PepSeA is the second aligner Ideation computes alignments with. It runs MAFFT — the reference multiple-sequence-alignment program — scored on your organisation's own monomer similarity matrix. Like center-star, it runs on its own for every dataset at import; unlike center-star, when you run it yourself it asks how it should be run. This page explains how it works and what those settings change.
You do not have to change anything. Every parameter has a default that produces a sound alignment, and on most datasets the default is the right answer. Read this when you want to know what you would be changing, why an alignment is waiting, or when a dataset is large enough that the defaults stop being free.
For how PepSeA compares with center-star and with importing a file, see Understanding Peptide Alignment. For the step-by-step instructions, see How to Align Peptides.
How PepSeA works
Scored on your organisation's monomers
MAFFT aligns sequences of letters. To align peptides built from non-natural monomers, PepSeA gives each monomer a code and scores every pair of monomers by how structurally similar they are. Those scores form the monomer similarity matrix (shown as the distance matrix in the alignment's progress timeline), computed from the structures of the monomers registered for your organisation.
That is what sets it apart from center-star, which grades a substitution by natural analog and physicochemical family: PepSeA compares the monomers' actual structures, so two non-natural monomers that are chemically close align together even when they derive from different amino acids.
A monomer the matrix does not know — an unresolved monomer, for example — still aligns, but only with itself: it scores as unrelated to everything else.
The matrix has to be up to date
The matrix covers the monomers your organisation has registered. When an import or a monomer sync brings new ones, Ideation rebuilds the matrix so they are scored properly, and an alignment that needs those monomers waits for the rebuild rather than running without them. The very first PepSeA alignment of an organisation builds the matrix itself, which is why it takes longer than the ones after it.
In the progress timeline, a wait for the matrix reads as Queued, and the rebuild itself as Refreshing distance matrix. If the rebuild fails, the alignments waiting on it fail too, with "The peptide similarity matrix rebuild carrying this dataset's monomers failed; the alignment would misplace them. Retry once the matrix has been rebuilt." — run the alignment again later.
A limit on non-natural monomers
MAFFT has a limited pool of codes to give monomers. Natural amino acids keep their own letters, so only distinct non-natural monomers spend from that pool, and PepSeA supports at most 208 of them in one dataset by default.
A dataset above the limit cannot be aligned with PepSeA. The Align with PepSeA menu entry is disabled and says why — for example "Dataset uses 231 distinct non-natural monomers; PepSeA supports at most 208" — and the PepSeA alignment queued at import fails with the same message. Center-star and importing a file are not affected.
Backbone first, then branches
The backbone is aligned in one MAFFT run. Side chains are then grouped by the backbone column they attach to — the same grouping center-star uses, see Understanding Peptide Alignment — and each group is aligned in its own run, with the same settings as the backbone. The progress timeline shows this as Aligning branches, with a counter of the groups done.
Your symbols are kept
MAFFT is sent each monomer under the name your monomer library knows it by, so that aliases are scored as the right structure. Before the alignment is saved, the symbols from your own file are put back: the alignment editor and the viewer show the monomers as you wrote them.
When it runs
At import, on your organisation's defaults
Every peptide import queues a PepSeA alignment named PepSeA alignment, next to the center-star one. Nobody is there to answer a dialog, so it runs on your organisation's default settings — the same values the dialog preselects. It may take longer than the center-star alignment, since it can wait for the matrix to be rebuilt for the monomers the import brought.
On demand, with the parameters dialog
Choosing New Alignment → Align with PepSeA queues the alignment immediately and then opens the PepSeA alignment parameters dialog. That order is deliberate: the slow part is preparing the monomer similarity matrix, and it starts the moment the alignment is queued. Time spent in this dialog is not time added to the alignment.
Start Alignment releases it. Closing the dialog with Cancel discards the queued alignment. If you close the tab instead, the alignment is closed on its own after 30 minutes and listed as Replaced, with the reason "The alignment parameters were never submitted, so the alignment was not started." Nothing was computed either way.
A PepSeA alignment requested while an earlier one on the same dataset is still queued — and not yet picked up by a worker — replaces it: the earlier row is marked Replaced and links to the newer one. An alignment that is already running is never replaced.
Seeing what an alignment ran with
A PepSeA row in the alignment list carries a small settings icon at the end of its summary line. Hover it to see the Strategy and Gap penalty the alignment was run with — useful when comparing two PepSeA alignments of the same dataset.
When it fails
A PepSeA alignment that cannot finish ends as a Failed row; open its status (the red icon on the row) to read why. The reasons you are most likely to meet:
| Reason | What to do |
|---|---|
| Too many non-natural monomers ("Dataset uses N distinct non-natural monomers…") | Use center-star, or import an alignment from a file. |
| The similarity matrix rebuild failed or never finished | Run the alignment again once the matrix has been rebuilt. |
| "PepSeA alignment failed (…)", or the PepSeA service is unavailable | Run it again; if it fails again, the PepSeA service may be down — contact your administrator. |
| The worker stopped, or never picked the alignment up | Run it again. |
| The organisation's PepSeA settings could not be read | Ask your administrator to check the PEPTIDE_ALIGNMENT_PARAMS setting. |
A failed alignment never affects the dataset or the other alignments. Delete the failed row and run a new one from New Alignment.
PepSeA occasionally returns a row it cannot place. Those peptides are kept, shown with their full sequence from the first position, and listed in a warning on the alignment; their columns do not line up with the others. Re-running the alignment does not change it — edit those rows in the alignment editor, or use another alignment for them.
The parameters, at a glance
Three parameters govern a PepSeA run. Which of them you see is set per organisation, so your dialog may show fewer than are described here.
| Parameter | Shown by default | Default value | What it decides |
|---|---|---|---|
| Alignment strategy | Yes | Accurate (ginsi) |
How much of the search space MAFFT explores before it commits to a column. |
| Gap penalty | Yes | 7.0 (Compact) |
How dearly the aligner pays for opening a gap. |
| Gap extension | No — pinned | 0.0 |
How dearly it pays for lengthening a gap it has already opened. |
Every value a PepSeA run uses comes from your organisation's PEPTIDE_ALIGNMENT_PARAMS setting. If that setting cannot be read, the dialog says so and offers no fields, rather than running the alignment on values nobody chose. Cancel, and ask your administrator to check it.
Alignment strategy
MAFFT is not one algorithm but a family of them, trading accuracy against runtime. Three are shown by default:
| Choice | MAFFT strategy | Shown by default | What it does |
|---|---|---|---|
| Accurate | ginsi |
Yes — and preselected | The default. Computes every pairwise alignment globally and refines the result iteratively, so it explores more alternatives than the others. |
| Fast | fftns |
Yes | The escape hatch. Skips the refinement entirely. |
| Automatic | mafft --auto |
Yes | Lets MAFFT choose, from the size of your dataset, between its accurate, intermediate and fast strategies. |
The difference in cost is large and was measured, not estimated: on a 1000-peptide dataset, Accurate took 172.5 s and Fast 13.4 s. Quality follows the same order. Fast is an escape hatch, not an equal option.
Automatic is the answer whenever you would rather not think about size. It applies MAFFT's own rule — accurate strategies on small sets, fast ones on large — so it degrades gracefully as a dataset grows instead of leaving you to notice.
Which strategy for which dataset
| Your dataset | Use | Why |
|---|---|---|
| A normal SAR series, up to ~200 peptides | Accurate | Your peptides are related along their whole length and of comparable size, which is what a global strategy is for. This is the default, so there is nothing to do. |
| More than ~200 peptides | Automatic | Above 200 an accuracy strategy is outside what MAFFT recommends it for, and Automatic picks a strategy that fits the size instead. |
| Thousands of peptides, or a first look you intend to redo | Fast | When you want the run back in seconds and will refine the interesting subset later. |
| You would rather not reason about size at all | Automatic | It applies MAFFT's own size rule for you, on this dataset and on the next one. |
| One conserved region, the rest divergent, or long insertions | linsi (L-INS-i) |
Aligns the shared region locally instead of forcing the whole sequence into one frame. Ask your administrator to enable it. |
| Several conserved regions separated by long variable stretches | einsi (E-INS-i) |
Built for exactly that shape — several anchors, long gaps between them. Ask your administrator to enable it. |
| Large, but quality still matters | fftnsi (FFT-NS-i) |
Fast's progressive pass plus a refinement pass: between Fast and Accurate on both cost and quality. |
| Very short peptides | nwns / nwnsi |
MAFFT's FFT approximation buys nothing on very short sequences; these skip it. |
If your dataset does not obviously match a row, Automatic is the safe answer. It is never the wrong one — it only ever picks a strategy MAFFT itself considers appropriate for the size.
Why Ideation warns you above 200 sequences
MAFFT recommends its accuracy-oriented strategies for fewer than 200 sequences. Past that, their iterative refinement is no longer the method its authors intend you to use: the cost grows roughly with the square of the sequence count, while the gain over a faster strategy shrinks.
So when you select Accurate — or any other accuracy-oriented strategy your administrator has enabled — on a dataset above 200 peptides, Ideation says so and points at Automatic. It is guidance, not a refusal: you can run it anyway. On a large enough dataset, though, an accurate run can outlive the time an alignment is allowed to take, and a run that times out gives you nothing.
Automatic never carries that warning, because choosing the right strategy for the size is exactly what it does.
The other six strategies
MAFFT names nine strategies and PepSeA accepts all nine. Six are hidden by default — not because they are worse, but because they differ along axes that need bioinformatics knowledge to choose on, and offering nine buttons would obscure the three that matter.
Your administrator can enable any of them, rename any of them, and change the order they appear in, through the PEPTIDE_ALIGNMENT_PARAMS application setting.
| MAFFT strategy | Accuracy-oriented | Use it for |
|---|---|---|
linsi (L-INS-i) |
Yes | Peptides that share one alignable region and diverge elsewhere, or that carry long insertions. |
einsi (E-INS-i) |
Yes | Peptides with several conserved regions separated by long variable stretches. |
fftnsi (FFT-NS-i) |
No | A large set where you still want a refinement pass — between Fast and Accurate on both cost and quality. |
mafft (FFT-NS-2) |
No | MAFFT's own default command. The same strategy as Fast, under its own name. |
nwns (NW-NS-2) |
No | Very short peptides, where MAFFT's FFT approximation buys nothing. |
nwnsi (NW-NS-i) |
No | The same, with a refinement pass. |
The two accuracy-oriented ones carry the same 200-sequence guidance as Accurate, and Ideation warns on them too.
If the strategy buttons you see do not match the three above, that is why: your organisation has configured its own set.
Gap penalty
How dearly the aligner pays for opening a gap. Higher values give a tighter frame with fewer gaps, at the risk of pushing residues that are not related into the same column.
This parameter does not behave like a smooth dial. It has plateaus, and the marks on the slider are where they are because they were measured:
| Value | Mark | What you get |
|---|---|---|
| 0 – 5 | Loose | Identical alignments across this whole range. Moving the slider inside it changes nothing. |
| 7 | Compact (the default) | The first value that visibly changes the alignment. |
| 15 | Very compact | Internal gaps collapse. Useful on a series you know is closely related. |
| 50 | — (past the slider) | The alignment breaks down. |
So: if you want a tighter alignment than the default, go to Very compact. If you want a looser one, go down into Loose, where any value from 0 to 5 gives the same alignment.
Gap extension
Not shown, and pinned at 0.0 on purpose. It was measured inert anywhere between 0 and 2, and destructive below −2: there is no setting of it worth offering.
Your administrator can change the pinned value, and can hide or expose any of the parameters on this page, through the same PEPTIDE_ALIGNMENT_PARAMS setting. A parameter your organisation hides still runs — at the value your organisation pinned.
Where to go next
- Understanding Peptide Alignment — the three ways to get an alignment, and which one to use
- Understanding Center-Star Alignment — the other computed method
- How to Align Peptides — creating, importing, reviewing and repairing an alignment
- Understanding Peptides in Ideation — HELM and monomers, backbone and branches