Codon usage matrix
The whole codon table for your region as a heatmap plus a sortable table — five metrics, two layouts, and a click that lands on the codon in the sequence.
What it does
Codon-usage analysis usually gives you one score for a whole coding sequence. That score tells you whether there is a problem; it never tells you which codon caused it. The matrix is the other half: every sense codon, coloured by the measure you choose, with the counts beside it.
Two views of the same numbers sit side by side. The heatmap is the picture — one cell per codon, grouped either by synonymous family or in the classic wall-chart arrangement. The table is the detail — codon, amino acid, count in your region, share of the family in the host, relative adaptiveness and RSCU, sortable on any column and filterable by typing.
Clicking a codon in either view selects its first occurrence on the sequence map, so the analysis puts you at the base rather than leaving you to search for it.
Going to a codon does not change what is being analysed. With the target set to the current selection, the three bases the click lands on are a place to look, not a new region: the scores, the rare-codon list and the profile go on describing the region you analysed, and the target only moves when you make a selection yourself.
When you'd use this
- Before ordering a synthetic gene. The score says "CAI 0.71"; the matrix says which family you are fighting and whether it is one rare codon used forty times or forty rare codons used once.
- When deciding what to optimise and what to leave. Codon optimisation is not free — it changes translation kinetics, and some rare codons are load-bearing. Seeing the whole family at once tells you whether a swap has a good alternative.
- When comparing your construct against the host you are expressing in. The RSCU pair — your region and the host table, on the same colour scale — is the direct comparison.
- When a construct expresses badly and the sequence is otherwise fine. A single family collapsed onto its rarest member is visible in one glance at the family layout.
What this computes
For every sense codon in the region, five numbers side by side:
- Count in your region
- Its share of the synonymous family in the host
- w, relative adaptiveness — the codon's frequency divided by the frequency of the most-used codon in the same family, so the family's favourite is 1 and everything else a fraction of it. These are the per-codon weights whose geometric mean is CAI, which is why this matrix and the CAI score in the panel above are two views of one calculation rather than two independent estimates
- RSCU in your region and RSCU in the host — the observed count divided by what it would be if every synonym in the family were used equally, i.e. the codon's share of its family times the family size, so 1 is no bias, above 1 is preferred and below 1 is avoided, independent of amino-acid composition
Both statistics are computed per synonymous family, which is what makes them comparable between sequences of different amino-acid composition, and stop codons and the degenerate cases are handled explicitly rather than folded in.
The region can be read on the minus strand, and it refuses rather than drawing nothing when it has no fetched per-codon host table to compare against — the curated organisms that ship offline are not enough for this matrix, and it says so.
Where it lives
The Codon Usage panel, below the analysis. The section header has a detach control that opens the matrix as a floating surface, which is where it is best used: given room, the heatmap and the table sit side by side instead of stacking in the rail.
It is part of the editor, not a calculator beside it. The region it analyses follows the panel's target — the whole sequence, the current selection, or a chosen feature — and clicking a codon selects bases on the map.
Inputs and outputs
In: a coding region and a fetched host usage table. The region comes from the panel's target control, and can be read on the minus strand. The host table is fetched from NCBI or Kazusa in the panel above.
Out: for every sense codon — count in the region, its share of the synonymous family in the host, relative adaptiveness w, RSCU in your region and RSCU in the host.
What it refuses, and what it says instead of drawing nothing:
- No fetched table. The matrix needs a full per-codon table; the curated organisms that ship with the workbench carry weights and rare-codon lists rather than a complete table, so the empty state names the two sources that do — NCBI and Kazusa — rather than leaving a blank card.
- A region that is not in frame. A region whose length is not a multiple of three has no codons to count, and the panel says so rather than analysing a shifted frame.
- A codon absent from your region has no observed RSCU — an em dash in the table, not a zero. Zero would be a claim about bias that a codon nobody used cannot support.
Controls
| Control | What it changes |
|---|---|
| Metric | Which of the five numbers colours the heatmap. The note under the control states what the selected one means and how to read 1.0. |
| Layout — By family | One row per amino acid, 1–6 codons wide. Every codon you could swap between is on one line, so the over-used member of a family is obvious. Ragged by nature: methionine and tryptophan have one codon each; leucine, serine and arginine have six. |
| Layout — Wall chart | The arrangement of a printed genetic-code chart: first base × second base down 16 rows, third base across four columns. Use it to find a codon by position rather than by scanning. |
| Click a cell or a row | Marks the codon in both views and selects its first occurrence on the sequence map. |
| Filter / sort (table) | Type to filter the 61 rows; sort by any column. Opens sorted by count, descending. |
| Target (panel-level) | Whole sequence, current selection, or a feature — and the minus strand for either. |
| Detach | Floats the section so the heatmap and table sit side by side. |
Algorithms you can choose
The five metrics are different questions, not display options:
| Metric | What it answers | Read 1.0 as |
|---|---|---|
| Relative adaptiveness (w) | Is this the host's preferred codon for its amino acid? w = f / fmax within the synonymous family — the weight CAI is built from (Sharp & Li, 1987) | the host's favourite codon in that family |
| RSCU — this region | Is your sequence biased, relative to using each synonym equally? | no bias |
| RSCU — host | Is the host biased, on the same scale? | no bias |
| Host usage fraction | What share of the family does this codon take in the host? | — (this one runs 0 to 1) |
| Observed count | How often does it actually occur in your region? | — (a raw count) |
The colour scale follows the metric, because the metrics have different neutral points. w and the host fraction run 0 → 1 and get a sequential ramp. RSCU is diverging around 1.0, where 1 means "used exactly as often as an unbiased sequence would", and it gets a diverging ramp with 1.0 at the midpoint. Putting RSCU on a sequential scale would bury the single landmark that makes it readable. RSCU's ceiling is the family size — 6 for leucine, serine and arginine — so the scale fits the data but never shrinks below 2, keeping 1.0 visibly mid-scale.
The science
RSCU (relative synonymous codon usage) is the observed count of a codon divided by what it would be if every synonym in its family were used equally. For a family of n synonyms it is the codon's share of the family multiplied by n, so it is 1 under no bias, above 1 when preferred, and below 1 when avoided — independent of amino-acid composition, which is the reason it exists.
w, relative adaptiveness, is each codon's frequency divided by the frequency of the most-used codon in the same family, so the family's favourite is 1 and everything else is a fraction of it. These are the per-codon weights whose geometric mean is CAI, which is why the matrix and the CAI score in the panel above are two views of the same arithmetic rather than two independent estimates.
Both are computed per synonymous family, which is what makes them comparable between sequences of different amino-acid composition. Stop codons and the degenerate cases are handled explicitly: a family wholly absent from the host table yields a neutral value rather than a division by zero, and a codon absent from your region yields no observed RSCU at all.
Limitations
- A host table is a snapshot of a database, not a property of the organism — a table computed over all coding sequences differs from one computed over highly expressed genes, and the second is what the classic CAI work used. Which table you fetched matters more than any metric choice.
- Bias is not the only cause of poor expression. mRNA structure, the ribosome binding site, GC content and repeats all matter, and none of them is visible here.
- Rare codons are not automatically bad. Some slow codons are functional — pausing sites, co-translational folding. The matrix shows you what is rare; whether to change it is your call.
- First occurrence only. Clicking a codon selects the first place it occurs in the region, not every place. Use the rare-codon list in the panel above for all positions of the ones that matter.
- Analysing the selection is not the same as browsing it. Once you select bases yourself the target has moved and the previous result no longer describes it, so the panel drops it and asks you to analyse again — that is the honest answer rather than leaving old numbers under a new region. Analyse a feature instead of a selection if you want to roam the sequence freely while the result stays put.
Related
In this group: Edit annotation · Enzyme · Insert sequence · Merge features · Manage enzymes · Dot plot · Dot plot grid · Melt curve · Amplification curves · Population genetics · RNA-seq · Site avoidance heat map · Protein coverage · Reference-gene stability · Digital PCR · Pairwise distance distribution · Nearest-neighbor distances · GC content distribution · Base composition per cycle · Pairwise distance grid · Per-cycle channel occupancy · Grammar assembly plan · Gibson assembly plan · Assembly plan build order · Golden Gate assembly plan · Guide track · Primer binding track · Hydropathy profile · Charge vs pH · Amino-acid composition · Hydrophobic moment · Optimization profiles · Digest gel · Virtual gel · Initiation-window fold · Antibody domain map · Electropherogram · Circuit netlist · Circuit transfer functions · Circuit construct · Construction graph · Protocol · Read Quality Profile · Read Alignment · Raw Signal · Verify vs Reference · Chromosomal microarray · Variant calls · Small molecule
Digital PCR
Turns partition counts into absolute copies per reaction with a confidence interval, and shows the ratios between targets without needing a standard curve.
Pairwise distance distribution
Every pair of barcodes in the set, counted by how far apart they are — the shape behind the single minimum-distance number.