Repository navigation
Interpreting results
mashID reports, for each sample, the reference whose sketch is best contained in the sample, plus
notes. This page shows what typical situations look like in summary_mashID.tsv and the per-sample
table, and what to do about them. Identity values below are for a database with sketch size 1000
or more and k=21; exact numbers vary with the database.
Sample Identity Shared_Hashes Median_Multiplicity Est_Depth Identification Note
S1 0.9997 994/1000 122 95.3 Mycobacterium bovis
Identity above 0.99, nearly all hashes shared, a depth consistent with the run, and no note. The per-sample table shows the same organism in the first rows; other organisms, if any, at identities far below the top hit. This is a species-level identification you can act on.
Sample Identity Shared_Hashes Median_Multiplicity Identification Note
S2 0.9995 995/1000 60 Escherichia coli Possible mixture with: Staphylococcus aureus
Per-sample table:
Rank Identity Shared_Hashes Median_Multiplicity Identification
1 0.9995 995/1000 60 Escherichia coli
2 0.9991 992/1000 9 Staphylococcus aureus
Two organisms both have almost all of their hashes present. With winner-take-all, shared k-mers go to
the best reference, so the second organism can only score this high with k-mers of its own: it is
in the sample. The multiplicities give the relative abundance, here roughly six to one. Whether this
is a contaminated culture, a polymicrobial sample or index hopping is for you to decide; mashID only
says both are there. -s multiplicity ranks by abundance instead of identity when the dominant
organism matters more than the best match.
Note: Possible mixture with: Mycobacteroides abscessus
Identification: Mycobacterium abscessus subsp. abscessus
If the "mixture" partner is the same species under another genus name, the database lists the
organism under synonyms and mashID counts them as different. Run make_mashID_db --check on the
database: it lists species that appear under several genera. Harmonise the names with a
--metadata table or rebuild with TaxIDs from NCBI (--assembly-report).
Identification: Mycobacterium tuberculosis Shared_Hashes: 943/947
A shared-hash denominator below the database's sketch size, or a reference length far below a genome
in the per-sample Description, means the top hit is a fragment: a 1 kb record is entirely contained
in any related sample. mashID ignores references shorter than --min-ref-length (100 kb) when the
database has a metadata sidecar, which it creates on first use, so this should not happen unless
--min-ref-length 0 was given. make_mashID_db --check lists such records.
Identity Identification Note
0.9862 Mycobacterium intracellulare Ambiguous: Mycobacterium paraintracellulare at identity 0.9851
The best hits of two organisms are within --ambiguity-margin of each other and neither clears the
mixture threshold: the database cannot separate them for this sample. Mash resolves species, and
these two are at the species boundary. Report the complex or group, or use a method with more
resolution (ANI against complete genomes, MLST, SNP typing).
MTBC members differ by well under 0.1% of their genome, and Mash identities among them differ in the
fourth decimal. Which member comes first is decided by which reference happens to share one more hash;
Identification will alternate between M. tuberculosis and M. tuberculosis variant bovis from
sample to sample. Treat the result as "MTBC" and assign the variant or lineage with a SNP-based tool.
Build the <database>.confusable.tsv sidecar (make_mashID_db --confusable) and mashID says so in
the output instead of leaving it to be remembered:
Identity 0.99989
Identification Mycobacterium tuberculosis variant bovis
Note Not separable in this database from: Mycobacterium tuberculosis, Mycobacterium
tuberculosis variant africanum, Mycobacterium mungi and 3 more
Do not read the runner-up's identity as the margin: under winner-take-all it is not one. See Outputs. The same applies to Bacillus cereus group members, Escherichia coli / Shigella, and other groups defined below the species level.
Identity Shared_Hashes Identification Note
0.9312 218/1000 Mycobacterium sp. XY-12
Identity between 0.90 and 0.97 with a small fraction of shared hashes means the sample is related to
the top hit but is not it: a species absent from the database, or a database for the wrong group.
Screen against a broad database (progenomes4 or refseq_bacteria) to find the genus, then a
dedicated one if it exists. Below 0.90 nothing is reported at all:
Identification: No significant hit in database
Check that the input is what you think it is (host reads, adapters only, an empty file) and try the broad databases.
Median_Multiplicity Est_Depth Note
2 2.4 Low coverage: median multiplicity 2
The identification can still be right, but with two observations per k-mer a contaminant at a few
percent is invisible and the identity estimate is noisy. Est_Depth says how deep the sample is if
it is the reported organism; a large gap between depth and multiplicity (many bases, low
multiplicity) suggests that much of the sample is something else, such as host DNA.
Screening a bacterial assembly or read set against refseq_viral reports every RefSeq virus whose
sketch is contained in the sample. A complete prophage shows as a phage at identity close to 1 with
nearly all hashes shared; a partial or diverged prophage shows at 0.9 to 0.97 with a fraction of the
hashes. Several phages in one genome each get a line in the per-sample table. The Identification
column of the summary is the best-contained virus, not "what the sample is".
| Observation | Read it as |
|---|---|
| Identity ≥ 0.99, no note | Species-level identification. |
Possible mixture with: X |
X is present too; check multiplicities for proportions. |
Ambiguous: X |
Database cannot separate the two; report the group. |
| Identity 0.90–0.97, few shared hashes | Related organism not in the database. |
No significant hit |
Not in this database, or not the expected kind of input. |
Low coverage |
Plausible but weakly supported; sequence deeper for contaminant detection. |