Marker coverage
A missing marker row is meaningful only in light of what the built database could search.
Check what was searchable
giftag infoOpen markers.tsv in the built giftag database to inspect one row for every gifter marker. Its columns are namespace, accession, name, source, status, profile and detail.
| Status | Meaning |
|---|---|
searchable |
A suitable profile and acceptance rule were available in this build. |
no_threshold |
The profile exists, but its source publishes no usable cutoff for it. |
missing_from_source |
The requested profile was absent from the selected source release. |
unsupported_namespace |
No supported sequence-profile source serves this marker namespace, such as EC. |
source_skipped |
The source was deliberately omitted with --sources. |
An unsearchable marker produces no annotation row and is indistinguishable from an absent marker to gifter’s evaluator. The ledger is the evidence needed to distinguish those cases when reading an unsupported GIFT. giftag_run.json also records the counts for the database used in that run. A positive gifter call still requires a complete curated route or machinery implementation; neither a marker hit nor a coverage count creates one.
Source rules
Each collection has its own accepted-hit rule. giftag records the profile, rule, score and threshold in every output row instead of combining scores across collections.
| Collection | Default release | Acceptance rule |
|---|---|---|
| KOfam | GenomeNet current, or an archive date supplied with --kofam-release |
Per-KO adaptive threshold from ko_list, applied to the full or best-domain score as in KofamScan. |
| NCBIfam / TIGRFAM | hmm_PGAP/20.0 |
Trusted cutoff. |
| Pfam | InterPro current | Gathering cutoff. |
| dbCAN families and dbCAN-sub | db_v5-2-9_5-5-2026 |
run_dbcan-style i-Evalue < 1e-15, HMM coverage > 0.35 and overlap resolution. |
The full dbCAN family library and every dbCAN-sub cluster of each requested family are searched, including competitors gifter does not itself use. Otherwise a weaker requested profile could win a region against a stronger unsearched profile, giving an overly specific marker. The default searches dbCAN-sub on every protein. --gate-subfamilies reduces work but can drop valid subfamily hits.
Source terms
giftag’s code is MIT-0 and its build redistributes no profiles. giftag build downloads profiles from their publishers onto your machine; their terms apply. The built giftag.json records the terms associated with the releases used.
- KOfam / KEGG: academic use is free; non-academic use, or an academic service to others, needs a licence from Pathway Solutions. See KEGG legal terms. Use
--sources ncbifam,pfam,dbcanif you need to omit KOfam, and read the resultingsource_skippedstatuses before interpreting absences. - dbCAN: the Open Data release states no restrictions on use; cite dbCAN3 and CAZy.
- NCBIfam: NCBI-built models are public domain; TIGRFAM-origin models are CC BY-SA 4.0.
- Pfam: CC0.
Validation and limits
The following is a 2026-10-04 snapshot, not a guarantee for other source or gifter database versions. It used Escherichia coli K-12 MG1655 and Bacteroides thetaiotaomicron VPI-5482, gene-called by giftag against gifter database 2026.35.1.
| Check | Result |
|---|---|
| Gifter markers searchable | 1,569 of 1,574 |
| dbCAN family calls versus run_dbcan on B. thetaiotaomicron | 458 of 458 gene-family pairs identical |
| dbCAN-sub calls versus run_dbcan on the kept clusters, using the same Z | 157 of 157 identical |
dbCAN-sub calls with --gate-subfamilies |
144 of 157 |
| Annotation speed on eight cores | About 45 seconds per genome |
The five unsearchable markers were three KOs without a KOfam threshold (K19098, K19099, K24159) and two EC numbers. This snapshot did not compare KOfam with KofamScan, and its dbCAN-sub comparison used only the clusters giftag keeps. The later benchmark covers both against the full libraries on eight genomes. Thresholds can also miss a real enzyme; for example, E. coli PyrC scores below KOfam’s conservative K01465 cutoff. These are reasons to examine search coverage and evidence before interpreting an unsupported capability as biological absence.
For build mechanics, overlap handling and the full validation rationale, see Methods and validation. For the comparison with KofamScan, run_dbcan, HMMER and InterProScan, see Benchmark.