bio-isoform-switching

What it does

Identifies shifts in which transcript a gene uses between conditions (DTU) and predicts downstream functional consequences, including potential NMD, ORF disruption, protein-domain changes, signal peptide alterations, and IDR changes. Distinguishes DTU from DGE and DTE and supports integration with external annotators to annotate switches. The workflow can output a consolidated set of isoform switches and their predicted functional impacts.

How it works

The skill centers on IsoformSwitchAnalyzeR and a DTU-focused pipeline:

• Import transcript-level expression (e.g., Salmon) and build a design matrix for conditions.
• Run a DTU test via either satuRn (preferred when replicates exceed 5) or DEXSeq, with a default rule: satuRn if any condition has >5 replicates; else DEXSeq.
• Pre-filter genes/isoforms based on expression cutoffs, retaining genes with multiple detectable isoforms.
• Perform a DTU test to identify switching genes, then optionally run a two-stage DTU workflow (DRIMSeq/DEXSeq + stageR) for proper OFDR control and to pinpoint which transcript drives the switch.
• Annotate functional consequences by extracting sequences and running external tools (CPC2, Pfam, SignalP, IUPred2A or DeepTMHMM) outside R, then re-import results.
• Compute predicted consequences via analyzeSwitchConsequences, including NMD_status and various structural/functional annotations (domains_identified, IDR_identified, signal_peptide_identified, etc.).
• Include an NMD prediction step based on the 50-nt rule, with caveats about last-exon PTCs and tissue-specific effects.
• Provide guidance for long-read inputs, uncertainty-aware DTE testing with swish, and how to interpret DTU in the context of functional impact.

When to use it

Use when investigating how splicing differences alter protein function or trigger NMD-mediated degradation, and when needing integrated functional annotations for DTU switches. The workflow is appropriate for standard DTU interpretation, with long-read input support and uncertainty-aware DTE analysis as options.

What it can touch

The workflow touches: IsoformSwitchAnalyzeR (R package usage), external annotation tools (CPC2, Pfam, SignalP, IUPred2A/DeepTMHMM), and outputs from Salmon/tximport/tximeta. It uses a pre-filter configuration (geneExpressionCutoff, isoformExpressionCutoff, IFcutoff) and requires running external tools on sequences, then re-importing results.

Caveats

• External annotation steps must be executed outside R; results must be re-imported for consequence analysis.
• NMD predictions are probabilistic; some predicted transcripts may escape NMD in practice.
• For very large cohorts (>200 samples) or boundary cases (exactly 5 replicates), results may differ depending on the chosen DTU method (satuRn vs DEXSeq).
• The 50-nt NMD rule has known exceptions (e.g., last-exon PTCs escape NMD; tissue-specific NMD factors).
• The workflow emphasizes checking whether an alternative form bears a PTC before interpreting directionality.