rnaseq-transcript-quantification

name	rnaseq-transcript-quantification
description	Alignment-free transcript-level quantification using Salmon or kallisto for fast and accurate RNA-seq expression estimates.

Skill: RNA-seq Transcript Quantification

Use When

• User wants fast transcript-level quantification without genome alignment
• User wants TPM and count estimates per transcript
• User needs input for downstream differential expression analysis (tximport/tximeta in R)
• User wants to compare Salmon vs kallisto quantification

Inputs

• Required:
  • Trimmed FASTQ file(s) (single-end or paired-end)
  • Transcriptome index (Salmon index directory or kallisto index file)
• Optional:
  • Tool choice: salmon or kallisto (default: salmon)
  • Library type for Salmon (default: A for auto-detect)
  • Number of threads (default: 4)
  • Output directory (default: ./quant_output)
  • Number of bootstrap samples (for downstream use with sleuth)

Workflow

1. Validate that the transcriptome index exists for the chosen tool.
2. If Salmon: run salmon quant with --validateMappings, --seqBias, and --gcBias for bias correction.
3. If kallisto: run kallisto quant.
4. If bootstraps are requested: add --numBootstraps (Salmon) or -b (kallisto).
5. Report: total reads processed, mapping rate, number of quantified transcripts.

Output Contract

• Quantification file:
  • Salmon: quant.sf with columns: Name, Length, EffectiveLength, TPM, NumReads
  • kallisto: abundance.tsv with columns: target_id, length, eff_length, est_counts, tpm
• Log file with mapping statistics
• Auxiliary files (Salmon aux_info/ directory or kallisto run_info.json)

Limits

• Salmon and kallisto must be installed and available on PATH.
• Transcriptome index must be pre-built before running this skill.
• Quantification is at the transcript level; use tximport in R to summarize to gene level for DE analysis.
• Not suitable for novel transcript or isoform discovery.
• For single-end reads with kallisto, fragment length and standard deviation must be provided.
• Common failure cases:
  • Transcriptome index built from a different annotation version than expected, skewing quantification.
  • Salmon library type auto-detection failing on very low-read-count samples.
  • kallisto single-end mode missing required --fragment-length and --sd parameters.