Skip to content

Run frameshift-sensitive VS marker screening

Frameshift screening searches the masked assembly with the shipped VS###### marker profiles. It can recover domains with a frameshift or in-frame stop codon that protein prediction can miss. ViroSync validates each domain against the viral marker database before it can seed a candidate region. The normal boundary and acceptance filters still apply.

The screen runs inside ViroSync with the standard Pixi environment. Native alignment uses up to --threads-per-worker processes per genome.

Configure the screen

The screen is off by default. Enable it for one run:

pixi run virosync \
  -i assembly.fna \
  -o results/frameshift \
  --config config/orchestration.yaml \
  --frameshift-screening

To enable it in a config file, set:

phase1:
  frameshift_screening_enabled: true

Use --no-frameshift-screening to disable the screen even when the config enables it.

Run the shipped example

The example uses three Trichomonas vaginalis G3 contigs from GCA_000002825.3.

pixi run example-frameshift
cat results/example-frameshift/batch_summary.tsv

With resource bundle v1.2.0, the row must report status=success, predictions=1, and accepted=1. The example checks the Phase 1 rescue path: it must write confirmed frameshift markers and rescued proteins. Its rescued markers are hits of the kinase model VS000087, which validate but never start a region, so no rescued protein reaches Phase 3. It is not a sensitivity benchmark.

Read the output

Each genome writes frameshift files under phase1/frameshift_screening/:

  • confirmed_frameshift_markers.tsv lists validated domains that can seed regions.
  • confirmed_frameshift_proteins.faa contains their amino-acid domain sequences.
  • frameshift_hits.tsv lists candidate domains and their alignment evidence.
  • frameshift_events.tsv records candidate event positions as zero-based, half-open genomic intervals. An interval marks the affected codon span; it is not an experimentally determined mutation site.
  • validation/validated_marker_hits.tsv records the evidence and validation result for each candidate domain.

Accepted calls are in the standard virosync_predictions.tsv and GFF3 outputs. An accepted per-EVE FAA can also contain a confirmed rescued domain.

Interpret the result

A confirmed rescued domain supports a disrupted viral marker locus. Its sequence is an aligned domain, not a repaired full-length protein or coding sequence. X marks an uncertain residue at a frameshift, stop codon, or ambiguous DNA. Event annotations cannot distinguish biological decay, programmed frameshifting, introns, and assembly errors.

The screen uses only the shipped VS profiles. Its six-frame seed search can miss loci whose conserved fragments are too weak to seed an alignment. Native alignment scores are not calibrated probabilities. Protein HMM E-values describe a peptide selected by the alignment procedure; they are not calibrated genome-search E-values. Viral-reference validation and coverage filters reduce unsupported candidates but do not establish a false-discovery rate. See the method for scoring and limits.