 8
## 9 HXT7p crtE_native ADH1t 9
## 10 TEF1p crtE_coopt ADH1t 10
## 11 PGK1p crtE_coopt ADH1t 11
## 12 HXT7p crtE_coopt ADH1t 12A single search() traverses all 12 paths simultaneously,
including every inter-part junction. BsmBI (CGTCTC) would
cut the cassette during the Golden Gate reaction; any internal site must
be eliminated before synthesis.
hits <- repo$search("CGTCTC", c(sg$id()), "dna")
n_sites <- length(hits[[1]]$matches)
cat("BsmBI sites found in the library graph:", n_sites, "\n")## BsmBI sites found in the library graph: 1The search returns unique site occurrences in the sequence graph, one per distinct location in the shared graph structure. That one site sits at the TEF1p→crtE_native junction and will be traversed by every construct that pairs those two parts, regardless of which terminator follows. Translating from graph site to blocked constructs happens in the design matrix below.
One site found, yet neither part contains BsmBI individually:
count_bsmbi <- function(seq) {
site <- "CGTCTC"
n <- 0L
for (i in seq_len(nchar(seq) - nchar(site) + 1L))
if (substr(seq, i, i + nchar(site) - 1L) == site) n <- n + 1L
n
}
all_parts <- c(promoters, cds_variants, terminators)
cat("BsmBI in individual parts:\n")## BsmBI in individual parts:for (nm in names(all_parts)) {
n <- count_bsmbi(as.character(all_parts[[nm]]))
flag <- if (n > 0L) " ← !" else ""
cat(sprintf(" %-18s %d site(s)%s\n", nm, n, flag))
}## TEF1p 0 site(s)
## PGK1p 0 site(s)
## HXT7p 0 site(s)
## crtE_native 0 site(s)
## crtE_coopt 0 site(s)
## CYC1t 0 site(s)
## ADH1t 0 site(s)tef1p_seq <- as.character(promoters[["TEF1p"]])
native_seq <- as.character(cds_variants[["crtE_native"]])
coopt_seq <- as.character(cds_variants[["crtE_coopt"]])
cat("\nTEF1p ends: ...", substr(tef1p_seq, nchar(tef1p_seq) - 11L, nchar(tef1p_seq)), "\n")##
## TEF1p ends: ... TGTATGGCGTCTcat("crtE_native starts: ", substr(native_seq, 1L, 12L), "...\n")## crtE_native starts: CATGGTTACTAT ...cat("crtE_coopt starts: ", substr(coopt_seq, 1L, 12L), "...\n")## crtE_coopt starts: ATGGTTACCTCT ...junction_bad <- paste0(tef1p_seq, native_seq)
junction_good <- paste0(tef1p_seq, coopt_seq)
cat("\nBsmBI at TEF1p + crtE_native junction:", count_bsmbi(junction_bad), "\n")##
## BsmBI at TEF1p + crtE_native junction: 1cat("BsmBI at TEF1p + crtE_coopt junction: ", count_bsmbi(junction_good), "\n")## BsmBI at TEF1p + crtE_coopt junction: 0The recognition sequence CGTCTC is split across the
junction:
...GCGTCT (the first six bases of the BsmBI
octamer)CATG... (supplying the terminal
C)Together they complete CGTCTC. This site exists on every
path through the graph that pairs TEF1p with crtE_native, regardless of
which terminator is chosen.
p2 <- plot(sg, rows = 20L,
colors = function(ann) ROLE_COLORS[[NAME_ROLE[[ann$name]]]])
m <- hits[[1]]$matches[[1]]
p2$go_to_match(m, center = TRUE)
p2$highlight_match(m, color = "red")
p2The highlighted region straddles the TEF1p→crtE_native edge. Because the site is stored once in the shared graph structure, a single highlight marks all paths that would be cut: here, both the CYC1t and the ADH1t variants of the TEF1p×crtE_native cassette.
design$blocked <- design$promoter == "TEF1p" & design$cds == "crtE_native"
blocked <- design[design$blocked, ]
clean <- design[!design$blocked, ]
cat(nrow(blocked), "blocked construct(s):\n")## 2 blocked construct(s):print(blocked[, c("construct_id", "promoter", "cds", "terminator")], row.names = FALSE)## construct_id promoter cds terminator
## 1 TEF1p crtE_native CYC1t
## 7 TEF1p crtE_native ADH1tcat("\n", nrow(clean), "clean constructs proceed to synthesis.\n")##
## 10 clean constructs proceed to synthesis.The two blocked cassettes (TEF1p × crtE_native, regardless of terminator) cannot be assembled by Golden Gate. The remaining ten are safe.
In practice, each part would be flanked by inward-facing BsmBI recognition sequences plus enough additional handle sequence to seat PCR primers and give the enzyme room to cut cleanly. The library graph holds the internal part content; the flanking architecture is applied at the cloning design stage.
Gen records every import and update as an operation. To attach the
screening conclusion to the import_library operation, open
the operations browser from the terminal:
gen operations --interactiveSelect the lycopene-cassette import entry to add a message, for example: “Screening complete: TEF1p×crtE_native blocked by junction-emergent BsmBI site at promoter-CDS boundary. Ten combinations cleared.”
To redesign the library with a different set of parts, create a new
branch from the operation just before the import. HEAD~1 is
shorthand for “one step back from the current operation” (analogous to
HEAD~1 in git):
gen checkout -b lycopene-v2 HEAD~1Then modify the parts list in R and call
import_library() again on the new branch. The original
branch and all its screening history remain intact; you can pick up that
experiment again at any point. ```