Skip to content

FAQs

Jump to bottom
mxwang66 edited this page Apr 20, 2026 · 1 revision

Edge cases

All genomes (both targets and non-targets) are identical

No signature will be found.

The log will show:

- expected k-mer absence in targets: 0.00000
- expected k-mer presence in non-targets: 1.00000

This means every k-mer found in the targets is also expected to be present in the non-targets, so Seqwin cannot identify any sequence unique to the targets.

All genomes (both targets and non-targets) share no k-mer

No signature will be found.

The log will show:

- expected k-mer absence in targets: 1.00000
- expected k-mer presence in non-targets: 0.00000

This means Seqwin cannot identify a region that is consistently present in the targets.

Targets are identical, and non-targets share no k-mer with the targets

The entire target genome will be reported as the signature.

The log will show:

- expected k-mer absence in targets: 0.00000
- expected k-mer presence in non-targets: 0.00000

In this case, the target genome is perfectly conserved within the target group and completely absent from the non-target group.

How to evaluate signature specificity

In signatures.csv, two columns are especially relevant for specificity: divergence and f_neg_hits.

These values are computed by aligning each signature against all non-target genomes with BLAST. If a genome has multiple hits of the signature, only the best one is reported.

For example, suppose there are 1000 non-target genomes. If a signature has a BLAST hit in 300 of them and no BLAST hit in the remaining 700:

  • f_neg_hits = 300 / 1000 = 0.3
  • divergence = D / (1000 * L), where D is the number of mismatches and gaps in the 300 genomes with a BLAST hit, L is the length of the signature.

This means that divergence should never be interpreted by itself.

For example, a low divergence value can occur in two very different situations:

  1. The signature has a BLAST hit in only a few non-target genomes (low f_neg_hits)
  2. The signature has few mismatches / gaps in most non-target genomes (high f_neg_hits)

These two cases have opposite biological meanings, so divergence must always be interpreted together with f_neg_hits.

When Seqwin may output no signature

We have not yet run a systematic simulation to define the exact range of sequence similarity for which Seqwin can or cannot recover signatures.

In practice, Seqwin may fail to find a signature in two common situations:

  1. The targets are too similar to the non-targets

    Example: carbapenem-resistant vs. carbapenem-susceptible P. aeruginosa.

  2. The targets are too divergent from one another

    Example: grouping P. aeruginosa and S. aureus together as a single target set.

In addition, Seqwin is designed to identify a single region that distinguishes the target group from the non-target group. It does not currently identify combinations of multiple regions that are only informative when considered together.

For example:

  • region A is present in 60% of target genomes
  • region B is present in 40% of target genomes
  • neither region is present in any non-target genome

In this situation, neither region A nor region B alone would be considered a signature by Seqwin, even though the combination of A and B could still distinguish the target group from the non-target group.

A preliminary solution to this limitation has already been developed and is planned for a future major release of Seqwin.

Clone this wiki locally