Manual Dashboard Review

Use this guide when the right decision depends on human judgment: a suspicious contig needs visual context, a breakpoint should be staged from a plot, or a curated FASTA needs an explicit recipe.

The key question is:

What should I decide in the browser, and what should I reproduce from the command line?

The Core Idea

chromo manual is a review surface, not a hidden automatic fixer. It creates a self-contained HTML dashboard where you can inspect dot plots, optional graph context, and optional chromo eval review events beside the contigs or junctions they describe.

The dashboard can export a FASTA directly from the browser, but the more reproducible path is to export a recipe JSON and apply it with chromo manual apply. That turns browser choices into an audited command-line FASTA and optional TSV report.

What The Dashboard Reads Or Writes

chromo manual reads:

reference FASTA,
assembly FASTA,
one coords or PAF file for that exact FASTA pair,
optional GFA graph context,
optional long-read PAF or GAF evidence,
optional chromo eval review table for task-specific queues.

It writes:

one self-contained HTML dashboard,
a browser FASTA download when sequences are available,
a browser recipe JSON download,
a reproducible FASTA and report when the recipe is applied with chromo manual apply.

The dashboard dot plots still come from the original coords or PAF file. After manual export or manual apply, re-align the manual FASTA before using it as the assembly input to chromo sort, chromo plot, chromo scaffold, or other alignment-dependent steps.

Dashboard Modes

Mode	Use it when…	Extra review focus
`chromo manual`	You want free-form browsing and curation.	Contig list, dot plots, graph context, breakpoints, removals, inversions, scaffold labels.
`chromo manual fix`	You have a `fix_review.tsv` from `chromo eval fix`.	Candidate split-piece queue and breakpoint evidence panels.
`chromo manual scaffold`	You have a `scaffold_review.tsv` from `chromo eval scaffold`.	Adjacent scaffold-junction queue and gap/overlap evidence.
`chromo manual gapfill`	You have a `gapfill_review.tsv` from `chromo eval gapfill`.	Graph fill queue and path-support evidence.

Task-specific manual dashboards do not replace the reviewed executor. They make review easier, then the accepted table or exported recipe still has to be applied explicitly.

What To Inspect First

ChromoSort manual dashboard with contig graph filters, selected-contig dot plot, and graph-neighborhood panel. — **Figure 1. Manual review surface.** The dashboard keeps the contig list, dot-plot evidence, optional graph context, and selected-event details together so browser edits can be made with the evidence visible.

Start with the left-to-right review path:

Confirm the dashboard was generated from the intended reference, assembly, and coords or PAF file.
Select a contig or review event.
Compare the dot plot to the assignment, breakpoint, scaffold, or gapfill table row.
Check optional graph, GAF, or long-read evidence panels when available.
Stage only the edits that match the evidence.
Export a recipe JSON for reproducibility.
Apply the recipe from the command line and save the report beside the FASTA.

Common Manual Actions

Action	Good use	Watch out for
Remove or restore a contig	Exclude unplaced, redundant, or clearly unwanted records.	Do not remove possible haplotypes or real duplications just because they overlap a reference span.
Move a contig or piece	Correct a reviewed order within a scaffold group.	Re-check the output with fresh alignment after export.
Invert a piece	Make a curated output match the reviewed orientation goal.	Whole-contig reverse orientation is often a normal sort option; internal inversions need evidence.
Add a breakpoint	Split a visually reviewed region.	Browser-staged coordinates should be reproduced with a recipe and audited in the apply report.
Label scaffolds	Group active pieces for scaffold export.	Use clear labels that match downstream naming conventions.
Remove unaligned	Quickly exclude records without current placement evidence.	Confirm whether unaligned sequence is expected, novel, or a name/alignment problem.

Free-Form Review Workflow

Generate a dashboard:

chromo manual \
  --ref-fasta reference.fa \
  --assembly-fasta assembly.fa \
  --paf paf/sample.paf \
  --gfa assembly_graph.gfa \
  --output-html review/sample.manual.html \
  --suggested-output-fasta sample.manual.fa

For large genomes, the HTML embeds metadata but not full assembly sequence by default. Open the dashboard, load the original assembly FASTA with the browser file picker, then export the FASTA or recipe. For small demos, add --embed-sequences.

Apply the recipe:

chromo manual apply \
  --assembly-fasta assembly.fa \
  --recipe chromosort.manual.recipe.json \
  --output-fasta results/sample.manual.fa \
  --report results/sample.manual.tsv

Then re-align:

results/sample.manual.fa
  -> minimap2 or MUMmer
  -> validation plots and downstream sorting

Task-Specific Review Workflow

Use this pattern when an algorithm should propose rows and a human should review the outliers:

chromo eval fix \
  --assembly-fasta assembly.fa \
  --paf paf/sample.paf \
  --all \
  --gfa assembly_graph.gfa \
  --read-paf reads_to_assembly.paf \
  --gaf reads_to_graph.gaf \
  --output-prefix review/sample.eval_fix

chromo manual fix \
  --ref-fasta reference.fa \
  --assembly-fasta assembly.fa \
  --paf paf/sample.paf \
  --review-table review/sample.eval_fix.fix_review.tsv \
  --gfa assembly_graph.gfa \
  --read-paf reads_to_assembly.paf \
  --gaf reads_to_graph.gaf \
  --output-html review/sample.manual_fix.html

The same structure works for manual scaffold and manual gapfill with the matching review tables.

Recipe And Report Audit

Read the manual apply report before trusting the FASTA:

Field	Why it matters
`piece_id`	Identifies the reviewed browser piece.
`source`	Records the original assembly contig.
`output_name`	Shows the FASTA record emitted by apply.
`scaffold`	Shows the browser scaffold label used for export.
`slice_start` and `slice_end`	Audit the source interval.
`strand`	Shows whether the piece was inverted.
`removed`	Confirms records intentionally excluded from output.
`export_mode`	Distinguishes piece output from scaffolded output.

Keep the HTML, recipe JSON, apply report, and manual FASTA together. They are the provenance bundle for the reviewed edit.

Cheat Sheet

If you need…	Use…
Visual review without sequence changes	`chromo manual` dashboard only.
A reproducible manual FASTA	Export recipe JSON, then `chromo manual apply`.
A spreadsheet queue in the browser	`chromo eval`, then task-specific `chromo manual`.
A browser FASTA for quick inspection	Browser export, then still re-align before downstream use.
Evidence panels for GFA, GAF, or reads	Provide those inputs when generating the dashboard.

Common Traps

Do not forget that the dashboard plot uses old alignment rows. Manual FASTA exports need fresh alignments.

Do not rely only on a browser download for provenance. Export and apply a recipe when the edit needs to be reproducible.

Do not assume a task-specific manual dashboard applies accepted rows. The matching executor still has to run.

Do not edit around a stale review table. Regenerate eval tables after changing the FASTA, assignments, graph, path-search settings, or primary alignment.

Do not treat graph and read panels as automatic sequence-changing authority. They are evidence for review.

What To Look At Next In ChromoSort

Use Spreadsheet Review Tables for chromo eval accept/reject workflows.
Use Chimeric Contig And Breakpoint Review before staging split edits.
Use Inversions And Orientation Changes before inverting or normalizing internal orientation changes.
Use chromo manual for the full command reference.