Written guide to one within-sequence comparison. Bioinformatics Codon Usage can compare sense-codon counts in contiguous windows of an uploaded coding sequence. This ACTB example has complete saved counts and coordinates plus an earlier independent R comparison. The windows are portions of one sequence, not genes, biological replicates or expression measurements.
1. Start with the declared coding sequence
The saved input is ACTB_NM_001101.5_CDS.fasta, derived from the annotated CDS at transcript positions 85–1212, one-based and inclusive. That is the same interval as 84–1212 in zero-based, half-open notation. The complete transcript is 1,812 bases; this input contains only the 1,128-base CDS, including its terminal TAG stop.
Those 1,128 bases form 376 complete triplets: 375 sense codons and one stop codon. Keep the exact source accession, extraction coordinates and sequence identity. This route reads frame 0 of the supplied sequence; it does not find a coding frame or remove untranslated regions for you. Loading the full transcript would change the question and the codon grouping.
The current window helper accepts unambiguous A/C/G/T bases, normalizes whitespace and case, and converts RNA U to T. It uses the app’s standard genetic-code mapping to distinguish sense and stop codons. It needs at least 100 complete codons for two real windows. Inputs that cannot support the calculation should remain unavailable or refused; the app does not create extra observations to fill a plot.
2. Select correspondence analysis and inspect the windows
Load the CDS, open Codon Usage in Bioinformatics, clear the other analysis selections and choose Correspondence Analysis (correspondence_analysis). The saved receipt covers this one route, not every codon-usage metric. Confirm the input-QC result before interpreting the output.
Current source divides every complete frame-0 codon into 2–10 contiguous windows, with at least 50 complete codons per window and balanced widths differing by at most one. For 376 triplets, this produces seven windows. Stops are excluded from their counts without moving the window boundaries. Trailing incomplete bases, if any, are reported separately; this example has none.
| Window / codon positions | Saved counts |
|---|---|
| W1: 1–54 | 54 sense codons, no stops |
| W2: 55–108 | 54 sense codons, no stops |
| W3: 109–162 | 54 sense codons, no stops |
| W4: 163–216 | 54 sense codons, no stops |
| W5: 217–270 | 54 sense codons, no stops |
| W6: 271–323 | 53 sense codons, no stops |
| W7: 324–376 | 52 sense codons, one excluded stop |
Positions in this table are one-based, inclusive codon indices within the extracted CDS. They are not transcript nucleotide positions. The seven windows cover all 376 triplets. Their counted row masses use the 375 sense-codon total, so W7 has a different mass from W6 even though both span 53 triplets.
3. Understand what the correspondence calculation represents
The exported input table has 7 × 61 = 427 count cells, one for each window and standard sense-codon category. Ten categories have zero total in this ACTB sequence, leaving 51 observed categories in the decomposition. The zero-total columns remain in the saved input table; they are omitted only from the correspondence calculation. No pseudocounts or padding rows are added.
The method forms relative frequencies from the count matrix, compares them with row-mass × column-mass expectations, and decomposes the resulting chi-square-standardized residuals. Its row principal coordinates describe window differences in that geometry. This is correspondence analysis of observed sense-codon counts; it is not an RSCU table or ordinary covariance PCA.
Use the row mass and count table when interpreting positions. A point farther from the origin has a different codon profile under the model’s weighting; its location does not measure expression, selection strength or clinical significance. A visual separation among windows from one sequence does not supply independent replicates or a hypothesis-test result. The method reference is Nenadić and Greenacre’s primary article on the R ca package.
4. Read inertia and axis orientation with the saved coordinates
This example has six informative axes and total inertia 0.8568384844108371. The first two principal inertias are approximately 0.23209009 and 0.19056662. Their shares of total inertia are 27.0868% and 22.2407%: together approximately 49.3275%.
The two-axis plot therefore displays about half of the total inertia, not all of the saved six-dimensional geometry. Keep the other four axes when a comparison requires the full coordinates. The percentages are inertia shares; they are not prediction accuracy, confidence or explained gene expression. If window profiles have no informative variation, the current route reports that state rather than inventing axes.
Axis signs can reverse between equivalent decompositions. A sign-flipped axis alone is not disagreement. The existing independent check used the unmodified R ca 0.72 implementation in R 4.5.3 and compared principal inertias and the complete 7 × 7 coordinate Gram matrix, which is unchanged by axis sign reversals. R was used as an independent reference; it is not asserted to be a product runtime dependency.
This guide reopened the saved R outputs and native exports: all 427 counts agree, all 42 coordinate fields retain their exported numeric values, and the six inertia and 49 Gram-entry comparisons pass the existing tolerances of 1e−12 absolute plus 1e−10 relative. These are checks of saved values and geometry arithmetic. No R program, SVD, correspondence fit or sequence codon-counting analysis was rerun.
5. Save all counts and coordinates, not only the plot
- After the selected route completes, open the viewer and use Export All. Keep the QC, method summary, window-coordinate and input-count tables with the figure.
- The accepted historical export set contains seven artifacts: one PNG, four CSVs, summary text and analysis-report JSON. A separate artifact manifest supplies their hashes and sizes. The receipt identifies the
_2export set; earlier exports are preserved. Match the manifest hash rather than choosing a file only by its name or modification time. - CA input counts contains all 427 integer count cells plus Scope, Ref and Formula: 430 parameter rows. CA window coordinates has 73 parameter rows: 42 axis coordinates, four fields for each window (first codon, last codon, mass and excluded stops), and three explanatory fields. Read the first/last codon fields when mapping a point back to sequence position.
- Correspondence Analysis has 24 rows, including total inertia and all six axis inertias and percentages. Input QC has ten rows. This guide compares all 537 saved CSV/native fields. The saved coordinates and inertias retain numeric decimal representations; rounded axis labels and percentages in this guide are display summaries.
- The analysis report lists one plot, four tables and one text result. No sequence object is exported in this example. Keep the CDS FASTA and its extraction/source receipt separately. These artifacts describe an inspectable result snapshot; no reloadable session is established by this receipt.
The count table and coordinate table are scalar parameter/value CSVs in this export. Do not assume their layout is already a rectangular 7 × 61 matrix: pair each Wn codon or Wn axis n label with its value. Preserve the window labels and declared order when assembling a matrix for later review.
Evidence and related ACTB workflows
Read the input, current-source and saved-reference inventory. Current app, base-window, viewer and launcher files match the accepted Bioinformatics payload. The historical app and viewer hashes differ and remain explicit. The native receipt records automated Load/Run/view/export actions, not a new continuous recording or current-layout acceptance.
The ACTB ORF guide explains coding coordinates and complete translated FASTAs; the protein-composition guide explains residue counts and rounded property summaries. These are different questions using related sequences. This codon example does not establish expression, selection, clinical, hosted, real Windows/macOS or whole-product acceptance.