Full-Length Isoform Analysis - Usage Guide

Overview

This skill discovers, classifies, filters, and quantifies full-length transcript isoforms from PacBio Iso-Seq/Kinnex (HiFi) and Oxford Nanopore (cDNA and direct-RNA) long reads. The central principle is that a long-read novel isoform is an artifact until proven otherwise: RT template-switching, intra-priming on genomic poly-A, and 5' RNA degradation manufacture novel junctions and truncated isoforms, so the SQANTI3/pigeon classification plus artifact filter plus orthogonal end/junction support (CAGE for 5' TSS, poly-A atlas for 3' TES, short-read STAR junctions) is the analysis, not a postscript. It covers the PacBio isoseq+pigeon pipeline (including the Kinnex skera-split step), SQANTI3 for any platform, and the ONT tools (IsoQuant, FLAIR, Bambu, StringTie2, FLAMES). Differential isoform usage is handed off to the alternative-splicing category.

Prerequisites

# PacBio path
conda install -c bioconda isoseq pbmm2 pigeon lima
# Cross-platform curation
conda install -c bioconda sqanti3
# ONT path
conda install -c bioconda isoquant flair bambu stringtie pychopper minimap2
# Reference annotation + genome, plus orthogonal support: CAGE refTSS BED, poly-A motif list,
# short-read STAR SJ.out.tab for junction validation

Quick Start

Tell your AI agent what you want to do: - "Run the PacBio Iso-Seq pipeline and classify the isoforms with pigeon" - "Deconcatenate my Kinnex data and build a filtered isoform catalog" - "Discover and quantify isoforms from my ONT cDNA with IsoQuant" - "Tell me which of my novel isoforms are trustworthy"

Example Prompts

PacBio Iso-Seq / Kinnex

"I have Kinnex HiFi reads. Run skera, lima, refine, collapse, then classify and filter with pigeon using my CAGE peaks and poly-A motifs, and report the FSM/ISM ratio and saturation."

ONT discovery and quantification

"These are ONT direct-RNA reads. Spliced-align them correctly, discover and quantify isoforms with IsoQuant, then curate with SQANTI3 and short-read junctions."

Artifact triage

"My catalog has 40% novel isoforms. Tell me whether that is real biology or an artifact problem, and which classes (ISM, NNC, mono-exon) to distrust."

Validation

"Validate my novel isoforms against CAGE 5' peaks, poly-A sites, and short-read junctions, and drop the intra-priming and RT-switching artifacts."

What the Agent Will Do

Identify the platform and chemistry (PacBio Iso-Seq/Kinnex vs ONT cDNA vs ONT direct-RNA) and choose the pipeline.
For Kinnex, deconcatenate with skera first; for ONT cDNA, orient with pychopper.
Build the isoform set (isoseq collapse, or IsoQuant/FLAIR/Bambu for ONT).
Classify against the reference (pigeon or SQANTI3) and apply the artifact filter with CAGE/poly-A/short-read support.
Report the FSM/ISM ratio, novel-class breakdown, and a saturation curve excluding singletons.
Quantify with an EM tool and hand differential isoform usage to alternative-splicing.

Tips

A high novel-isoform fraction is a red flag, not a success - it usually means an under-powered filter or degraded RNA.
ISM fraction is an RNA-integrity thermometer; do not report ISMs as novel without CAGE 5' support.
NIC (novel combination of known junctions) is far more trustworthy than NNC (novel splice site); mono-exon novels are the false-discovery sink.
The Iso-Seq binary is isoseq (renamed from isoseq3 in v4); pigeon is the classifier/filter, not a quantifier, and it consumes the collapsed.sorted.gff after pigeon prepare, not a BAM.
Kinnex/MAS-seq data must be deconcatenated with skera split before lima.
Match the minimap2 preset to chemistry; -uf is for stranded direct-RNA/Iso-Seq, not unoriented ONT cDNA.
Long-read isoform quantification needs EM (Bambu/IsoQuant/NanoCount); raw FLNC counts are not short-read-equivalent abundances, and discovery is depth-unsaturated.

long-read-alignment - Spliced alignment presets for cDNA/direct-RNA
basecalling - Direct-RNA chemistry and cDNA vs direct-RNA tradeoffs
alternative-splicing/long-read-splicing - Long-read splicing (define the boundary)
alternative-splicing/isoform-switching - Differential isoform usage downstream
rna-quantification/tximport-workflow - Transcript-level quantification downstream
genome-annotation/eukaryotic-gene-prediction - Isoforms as annotation evidence