PRForm guide

Run PRForm and interpret the results

PRForm scores every nucleotide in a viral FASTA record for the three calibrated classes: no PRF (0), -1 PRF, and +1 PRF.

Back to PRForm

Quick start

Paste a FASTA record or select a FASTA file.
For a reproducible first run, use one of the three test-data buttons.
Choose options, then select Run.
Use the ranked candidates, Genome Track, and downloads to inspect the result.

The test buttons load the full Influenza A genome (all segments), the full SARS-CoV-2 genome, and a Nanovirus no-PRF control.

Input and options

Top-K peaks: Enter a positive whole number or use the up/down arrows. There is no fixed 10-peak limit. Requests exceeding the longest input sequence are reduced to its number of positions. Candidates may still be classified as no PRF; larger counts increase processing time and download size.
Genome FASTA: One or more nucleotide records in FASTA format. Each record has its own result card. The original sequence remains available independently of Prokka.
Trusted annotation GFF: Optional curated annotation. When supplied, it is retained instead of running Prokka.
Run Prokka annotation: Requests annotation. If annotation fails or produces no usable features, completed predictions remain available and the results page shows an annotation warning with a diagnostic log.
Score both strands: Scores the input and reverse-complement orientations, retaining the winning orientation while keeping positions in the submitted genome coordinates.
Generate 6-frame translation: Shows open reading frames and stop codons in all six reading frames.

Candidate calls and scores

-1 or +1: For a model-supplied call, this is the model’s predicted frameshift class. For an annotation-supplied call, the direction comes from the annotation and may differ from the model’s highest-probability class.
0 / no PRF: The candidate was classified as no frameshift. A summary saying there are no PRF calls among returned candidates applies to those candidates only, not every position in the sequence.
Candidate order and coordinates: Model candidates are ranked by P(-1) + P(+1), while the model class is the largest of P(0), P(-1), and P(+1). Annotation-supplied candidates use the annotation’s coordinate order. When the source is unspecified, the interface preserves the returned order without claiming a model ranking. Positions start at 1 on the submitted sequence, including reverse-strand calls.
Maximum per-position P(PRF): For current results, the largest P(-1) + P(+1) at a single position in the model tracks. This is not a separately calibrated probability that the whole sequence contains a frameshift. Older results show “Legacy sequence probability” when that meaning is known; otherwise they show “Stored score (meaning unspecified)”.
Class probabilities: P(0), P(-1), and P(+1) describe the model tracks at the selected position. P(PRF) is the sum of the two available frameshift class probabilities. Unavailable values are shown as n/a, distinct from a returned probability of zero. Missing plot values appear as gaps. Annotation confidence scores are not substituted for these probabilities. The View note distinguishes annotation strand from model-track strand when they differ.

Genome Track

Annotation rows show stranded coding features from the supplied or generated GFF on separate + and − rows. Unstranded and noncoding features are omitted from this compact genome view.
Frameshift products use green for the annotated/original frame and the shared sequence before the slip site, blue after a -1 shift, and orange after a +1 shift. These are translation previews at the selected position and shift. Moving the marker or changing the shift does not change the returned candidate call or probabilities.
PRF probability per nucleotide shows separate P(-1 PRF) and P(+1 PRF) bar tracks. At close zoom, bars show individual positions. When zoomed out, each bar shows the highest class probability in a group of positions; the two tracks can reach their maxima at different positions.
6-frame ORFs and stops appears when six-frame output was requested and is available for the selected record. Its loading and availability states are separate from annotation and original-sequence access.
Sequence and product availability are separate. A product panel without usable coding context does not mean the original nucleotide sequence is missing.
Use Zoom in or the mouse wheel. At close zoom, aligned DNA triplets and translated amino acids appear directly in the upper Frameshift products rows.

Downloads

New jobs use the formats below. Previously saved downloads retain their original format.

Candidates (CSV) contains model-ranked or annotation-supplied positions, types, and separate minus1_PRF_prob and plus1_PRF_prob model values. Add the two available probabilities at a position to obtain P(PRF); there is no separate sequence-probability column.
Probability tracks (HDF5) contains full per-nucleotide arrays named no_PRF_prob, minus1_PRF_prob, and plus1_PRF_prob.
Annotation GFF, when present, is the annotation used in the Genome Track.
Annotation log records diagnostics when Prokka was requested and is also included in the job ZIP.
Six-frame translations (FASTA) contains the generated translations. ORF and stop-codon context is shown separately in the Genome Track.
Job files (ZIP) bundles the files generated for that job.