docs(star-align): ultra-detailed comments on the window model, extension and stitching - #16
Open
BenjaminDEMAILLE wants to merge 3 commits into
Open
docs(star-align): ultra-detailed comments on the window model, extension and stitching#16BenjaminDEMAILLE wants to merge 3 commits into
BenjaminDEMAILLE wants to merge 3 commits into
Conversation
BenjaminDEMAILLE
changed the title
…ndow model Explain the part of the aligner that decides everything else: what a genomic window is and why grouping seeds before stitching is what makes the problem tractable (stitching is quadratic, so 50 seeds per window many times beats every seed at once, which is what closed D12), why only low-multiplicity anchors may open a window while repetitive seeds may only join one, why a window may never straddle a chromosome boundary in the concatenated coordinate space, why the flanks are padded, what the diagonal test in the overlap dedup actually means, and why eviction drops the shortest non-anchor seeds rather than the oldest. Also the Rust behind it: loop labels, enums as closed sets the compiler forces you to handle, Drop as cleanup that cannot be forgotten on any exit path, and why the tie-break comparison must never return Equal. Comments only; no code, no behavior, no output bytes changed. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Explain the two operations every alignment is built from. extendAlign is local alignment without a matrix: walk base by base, remember the best-scoring prefix, which is what lets a mismatch be crossed when enough matches follow while a run that never recovers is simply excluded; the mismatch budget grows with the length already aligned and is capped in absolute terms. stitchWindowAligns is a subset search over the window's seeds, explored as a binary include/exclude tree, with a failed stitch returning a large negative score so the whole subtree is pruned. That pruning plus seedPerWindowNmax is what keeps an exponential search space affordable, and the fixed seed order is what makes the winner among equals reproducible. stitchAlignToTranscript makes one decision, spelled out: the relationship between the read gap and the genome gap is what the gap is (mismatch, deletion, intron with a motif check, insertion, or the unsequenced middle of a fragment). Comments only; no code, no behavior, no output bytes changed. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The exclude-branch guard reads as if STAR's WA_Anchor==2 force-include were live. It is not: WlastAnchor is an unsigned counter initialised to -1, so no align is ever marked, as the note in stitch_pieces already records against an instrumented STAR build. Say so where the guard is, and say why it is still ported. Window span said 'a few hundred kilobases'; with the default 64 kb bins, 9 bins of anchor distance and 4 bins of flank each side it reaches ~1.1 Mb. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
BenjaminDEMAILLE
force-pushed
the
docs/comments-align
branch
from
August 5, 2026 10:23
863f197 to
fdb9580
Compare
Contributor
Author
|
Self-review of the stack turned up two wrong claims in this PR; both are fixed in the follow-up commit.
|
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Stack 4/5, on top of #15. Comments only; no code, no behavior, no output bytes changed.
extendAlignas local alignment without a matrix, and the include/exclude tree institchWindowAlignswith its pruning.stitchAlignToTranscriptmakes: read gap vs genome gap is what the gap is.Gate green (
scripts/check.sh).