Support a coordinate-to-accession map to a graph, mapping an annotation label+coordinate to an accession by hmusta · Pull Request #549 · ratschlab/metagraph

hmusta · 2025-10-08T14:55:50Z

An extension to automatically transform query results from column-based to accession/header-based labels

Usage:

1. Construction

After constructing a coordinate-aware annotation, run another annotation step with flags --anno-filename --index-header-coords to index the headers and construct the CoordToHeader mapping:

metagraph annotate --anno-filename --index-header-coords -i graph.dbg -o annotation file_1.fa file_2.fa ...

2. Query

When the CoordToHeader (.seqs) index is constructed (e.g., annotation.row_diff_brwt_coord.annodbg + annotation.seqs), the mapping is done automatically for the usual query command, e.g.,

metagraph query -i graph.dbg -o annotation.row_diff_brwt_coord.annodbg --query-mode matches query.fa

To disable the mapping and report the original file-based coordinates/counts/labels, pass flag --no-coord-mapping.

metagraph/src/graph/graph_extensions/row_tuples_to_id.cpp

karasikov · 2025-10-25T13:53:28Z

It would be good to have some Usage commands, how to run this.
(If not in the docs, at least in the PR description above.)

metagraph/src/graph/graph_extensions/row_tuples_to_id.hpp

metagraph/src/graph/graph_extensions/row_tuples_to_id.cpp

… of the same coords

karasikov · 2026-01-02T17:44:52Z

Stats are printed in this format:

[2026-01-02 18:40:56.487] [info] Statistics for coord-to-accession mapping 'refseq.seqs'
=============== COORD-TO-ACCESSION STATS ===============
columns: 85375
total sequences: 32881371
total k-mers: 1703657822378
=================== PER-COLUMN STATS ===================
column 0:
  sequences: 215264 (NZ_LPZT01000038.1  NZ_LNTR01000002.1       NZ_LOHN01000001.1       ...)
  k-mers: 5816602042
  k-mers per sequence: 27020.8
column 1:
  sequences: 493377 (NZ_LXGF01000074.1  NZ_LXGF01000102.1       NZ_LXGF01000110.1       ...)
  k-mers: 20332752987
  k-mers per sequence: 41211.4
column 2:
  sequences: 53868 (NZ_UCIF01000001.1   NZ_UCIF01000002.1       NZ_UCIF01000003.1       ...)
  k-mers: 4857890014
  k-mers per sequence: 90181.4
column 3:
  sequences: 6480 (NG_042103.1  NC_027945.1     NW_004948952.1  ...)
  k-mers: 3942536088
  k-mers per sequence: 608416.1
column 4:
  sequences: 15416 (NW_020939779.1      NW_020939780.1  NW_020939781.1  ...)
  k-mers: 3486134738
  k-mers per sequence: 226137.4
column 5:
  sequences: 10230 (NT_107461.3 NT_107462.3     NT_107463.4     ...)
  k-mers: 10818493389
  k-mers per sequence: 1057526.2
...

adamant-pwn

Overall looks good. Some comments so far.

metagraph/src/annotation/coord_to_header.hpp

metagraph/src/annotation/coord_to_accession.hpp

metagraph/src/annotation/coord_to_accession.cpp

metagraph/src/graph/annotated_dbg.cpp

metagraph/src/cli/annotate.cpp

metagraph/integration_tests/test_query.py

…the flags in CLI

adamant-pwn

Looks good to me, except one misaligned (?) indentation (see comment).

metagraph/src/cli/load/load_annotated_graph.cpp

…ts-sum queries

…inate to an accession (#549) * use k-mer coordinates to decode accession IDs when labels represent chunks/files * print warning if no CoordToHeader mapping (.seqs file) was found for coordinate annotations * print num accessions in server stats when the mapping is active * added coord-to-header tests * hide query mode `labels` (labels without label counts) by default; shown with flag --advanced --------- Co-authored-by: Harun Mustafa <hmusta@users.noreply.github.com> Co-authored-by: Mikhail Karasikov <karasikov@users.noreply.github.com>

hmusta added 5 commits October 7, 2025 23:16

use coordinates to decode accession IDs when labels represent chunks

6a28828

fixes

3323bfc

fix loading

89b2a29

fixes

a8e089b

fix

819d6e6