NMD & Transcript Fate
Predicting Nonsense-Mediated Decay susceptibility and truncation severity.
Nonsense-mediated decay (NMD) is a surveillance pathway that degrades mRNAs whose stop codon is recognised as premature — typically one sitting well upstream of the last exon–exon junction. SPLISOFORMS predicts each isoform's NMD susceptibility from the established 50-nucleotide rule, and separately tracks how truncated a surviving isoform is relative to its canonical counterpart.
The 50-Nucleotide Rule
An isoform is flagged as an NMD target when it carries a premature termination codon (PTC) — a stop lying more than 50 nt upstream of the final exon–exon junction (Nagy & Maquat, 1998). Two well-established exceptions let a transcript escape decay even with such an upstream stop:
- 5′-proximal escape: the stop codon lies within 150 nt of the start codon, close enough for the ribosome to reinitiate.
- Single-exon transcripts: with no exon–exon junctions, there is nothing to mark the stop as premature.
Status Values
Every isoform receives one of three calls. Two come directly from the 50-nt rule; the third, Unknown, is used whenever the rule cannot be applied.
NMD Target
A premature stop codon lands more than 50 nt upstream of the last exon–exon junction, outside any escape zone.
Transcript likely degraded — protein not produced.
NMD Escape
The stop sits in the last exon, within 50 nt of the last junction, or in the 5′-proximal zone — or the transcript is single-exon, with no junction to trigger decay.
Transcript likely translated.
Unknown
The 50-nt rule couldn't be applied — the exon or CDS coordinates it needs are missing, or the ORF is too short for the rule to be valid.
No call is made, rather than a guess.
When information is missing, the call is Unknown
The 50-nt rule needs both the transcript's exon structure and its coding-sequence (CDS) coordinates. When those aren't available — most often for novel isoforms whose ORF hasn't been resolved yet — the isoform is left Unknown rather than guessed.
This currently affects the breast-cancer (Head et al. 2026) dataset: its TransDecoder ORF/CDS predictions are not yet loaded, so those isoforms remain Unknown until that data lands. Short translated ORFs (smORF / uORF candidates) are also routed to Unknown, because the 50-nt rule was never validated for them — see below.
Truncation Severity
Protein-length divergence vs. the canonical sequence is reported astruncation_severity. A short isoform may have a legitimate alternative TSS without triggering NMD, whereas a full-length isoform could still harbor a PTC.
≥ 90% of canonical length.
70–90% of canonical length (e.g. an alternative transcription start).
< 70% of canonical length — suggests a premature stop or a major exon skip.
Microproteins & uORF Candidates
The 50-nt rule was developed for canonical protein-coding transcripts, not for short translated ORFs — applying it to them would produce misleading calls. Such short ORFs are increasingly recognised as a genuine layer of the proteome — the expanding catalogue of microproteins and peptideins. SPLISOFORMS flags these candidates separately and routes them to Unknown.
Isoform < 100 aa and canonical < 150 aa. May encode bona fide microproteins or peptideins.
Isoform < 50 aa with a long canonical (≥ 150 aa). Heuristic flag for NMD-resistant upstream ORFs.