Fragment analysis

1. 01Reads a fragment-analysis ABIF (.fsa/.hid) — the half of the format that carries no base calls, which a Sanger trace reader refuses as a damaged file
2. 02All five sizing curves the instrument vendor names, not one with a flag: Local Southern, Global Southern, 2nd- and 3rd-order least squares, and a natural cubic spline
3. 03The size standard read out of the file's own analysis, so the ladder is the run's, not a table typed in somewhere else
4. 04The ladder CHANNEL measured rather than assumed, by matching the standard's own scans against every channel's peaks — with the evidence shown
5. 05Sizing accuracy measured on your run by leave-one-out over the standard's own fragments, in bases, beside every size
6. 06A size produced by a method's documented fallback says so, because an extrapolated length is not comparable with an interpolated one
7. 07Four- and five-dye chemistry, including the fifth dye's separate ABIF tags
8. 08The analysis offset honoured, so a run whose analyzed channels start later than the collected ones is sized correctly rather than silently wrongly
9. 09Off-scale peaks flagged: position and size still read, height reported as the floor it is
10. 10An electropherogram on an axis in BASES rather than scans, with the standard's own fragments drawn on it
11. 11Shared or per-channel scaling, because a size standard run beside taller allele peaks is invisible on one axis
12. 12Peak threshold and an analysis range in bases, with exact accounting of every peak they excluded
13. 1347 marker kits bundled, every one NCBI OSIRIS publishes — alleles named from the kit's own public-domain ladder definition, with the locus and the deviation in bases
14. 14Every peak that is NOT an allele named by kind: no locus, off ladder, ambiguous between two, or a by-product
15. 15Adenylation and stutter recognised by the instrument authority's own rules and thresholds, each reported with its parent peak and height ratio, both adjustable
16. 16One allele, one call: a second peak at an already-called allele is reported with its reason rather than listed twice
17. 17Each locus's repeat unit measured off its own ladder rather than tabulated

packages/ports/assay/fragment plus the fragment reader in packages/ports/formats/trace. The reader is graded against Biopython's generic ABIF walk over every bundled file; peak detection is graded byte-for-byte against Fragman 1.1.0 (CRAN), the maintained R package for fragment analysis, run on the same real files; the spline and the regressions against R's own stats::spline and stats::lm; and the sizing against the one authority that cannot be argued with — the size standard's own fragments, whose lengths are known by construction. The methods themselves are implemented from Applied Biosystems' DNA Fragment Analysis by Capillary Electrophoresis user guide. The panel runs on @ports/fragment.

Everything runs client-side. Files are parsed in the browser and projects are saved on this machine, so an unpublished sequence is never uploaded.

Why does my .fsa open as garbage, or not at all?

Because .fsa means two different things. NCBI's submission tools call a plain nucleotide FASTA .fsa, and Applied Biosystems writes a fragment-analysis electropherogram under the same suffix. Those are text and binary respectively, so whichever one your software guessed, it is wrong half the time. Here the contents decide, not the name.

What is the difference between an .ab1 and an .fsa?

They are the same container holding different documents. An .ab1 is a sequencing read: dye channels the instrument has already interpreted into base calls, each with a confidence. An .fsa is a fragment run: dye channels and a size standard, with no bases anywhere in it. A reader built for one cannot read the other, which is why they open in different places here.

Which sizing method should I use?

Local Southern unless you have a reason, which is what the instrument software defaults to. It uses the four standard fragments nearest each peak, so it is accurate where it applies and falls back where it cannot. Switch to Global Southern when you suspect one standard fragment ran anomalously — it fits one curve to all of them, so a single bad point barely moves it, at the cost of fitting the run as a whole less well.

How accurate are the sizes?

The panel measures it on your own run rather than quoting a figure. Each standard fragment is removed in turn and sized from the others, and since its real length is known by construction, the difference is a real error. That number is reported beside the results, and a difference between two peaks smaller than it is not a difference the run can resolve.

Does it call alleles or genotypes?

Yes, when you name a marker kit. Choosing one names each peak's locus and allele from that kit's own ladder definition, and every peak it cannot name says why: no locus, off ladder, ambiguous between two alleles, or a by-product of a taller peak. Without a kit the run is still a complete measurement — it reports lengths.

Why is the same allele listed twice?

It is not. A real fragment usually appears as a pair of peaks one base apart, because the polymerase adds a non-templated adenine to most of the product and not all of it, and a repeat also leaves a stutter peak one repeat short. Both are recognised as by-products of the taller peak and named as such, with the height ratio that classified them, and both thresholds are yours to change.

Why does amelogenin say 1 and 2 instead of X and Y?

Because that is what the kit definition says. It declares the two alleles positionally, at 103 and 109 bases, and carries no mapping to those letters anywhere in the file. Those two are the X and Y amplicons. The label shown is the kit's own rather than a translation of it, and the panel marks the locus as positionally labelled so the number is not read as a repeat count.

Is anything uploaded?

No. The file is read and sized in your browser.