Alignment-Free Quantification - Usage Guide

Overview

Salmon and kallisto perform transcript quantification directly from FASTQ files without traditional alignment, making them much faster than alignment-based methods.

Prerequisites

# Salmon
conda install -c bioconda salmon

# kallisto
conda install -c bioconda kallisto

Quick Start

Tell your AI agent what you want to do: - "Quantify my RNA-seq samples using Salmon" - "Build a Salmon index with decoy sequences" - "Run kallisto on my paired-end FASTQ files"

Example Prompts

Index Building

"Build a decoy-aware Salmon index from the human transcriptome and genome"

"Create a kallisto index from my transcripts.fa file"

Quantification

"Quantify paired-end reads for sample1 using Salmon with bias correction"

"Run Salmon quant on all my FASTQ files in batch mode"

Output Interpretation

"Explain the columns in the Salmon quant.sf output file"

"What's the difference between TPM and NumReads in Salmon output?"

What the Agent Will Do

1. Download or locate the reference transcriptome (Ensembl/GENCODE)
2. Build the index (optionally with decoy sequences for Salmon)
3. Run quantification on each sample with appropriate parameters
4. Check mapping rates and quality metrics
5. Organize output files for downstream analysis with tximport

Obtaining Transcriptomes

Ensembl (Recommended)

# Human
wget https://ftp.ensembl.org/pub/release-110/fasta/homo_sapiens/cdna/Homo_sapiens.GRCh38.cdna.all.fa.gz

# Mouse
wget https://ftp.ensembl.org/pub/release-110/fasta/mus_musculus/cdna/Mus_musculus.GRCm39.cdna.all.fa.gz

GENCODE

# Human
wget https://ftp.ebi.ac.uk/pub/databases/gencode/Gencode_human/release_44/gencode.v44.transcripts.fa.gz

Building Decoy-Aware Salmon Index

# Get genome and transcriptome
wget genome.fa.gz
wget transcripts.fa.gz
gunzip *.gz

# Extract chromosome names as decoys
grep "^>" genome.fa | cut -d " " -f 1 | sed 's/>//g' > decoys.txt

# Concatenate (transcriptome first!)
cat transcripts.fa genome.fa > gentrome.fa

# Build index (k=31 suits reads >=75 bp; lower to ~23-25 for ~50 bp reads)
salmon index -t gentrome.fa -d decoys.txt -i salmon_index -k 31 -p 8

Understanding Output

Salmon quant.sf

kallisto abundance.tsv

TPM vs Counts

• TPM - A within-sample proportion (sums to 1e6 in every sample). Good for ranking/visualization, but invalid for cross-sample differential expression because a few up-regulated genes inflate the fixed total and depress every other TPM (composition bias).
• Counts - Import the estimated counts (NumReads / est_counts), not TPM, into DESeq2/edgeR via tximport, which carries the average-transcript-length offset.

Why the decoy-aware index matters

A transcriptome-only index has no home for reads from introns, unannotated transcription, or intergenic DNA, so they get misassigned to whatever transcript shares sequence, inflating its count. The genome added as decoys lets the quantifier discard those reads instead. This is the single highest-impact accuracy choice; always use it when the genome is available.

Verifying library strandedness

• -l A auto-detects, but a library with weak strand signal can be miscalled. Always read lib_format_counts.json and confirm one consistent type across samples.
• dUTP / Illumina TruSeq Stranded / NEBNext Directional report as ISR (maps to featureCounts -s 2, reverse). Forward is ISF (-s 1); unstranded is IU (-s 0).

Tips

• Use the decoy-aware Salmon index for best accuracy; do not pass the deprecated no-op --validateMappings.
• Enable bias correction with --gcBias --seqBias; it matters most when library-prep batch is confounded with condition.
• Generate inferential replicates (--numGibbsSamples 20 for Salmon, -b 100 for kallisto) ONLY for transcript-level DTE/DTU downstream (swish, sleuth, catchSalmon); gene-level work does not need them.
• Check mapping rates - should be >70% for a matched reference.
• Match transcriptome version to the GTF annotation release.