Pairwise alignment

The Align tab aligns another sequence to the open document, for example a Sanger read against the plasmid, or a synthesised fragment against its design. In front of a protein it aligns two proteins, scored by BLOSUM62.

How to align

1. Paste the other sequence into the box: bare bases, a FASTA record or a GenBank record all work. Or drop a file on the box, or click Choose file…: GenBank, FASTA, SnapGene, AB1 and FASTQ files (gzipped too) are read into the box without opening a tab. To align against a document already open in another tab, pick it from Open tab… (the same kind only: protein against protein); its current sequence is used, unsaved edits, topology and, for a read, its qualities and trace included. Choose or drop several files at once to align all of their records together (below).
2. If the box holds several records (a FASTA file of several reads, say), pick the one to align from the list that appears — or align them all, see Aligning a batch of reads.
3. Click Align, the button just under the box and the file controls. Both orientations of the other sequence are considered and the better one is shown; the heading says when it was the reverse complement. The defaults suit most sequences; the Options group under the button holds the rest:
4. In Options, choose Global (end to end) to align the whole of both sequences (Needleman–Wunsch), or Local (best region) to find the best-matching stretch (Smith–Waterman). Local is the right choice for a read against a plasmid, and is chosen for you when the box holds a read (a file with base qualities) or a sequence under half as long as what it is aligned to; the note in the Options group says so. Global would score such a sequence across the whole of the plasmid and report a perfect read at 17% identity. A mode you choose yourself stays chosen while the Align tab is open, whatever you then put in the box.
5. Tick Against selection only to align against the selected part of the document instead of all of it (it is greyed out until something is selected). Click Align again to re-run with the new options.

The result appears in a Result group under Options, and reports the score, percent identity, the number of columns and of gap columns. It does not print the alignment itself, which is too wide for a side panel: click Large view to read it in a window of its own, along the whole sequence (see below). Identity counts only the identical columns.

Point at the result's summary line (a bordered button with a › at its end; hover it, or Tab to it) to see where the alignment sits: the region it covers is drawn in the map and the sequence view, in the same dashed purple as other previews, without touching your selection. A region that runs through the origin of a circular plasmid is drawn at both ends. Click it (or press Enter) to select the covered bases, so you can annotate or copy them, and scroll to them; the highlight steps aside for the selection. On a phone, tap to select. The highlight goes when the pointer or focus leaves, when you edit, or when you leave the tab. The same works for each read in a batch's list, where the whole row is the target, not just the name.

A result belongs to its input

A result stays on screen only while it describes what is in the panel. Pasting or editing the sequence box, choosing or dropping a file, taking another open tab or picking another record clears the result and its group (or the batch's list) and stops an alignment that is still running, so an old result never looks like the new input's. Changing Mode or Against selection only is different, since people tweak these while comparing: the result stays but is dimmed, "Input changed, align again" shows under the Result group's name, and its summary line and Large view buttons are disabled until you click Align again. A Large view window that is already open keeps its own result until you close it.

Aligning a batch of reads

A FASTQ file of reads, or a FASTA of several Sanger reads, can be aligned all at once — the usual way to check a batch of clones.

1. Drop the file on the box, or click Choose file…. The note under the box says how many records it has. To check clones sequenced as separate files, pick or drop all the files at once, in any mix of formats: their records become the samples, in file order, and the box is emptied (its text would be unwieldy). Rows are named by record and file. A new pick replaces the samples loaded, as does typing in the box; a file that cannot be read is named and the others kept. The 96-record limit counts across all the files.
2. Leave the mode as it is, and each record is aligned in the mode it would get on its own: Local for a read or anything under half the document's length, Global otherwise. A mode you choose is used for every record. Tick Against selection only if the reads should be looked for in part of the document only.
3. Choose All N records in the record list under the box; the Align button then reads Align all. Click it. Each record is aligned against the document in turn, in the background; the bar counts reads (12 of 96) and Cancel stops the batch, keeping the reads already aligned. Leaving the Align tab cancels it too.
4. The reads are listed as they finish, in a Results group, one row each: the name, its length and whether it aligned reversed, the identity, the number of differences on confident bases (Q20+ diffs, following Confident from; a read without qualities counts all of its differences, as Diffs), and the stretch of the document it covers. Click Identity to sort lowest first — the clones to look at — then highest first, then back to the file's order.
5. Click a read's name to pick it, then click Large view of all (the highlighted button above the list): the window stacks every read, with the one you picked already selected, its score line showing.

A read that could not be aligned (nothing good enough after trimming, too large) is listed with the reason, and the rest carry on. At most 96 records (a plate) are aligned at once; for a larger file the note says so and the list has no All records choice: pick one record, or split the file. A batch is aligned in a band around the words each read shares with the document, then checked against every alignment there that could score higher (see long reads below), which takes about a second for a plate of Sanger reads against a 5 kb plasmid. A read gets the score it would get aligned on its own, unless it matched better somewhere else in the document that its words do not point to; where two alignments score the same (an indel in a repeat can sit in either copy), the batch may show the other one.

Assembling reads into a contig

Reads of an insert whose sequence you do not know yet, or reads you want one consensus from, can be joined with no reference. Load the reads as for a batch (a FASTQ, AB1 files or a FASTA of two or more records, at most 96) and click Assemble reads, beside Align. Nothing is aligned against the document; the document is not used at all.

Each read is first trimmed by its qualities (Trim poor ends, in the Reads group, as for an alignment). The longest read starts a contig, and every other read is joined to it if it overlaps an end, or lies inside it, by at least 25 bases at 90% identity or better. A read may be on either strand: those turned over to fit are paler in the layout. Reads that join nothing start contigs of their own, listed in the Contig picker, biggest first. A read with nothing left after trimming is counted as left out.

The Assembly group shows the picked contig: its length, depth, a layout with a bar for each read and a red mark at each disagreement, and the disagreements listed with the position, the call and each read's base and quality there. The consensus is called from every read at once, weighted by quality: one high-quality base outweighs several poor ones; where reads at Q20 or better differ and none is clearly right, the consensus has an IUPAC code (R for A or G, and so on); a base most reads lack is left out. A disagreement is such a code, or a base a confident read calls differently from the consensus. Reads without qualities (a FASTA) count every base as Q20, which the note says.

Save consensus as document opens the consensus in a new tab, named after the first read, with its call qualities when the reads had any. To check it against the intended construct, open that tab's Align tab and drop or paste the construct there.

Assembly is greedy and meant for a few to a few dozen Sanger reads of one region. It does not resolve repeats, and a read is placed at the first contig it fits.

Verifying a plate of clones

A whole-plasmid sequencing run returns one consensus a clone, often a plate of them, and the question is which clones are the construct you meant to make. File ▸ Verify clones… answers it for the whole plate at once.

1. Open the construct each clone should be (or several constructs, for a mixed plate) in their own tabs. The dialog lists the open DNA documents; tick the expected ones. The document in front is ticked to begin with.
2. Add consensus files… takes FASTA, GenBank, SnapGene and the other formats the app opens, as many files as you like and several records to a file, up to 384 clones. A file with one record is named by the file, one with several by the file and the record.
3. Verify checks them one after another; Stop keeps what is done.

Each clone is lined up with its construct through the origin and on either strand, so a consensus written from another start, or reverse complemented, is not a difference. It is then aligned end to end, and the table gives the clone's Verdict, its Identity, and what differs, each difference placed on the construct's features:

A difference is an SNV (one base), a substitution (several adjacent bases), an insertion or deletion (fewer than 20 bases) or an extra or missing region (20 or more), with its position in the construct's numbering. In a feature it reads as "frameshift in CDS bla" (an insertion or deletion of a number of bases that is not a multiple of three in a CDS), "in-frame deletion of 3 bp in CDS bla", "promoter pLac changed" or "CDS bla deleted". An insertion at the very edge of a feature is next to it, not in it. Ambiguity codes in the consensus (an N) are not counted as differences; the row says how many there were.

Mixed plates. With several constructs ticked, each clone goes to the one it shares the most sequence with. To send clones by name instead, type text in the box beside a construct: a clone whose file name contains it (case does not matter) goes to that construct, the longest text winning. A clone with no match by name is placed by sequence.

Click a clone's name to open Compare with… between it and its construct, as for any two documents. Export CSV saves the table, a row a clone with every difference in the last column. Nothing leaves your browser and no file is changed.

Aligning a read with its qualities

An AB1 or FASTQ file dropped on the box (or chosen with Choose file…) keeps its base qualities for the alignment; the note under the box says with base qualities. They last while the box holds the file's text: edit it and it is read as plain bases again. See Sequencing reads for what the qualities are.

A Reads group, under Options, appears when the box (or the document) holds a read.

• Trim poor ends, on by default, cuts the read's unreliable start and tail before aligning — the first 20–50 bases and the end of a Sanger read, typically. It keeps the stretch whose bases are mostly better than Q13 (a 5% chance of error; see Trim at below), by Mott's algorithm as phred uses it, so a single poor base inside a good stretch stays. The result says how many bases went from each end; untick it to align the whole read.
• The differences, by confidence. Under the score line, a line says how many differences from the document sit on bases the read was sure of (Q20 or better, one error in a hundred, unless set otherwise) and how many on poor ones. Each confident difference is listed with its position in the document and its quality; click one to select it there. Those are the ones worth a look; the poor ones are usually the sequencer, not the clone.
• Where confident starts, and where trimming cuts. In the Reads group, Confident from sets the quality a base must have for a difference on it to count as confident, from Q10 to Q50: Q20 suits Sanger reads, a nanopore service's consensus (Q40 and up) wants Q40, raw nanopore reads Q10 or Q13. The count and the list follow it at once, without aligning again, and so does the status bar's share of good bases for an opened read. Trim at sets the error rate trimming keeps bases better than: Q13 (5%) is phred's usual, Q20 (1%) or Q30 (0.1%) trim harder, Q10 (10%) keeps more of a noisy read; it applies the next time you align. Both are remembered in this browser with the view preferences.

The panel does not draw a read's poor bases or an AB1 trace; the large view does (below).

When the document is the read

Opened the AB1 or FASTQ itself, and want to check it against the plasmid? Paste or drop the plasmid (or any reference) into the box. While the open document is a read — opened from an AB1 or FASTQ, its bases unedited — and the box holds a sequence without qualities of its own, the note above the box says … is a read, and the document is aligned _to_ the box: the box is the reference, the document the read. Its qualities and trace are then used as above — trimming, confident and poor differences.

• The reference is the box's sequence and the read the document, which is turned into its reverse complement when that is what aligned.
• Each confident difference is named at its place in the read, with the reference position beside it: Mismatch at 36 (pRef 41), Q40. Clicking it selects that base in the document, the read. Select aligned region in this document selects the stretch of the read that aligned.
• A GenBank reference marked circular is aligned through its origin, as a circular document is.
• Against selection only (in Options) does not apply: the whole read is aligned.
• Untick This document is the read (in Options) to align the box's sequence to the document the usual way round, without the qualities. A box that holds a read of its own (a dropped AB1 or FASTQ) is always aligned the usual way.

Aligning several sequences together

Pairwise alignment lines two sequences up. To compare several variants of a construct, or several homologous coding sequences or proteins, align them all at once: paste a multi-record FASTA or GenBank text into the box, or drop or choose several files, and the Multiple alignment group appears under the Align button.

1. Put three to fifty sequences in the box (one record each). Include this document adds the open document as one more, so a box of two records is enough for three.
2. Click Align N together. It runs in the background, with a progress bar and Cancel, and opens a window of its own.
3. The window shows every sequence in one column space. Over them are the consensus (the commonest letter in each column; blank where most sequences have a gap) and a bar of how conserved the column is. Each column's cells are shaded by the share of sequences that carry the commonest letter, so conserved stretches stand out. For bases, the Shading menu can instead colour each base A, C, G and T. Names stay in place as you scroll sideways.
4. Save aligned FASTA and Save Clustal download the alignment; Copy FASTA copies it. Clustal files carry the usual conservation line (* where every sequence agrees, : where every pair of residues of a protein scores above zero in BLOSUM62).

Bases are scored as in pairwise alignment, and proteins (when the open document is a protein) by BLOSUM62. The method is progressive: a guide tree is built from how many short words each pair of sequences shares, the sequences are joined along it, and the result is refined by re-aligning the two sides of each branch where that scores better. Gaps at either end of the alignment are cheap, so a fragment is not penalised for being short. It does not match the best programs (MAFFT, Clustal Omega) column for column; on simulated sets it finds about 96% as many of the true residue pairs as MAFFT does, and it is exact for sequences that differ by a few substitutions and indels. The sequences are taken as they are: it does not reverse-complement one that is on the other strand, and a circular plasmid is aligned from the first base it is written from. All of it runs in your browser.

Scoring and limits

Scores use match +5, mismatch −4, gap open −10 and gap extend −0.5 (the EMBOSS DNAfull scheme). Ambiguity codes score by the same scheme, by the bases they stand for: A against R (A or G) +1, A against N −2. An N is therefore neither a match nor a full mismatch, and a run of them does not attract the alignment.

Alignment runs in a background thread and the interface stays responsive. A long one shows a progress bar with the percentage done and a Cancel button; leaving the Align tab cancels it too. For large inputs the orientation is picked first from the short words the two sequences share, so only one alignment runs; when neither orientation clearly wins, both are aligned.

A long read against the plasmid it came from — a nanopore read of 10 kb, say — is aligned in a band around the words the two share rather than over every pair of bases, which takes a fraction of a second rather than several. The answer is then checked: the score found limits how many gaps any better alignment could have, and so how far from the band it could stray, and that whole region is searched too. A Global alignment is then the best there is; a Local one is the best among alignments that run through the stretch the shared words point to. The check is skipped when it would cost as much as aligning in full, mostly for a long, noisy read in Local (a 10 kb nanopore read, say): its alignment is then the best the band found, which is usually but not always the best there is, since a tandem repeat or a long insertion near an end of the read can lead the band astray. Such an alignment is marked: its heading says "not checked against the best possible alignment", and in a batch of reads its identity carries an asterisk with the same words on hover. Two sequences that share too little for a band, and would need more than 150 million cells in full (about 12 kb × 12 kb), are refused to protect the browser's memory; align against a selection for those.

Reads through the origin

On a circular document, a Local alignment against the whole document finds a read that runs through the origin, such as a whole-plasmid nanopore read that happens to start in the middle. That holds for a read with only a few bases before the origin too, singly or in Align all: those bases, and any difference among them, are part of the alignment. Positions are numbered as the document's, going from its last base back to 1, and Select aligned region in this document selects across the origin.

In the large view, reads that run through the origin and reads that stop short of it are combined by base: coverage, the feature verdicts' read and strand counts, and the "samples disagree" flag treat a base as one base whichever side of the origin a read reaches it from. The protein effect of a difference in a coding sequence across the origin is read the same way, including a codon split by the origin, when the read that carries the difference runs through it; a read that stops at the origin holds only part of such a codon and shows no effect for it.

Large view

Large view on a result (the highlighted button under the score line) opens the alignment in a window of its own, and Large view of all on an Align all batch stacks every aligned read at once, starting with the read picked in the list selected. The document sits at the top with a position ruler and each sample is a row under it, all moving together under one scroll, so a column reads straight down through the document and every sample. Differences from the document have a coloured background, by where the column falls in the document: red in a CDS or an ORF (a difference that may change a protein), blue in any other feature, grey outside every feature. A column in several features takes the highest, CDS first. A key sits in the toolbar, and the heading counts the differing columns in each. Colour is not the only cue: a changed base is a plain block, a base the sample lacks (a dash) has a bar along its foot, a base the sample adds has one along its head, and a match through an ambiguity code is a paler block. The colours are from the whole document, whether or not Features and ORFs are on; ORFs count once the ORFs panel has found them. With no document behind the reference (a pasted sequence) every difference is grey and there is no key.

• The strip above the rows spans the whole alignment. It marks every difference, in the same three colours, and shows the stretch now in view as a rectangle; click or drag on it to move there. A thin coverage band along its foot shows how many reads cover each column at good quality: bare for none, light for one, dark for two or more, so a gap shows at a glance. A read counts where its base is at or above the confident quality (Confident from, in the Reads options); a read with no qualities counts everywhere it reaches, and a read of either strand counts.
• Samples that disagree. With two or more samples, a column where samples carry different bases from each other (not only from the document) is marked three ways: a solid foot under the strip's marks, a small triangle under the ruler and a faint tint down the column. The status row counts them ("3 columns where samples disagree"). Only bases at or above Confident from count (a read with no qualities counts everywhere), a deletion in one sample against a base in another is a disagreement, an ambiguity code such as N takes no part, and the empty cells a sample's insertion leaves in the others are ignored. Disagreement between reads is more often a base-calling error than a real change. A difference that every covering sample shares (they all read G where the document has A) is not marked: that agreement is what makes it credible, and the list notes it as "samples agree".
• Under the toolbar, when the reference is a document, a verification line sums up the features (the source feature and ORFs are left out) in one sentence: "All 27 features confirmed by all 5 reads", or "24 of 27 features confirmed by at least 3 reads, forward strand only". A feature is confirmed when every base of it is covered at good quality and nothing differs; the number of reads is the fewest over the bases of any confirmed feature. Only the features that need a look follow it, each clickable to bring it into view: "AmpR: 1 difference" (a poor-quality base does not count either way), "x: 17 of 20 bases covered" when only part is reached and nothing in that part differs, and "ori: not covered". Differences come first, then partly covered, then not covered; with more than five, the first four are named and and N more opens the table below with the problems first. The All features button, beside List in the toolbar, opens a table of every feature, one row each, between the line and the alignment (it scrolls inside itself and closes again from the same button; opening it closes the List of differences, and List closes it). Its columns are the Status with a mark (✓ Confirmed, ✕ and the number of differences, ~ Partly covered, ○ Not covered), the Feature and its Type, the Position in the document (1-based and inclusive, as the ruler numbers it; a feature through the origin of a circle reads, say, 6801–120; a join runs from its first part to its last), Reads (the fewest reads at good quality over its bases, 0 when some base has none), Strands (both, forward only, reverse only, or mixed when every base is read but neither strand reaches them all; empty when some base is not covered) and Bases covered ("17 of 20"). The rows are in document order; Problems first puts differences, then partly covered and not covered features above the confirmed ones. Click a row, or its name (a button, so Tab and Enter reach it), to bring the feature into view. Copy writes the table as tab-separated text, in the order shown. A feature across the origin of a circle counts each base once. These lines say what the alignment shows, not that the construct is right: a base no read covers is unchecked, not good.
• The toolbar has five groups. At the left, ‹ Prev and Next › (Alt+N, Alt+Shift+N, or whatever Next change is bound to in Format ▸ Keyboard shortcuts…) go from one difference to the next and wrap round at the ends, with a count beside them ("3 of 41"). A run of neighbouring differing columns (a 5-base gap, say) is one stop, and the whole run is marked. Beside the count, Filter chooses which differences they stop at (see below), List opens the table of differences described below and All features the table of features (one at a time), and Go to and Find open a small form under the toolbar (see below), Export opens another for saving or copying the alignment (see below), and, with more than one sample, Samples sorts and hides them (see below). At the right, the Show group switches Features, ORFs, Amino acids and Trace on and off; they stay as you left them when you open the window again. The forms under the toolbar (Filter, Go to, Find, Export, Samples) close when you click anywhere outside them, on Esc, or with × at the end of their heading; the List and All features tables stay open until you press their button again. Under the toolbar, the status row has, at the left, the picked sample's score line and buttons, and at the right the colour key.
• Filter (beside the count) narrows the differences ‹ Prev and Next › stop at, for when most of them are poor calls at the ends of reads. Where stops only at differences In a feature (any feature, a CDS included) or In a CDS or ORF, as the shading classes them; Good quality only only where a sample carrying the difference has its base there at or above Confident from (a read with no qualities, and a deletion, count as good); In the picked sample only only at the differences the picked sample carries (with none picked, every sample counts). They combine: with both of the last two, the picked sample's own base must be good. Skip reviewed (on unless you turn it off) passes over the differences marked reviewed or taken (see below). The count then reads "12 of 41 differences", and "3 of 12" on a stop; the button reads Filter (on) when the first three leave anything out. The filter only narrows the stops: the overview, the alignment and the List still show every difference. It is remembered between openings like the Show buttons.
• Acting on a difference. Once Next, Previous or the List has stopped at a difference, the status row shows two buttons for it. Reviewed marks it as looked at (a known poor call, say): it is drawn faded in the alignment and the overview, Next and Previous skip it (unless Skip reviewed is off in Filter), the List notes it "reviewed", and the verification no longer holds it against its feature but counts it apart ("AmpR confirmed by 2 reads, 1 reviewed"; "Confirmed (1 reviewed)" in the table). Press it again to take the mark off. The List has a Reviewed box on each row that does the same. Take _sample_'s bases writes the picked sample's bases over that difference into the document as one ordinary edit: Undo takes it back, it is marked as a change like any other, and features move with it. A gap in the sample deletes the document's bases there; bases the sample has and the document lacks are inserted. The bases are written in the document's case. The difference is then marked "taken into the document", faded and skipped like a reviewed one, and the bases written are selected in the document. You can take several differences one after another; the window keeps track of how each edit moved the positions after it. It refuses (and says why, on the button and beside it) when the sample does not reach the whole difference or matches the document there, when the difference runs through the origin of a circle, when the document is the read, and when the document has been changed in any other way since the last take (undo included): Align again to go on. The alignment itself is not redone, so it still shows the document as it was aligned. Marks last as long as the alignment result: closing the window and opening it again keeps them, and aligning again starts afresh. They are not saved with the document.
• Sort and hide samples. With more than one sample, Samples (in the toolbar, after Export) opens a small form. Sort by orders the rows: As aligned (the order they were given), Identity, highest first, Name (numbers in names count as numbers, so read2 comes before read10) or Position on the reference (where the alignment starts); equal ones keep their aligned order. Pick a sample and press Hide in the status row to take it out of the window; the button then reads "Samples (1 hidden)" and the status line says "1 hidden". The form lists the hidden samples under "Hidden: 1 of 4", each with Show, and Show all. The last sample shown cannot be hidden. The rows you see are the only rows everything else works on: ↑ and ↓ walk them in the order shown, the differences and their count, the list, the coverage band, the marks where samples disagree, Find, Export and the verification lines are all computed from the shown samples alone. A hidden sample therefore does not count towards a verification: with some hidden, a note above the lines says "Verification from 3 of 4 samples". A column that only a hidden sample's insertion opened is not in the picture either. Sort and hidden samples are for this opening of the window and are not remembered.
• Go to (Ctrl+G) takes a 1-based position, numbered as the ruler numbers the document, and scrolls to its column and marks it, however many columns another sample's insertion has opened before it. A position past the end of the document (or of a circular one) is refused with the last position; one in the document but outside the stretch the alignment covers (when the document's selection was aligned) says which positions are. When the document is the read, positions are those of the reference you pasted. Find (Ctrl+F, the editor's Find key, which here belongs to the window) looks for a motif, IUPAC codes allowed (GAATTC, RGATCY), at least three bases, on both strands, in the Reference or in the sample chosen in In (the sample you picked on the alignment; choosing another there picks it). A sample is searched on its own bases, so a motif is found across a deletion, and the marked columns include the gap they span; the reference is searched across another sample's insertion the same way. The form says "2 of 5"; ‹ and ›, Enter and Shift+Enter step through the matches, wrapping round, and each is marked and scrolled into view (a match on the reverse strand says so). A motif across a circle's origin is found in the reference only when a sample reads through the origin, which makes the reference run on. Esc closes the form first and the window on the second press; neither key reaches the editor underneath.
• Export takes the alignment out of the window for a notebook, a slide or an email. The form's Columns are those on screen when it opens (1-based alignment columns, the gap columns another sample opened included); type a range, or press On screen or All. Save SVG and Save PNG download a picture of those columns as the window draws them: the names, the ruler, the feature track, the reference and every sample with the differences shaded, the marks where samples disagree, and the amino-acid strips and chromatograms when they are switched on. The picture uses the window's colours and theme, and nothing picked or marked in the window. It is named from the document and the columns (pUC19-alignment-1-120.svg). A picture is drawn whole, so a very large one is refused with the reason: an SVG up to 100,000 bases (columns times rows, the document's row counted), a PNG up to 400,000 and about 16,000 pixels on a side (it is drawn at twice the size when that fits, for sharp text). Choose fewer columns, or copy the text. As text copies the columns in blocks of the width in in blocks of (60 by default, 10 at least): in each block the document's row and a row per sample, each with its name and the 1-based number of its first and last base in the block, then a match line under the document where | marks a column in which every sequence that reaches it has the same base (a gap agrees with a gap, case is ignored). The document's numbers are the ruler's; a sample's count its own bases along the stretch aligned, so a read's numbers start at 1 for its first aligned base, and are left blank where the sample has no base in a block. Names are cut to 24 characters. As aligned FASTA copies the same columns as one record per sequence, the document first, all the same length with - for a gap and for the columns a sample does not cover, 60 to a line, ready for any tool that reads an alignment. A file leaves the app as a download; copying is to the clipboard.
• Between the ruler and the document row runs a feature track: the document's features as bars, an arrow for the strand, a join tied by a line, and one across the origin of a circular document drawn where a read through the origin sees it. Read a difference against the bar above it to see whether it falls in a gene, a promoter or nothing; hover a bar for its name and type. The Features and ORFs buttons in the Show group (pressed means on) switch each on. ORFs are the ones the ORFs panel finds (at its minimum length), outlined and dashed in their own lanes under the features. A feature that an insertion in a sample splits round is drawn across the gap columns. Up to eight lanes are drawn; the score line counts what did not fit. When the document is the read and the reference is the sequence in the box, there is no track, since a pasted sequence brings none of its own.
• A sample that has features of its own (an open tab, a GenBank or SnapGene file, or the document when it is the read) gets them drawn in lanes under its row, marked Own features beside its name, with the same Features button. They are placed through the sample's alignment: each bar sits over the columns its bases aligned to, so a feature the sample carries at another place than the document does stands offset from the bar in the track above, and one the sample lacks has no bar under it. A sample that aligned reversed has its features turned with it, strand included. Only the part of a feature the sample aligned over is drawn; one wholly in a trimmed or unaligned stretch is left out. Hover a bar for its name; up to three lanes per sample, and the status line says how many of the picked sample's features did not fit.
• The Amino acids button (shown when the document has a CDS in the alignment; Alt+T in this window, as Translations is in the views) draws a strip of residues under the document row and under each sample: the document's own, then what the sample's bases make of the same codons, read with the feature's genetic code and strand. A residue that changed is shaded (another residue, a stop gained or lost); one the same letter from other bases is shaded faintly; a - marks a codon with a base missing or inserted inside it; a codon the sample does not reach is left empty. The codons are the document's, so a frameshift shows as - where it starts and as changed residues after it, not as a re-read frame.
• List opens a table of the differences, one row per run of differing columns, between the verification lines and the alignment (it is closed by default and remembered like the Show buttons). Its columns are the Position in the document (1-based, as the ruler numbers it; an insertion is put at the base before it, and on a circular document a read through the origin counts on from 1 again), the Change from the document's bases to the sample's (A→G, - for none, long runs cut to their ends and a length), the Samples that carry it (with several carrying different bases, each change is listed), the Feature it falls in (a CDS or ORF first, then any other feature, else none), the lowest read Quality over the differing columns (reads with qualities only), the Protein effect inside a CDS and a Note ("samples disagree" where the region has a column marked as above, "samples agree" where two or more samples carry the same change and none disagrees, else blank; "reviewed" or "taken into the document" is added for a marked difference), and a Reviewed box (see above). The effect is silent, a missense such as p.K42R, a nonsense p.K42*, a stop lost p.*42K, frameshift for an insertion or deletion that is not a multiple of three bases, or in-frame indel. Codons are read in the CDS's own frame, strand and genetic code, so a reverse-strand CDS is translated after reverse complementing. Outside a CDS the effect is blank, and where a region crosses more than one CDS each is named. Click a row, or press Enter on its position, to go to it as Next does and pick a sample that carries it. Copy puts the whole table on the clipboard as tab-separated text with a header line, ready for a lab notebook or a spreadsheet. With no document behind the reference there is no feature or effect.
• Click a name, or click into the alignment and press ↑ or ↓, to pick a sample. Its score and identity show under the toolbar, with Select in document beside them to select the region it aligned to.
• Esc closes the window and returns you to where you were.

An insertion in one sample opens gap columns in the document row and in every other sample, so the rows stay in register. This is each sample aligned to the document alone, laid side by side; it is not a multiple sequence alignment, and two samples are not aligned to each other. The A read's bases below the confident quality sit on a grey block, faded, so a difference there is easy to discount. A read opened from an AB1 file has its chromatogram drawn under its row, each base's peak under its letter and the base qualities as faint bars behind; press Trace in the window's toolbar (shown when any sample has one, on by default) to hide it and see more rows.