Skip to content

Supported profilers

Six profilers, each importable and executable. Import paths accept the formats below; execution (gmeta run -m …) shells out to the tool from the pixi environment.

kraken2 + bracken

  • Import: kreport (tree format) and/or bracken species table.
  • Execute: gmeta run -m kraken2 runs kraken2 then bracken; read length for bracken comes from the database's databaseXXXmers.kmer_distrib file.
  • kreport column order matters: percent | clade reads | direct reads | rank | taxid | indented name. gmeta pairs species to their genus via the name indentation — real genus clade counts.

centrifuge

  • Import: native report (name taxID taxRank genomeSize numReads numUniqueReads abundance) or any kraken-style kreport (e.g. from centrifuge-kreport).
  • Execute: gmeta run -m centrifuge classifies, converts with centrifuge-kreport, and reuses the kraken2 kreport path — the 识别网 export gets the same tree-exact genus pairing.

kaiju

  • Import: kaiju2table output (file percent reads taxon_id taxon_name). Pass the species table as the profile; --kaiju-genus adds real genus counts. Lineage names ("Bacteria; Escherichia; Escherichia coli") are parsed for genus hints.
  • Execute: gmeta run -m kaiju runs kaiju then kaiju2table for both species and genus ranks.

metaphlan

  • Import: profile with #clade_name NCBI_tax_id relative_abundance … rows; rel_ab_w_read_stats additionally fills 种序列数 (estimated reads).
  • Execute: gmeta run -m metaphlan runs metaphlan -t rel_ab_w_read_stats with --bowtie2db $DB_DIR/metaphlan.

kmcp

  • Import: kmcp profile 17-column TSV; requires the taxid/rank/ taxname columns (present when the database was built with a taxid map). Per-reference rows are aggregated by taxid.
  • Execute: gmeta run -m kmcp runs kmcp search (single-end mode for sensitivity, per the official recommendation) then kmcp profile --level species. Needs the NCBI taxdump under $DB_DIR/taxdump.
  • Note: KMCP percentages are per-reference; cross-check totals against the 审核明细 sheet.

sylph

  • Import: profile TSV (Genome_file, Adjusted_ANI, Taxonomic_abundance).
  • Execute: gmeta run -m sylph runs sketch + profile; confidence bands follow ANI (high ≥95%, medium ≥90%).

Format sniffing

net build auto-detects each input file; --profiler forces one. The sniffer distinguishes headers (kaiju, metaphlan, kmcp, bracken, sylph) from headerless layouts (centrifuge's name-led columns, kraken2's percent-led kreport).