bio-long-read-splicing

What it does

The skill instructs an agent to analyze alternative splicing using long-read RNA-seq data from PacBio Iso-Seq and ONT platforms, focusing on full-isoform resolution. It guides the agent to employ a multi-tool workflow including FLAIR (correct/collapse/quantify/diffSplice), IsoQuant (de-novo or annotation-guided isoform discovery), Bambu (annotation-aware discovery + quantification), SQANTI3/SQANTI-LR (isoform classification with artifact flags), rMATS-long (event calling on long-read isoforms), and minimap2 with platform-specific presets. It also addresses microexon detection, recursive splicing, complex multi-exon isoforms, and DTU without transcript-quantification uncertainty. It provides explicit commands for long-read splice-aware alignment, then a stepwise analysis pipeline for correction, collapse, quantification, and differential testing, plus discovery and classification workflows. It includes explicit example commands for FLAIR workflows, IsoQuant discovery, Bambu quantification, SQANTI3 QC, rMATS-long preprocessing and execution, and DTU analysis with DRIMSeq/DEXSeq/stageR. It documents platform-specific minimap2 parameters and when to use -uf, and it notes reference versions compatibility checks and required memory. It also delineates use-case triggers (when long-read resolution is needed due to short-read limitations).

How it works

The skill structures a concrete, step-by-step workflow:

• Align reads with minimap2 using platform-specific presets (splice:hq for HiFi; -k14 for ONT; and -uf for direct RNA or stranded libraries).
• Run a FLAIR workflow: flair correct (with short-read junctions if available), flair collapse, flair quantify, and flair diffSplice to obtain per-event PSI and statistics.
• Use IsoQuant in a de novo or annotation-guided mode to discover and quantify full-length isoforms, producing a GTF and counts.
• Apply Bambu via R to jointly discover and quantify isoforms with a calibrated Novel Discovery Rate (NDR) parameter, returning SE objects with counts.
• Classify isoforms with SQANTI3 (and SQANTI-LR for long-read data), generating classifications like FSM/ISM/NIC/NNC and artifact flags, and filter accordingly.
• Perform rMATS-long for differential isoform analysis on long-read data, including preprocessing steps (organizing gene info by chromosome, simplifying alignment info, organizing by gene/chr, detecting splicing events, and generating a final rmats_long.py run).
• Describe downstream differential and DTU analysis: DRIMSeq + DEXSeq + stageR for DTU on long-read counts derived from Flair or Bambu.
• Provide single-cell long-read guidance: combining 10X short reads with PacBio MAS-Iso-seq via FLAMES or skera-based barcode rescue.
• Include platform notes and version compatibility checks for tools and dependencies, with explicit commands and options.

The instructions are explicit about the exact tool names, script invocations, options, and file naming conventions, and they preserve the order of steps to achieve end-to-end analysis.

When to use it

Use when short-read AS limitations (anchor-length issues for microexons, complex multi-exon isoforms, recursive splicing, transcript ambiguity) necessitate full-isoform resolution from long reads. Applicable across: bulk Iso-Seq discovery, bulk ONT cDNA/RNA analysis, end-to-end differential analyses, joint discovery + quantification, de novo discovery for non-model organisms, event-level differential splicing on long reads, DTU on long-read counts, and single-cell long-read splicing workflows.

What it can touch

The skill references and prescribes the following tools and inputs: FLAIR, IsoQuant, Bambu, SQANTI3, SQANTI-LR, rMATS-long, minimap2, samtools, DRIMSeq, DEXSeq, stageR, FLAMES, skera, limas/isoseq3 pipelines, and references/annotation files (reference.fa, gencode.v45.annotation.gtf). It requires mixed tool types and specifies commands and parameters for each.

Caveats

The skill notes version compatibility checks for Python, R, and CLI tools and emphasizes verifying signatures and flags before usage. It describes memory requirements (e.g., IsoQuant memory >= 64 GB for atlas-scale runs) and cautions about platform-specific alignment flags (the -uf flag semantics) and the difference between splice:hq and splice presets. It does not make performance guarantees and confines claims to described workflows, parameters, and intended use cases. License is NOASSERTION for the skill folder; no explicit licensing for each tool beyond that context is stated here.