Quantification Output Parsing

quantification-output-parsing

Summary

Parse and import transcript-level abundance, count, and length estimates from RNA-seq quantification tools (salmon, sailfish, kallisto) into structured matrices suitable for downstream statistical analysis. This skill bridges pseudoalignment/quasi-mapping quantifiers and gene/transcript-level differential expression analysis.

When to use

You have transcript-level quantification output files (e.g., quant.sf from salmon, abundance.tsv from kallisto) and need to convert them into abundance, count, and effective length matrices for gene-level or transcript-level differential expression analysis in R packages such as DESeq2, edgeR, or limma-voom.

When NOT to use

• Input quantification files are from alignment-based methods (e.g., STAR, HISAT2, Bowtie2 BAM files); use a dedicated BAM-to-count tool instead.
• You already have pre-computed gene-level count matrices; tximport is unnecessary.
• Your downstream analysis requires exon-level or junction-level estimates; tximport aggregates to transcript or gene level only.

Inputs

• quantification output files (quant.sf, abundance.tsv, or equivalent from salmon/sailfish/kallisto)
• tx2gene data.frame (transcript ID to gene ID mapping)
• sample metadata or file paths linking samples to quantification outputs

Outputs

• abundance matrix (transcript or gene level, depending on txOut parameter)
• counts matrix (transcript or gene level)
• length matrix (average transcript length, weighted by sample-specific abundance)

How to apply

Use tximport with the type argument set to the quantifier name (salmon, sailfish, or kallisto) and specify txOut parameter based on your analysis goal: set txOut=TRUE to retain transcript-level matrices, or txOut=FALSE for direct gene-level summarization. Provide a tx2gene data.frame mapping transcript IDs to gene IDs; tximport will then read all quantification files and return a list containing 'abundance', 'counts', and 'length' matrices. The length matrix represents average transcript length weighted by sample-specific transcript abundance estimates and should be used as an offset in downstream count-based analysis. For faster file reading, enable the readr package dependency. If performing transcript-level analysis, verify downstream tools can accept the txOut=TRUE output structure; if performing gene-level analysis, confirm that tx2gene mapping is complete and unambiguous.

Related tools

• tximport (Core function to read quantification files, parse transcript-level estimates, and summarize to gene or transcript level) — https://github.com/thelovelab/tximport
• readr (Optional dependency to accelerate file reading in tximport)
• tximeta (Extended package wrapping tximport, adds automatic annotation metadata)
• salmon (Upstream quantifier producing transcript-level abundance and count estimates compatible with tximport)
• DESeq2 (Downstream package for gene-level differential expression analysis using tximport count matrices and length offsets)
• edgeR (Downstream package for gene-level differential expression analysis using tximport count matrices)

Examples

txi <- tximport(files=c('sample1/quant.sf', 'sample2/quant.sf'), type='salmon', tx2gene=tx2gene, txOut=FALSE)

Evaluation signals

• Verify that the returned abundance, counts, and length matrices have dimensions (transcripts/genes) × (samples) with no missing values for expected features.
• Confirm gene-level matrices from txOut=TRUE followed by summarizeToGene are identical (via all.equal() in R) to those produced by txOut=FALSE on the same input files.
• Check that the length matrix is numeric, strictly positive, and reflects weighted average transcript lengths (should vary across samples if differential isoform usage is present).
• Validate that tx2gene mapping covers all transcripts in the quantification files and produces no unexpected many-to-many relationships.
• Ensure downstream offset calculation (e.g., log(length)) produces finite values and does not introduce systematic bias in downstream differential analysis.

Limitations

• tximport requires a complete and accurate tx2gene mapping; incomplete or misaligned mappings will silently drop transcripts or produce incorrect gene-level summaries.
• The length matrix corrects for changes in gene length due to differential isoform usage, but does not account for other sources of length bias (e.g., gc content, sequence composition); users should consider additional offset matrices or methods for those confounders.
• Transcript-level analysis requires downstream tools that accept txOut=TRUE output structure (list of matrices); standard DESeq2 / edgeR workflows assume gene-level input and may require adaptation.
• tximport does not validate quantification file integrity or detect corrupted / incomplete quant.sf files; upstream quantifier errors will propagate into downstream analysis.

Evidence

• [readme] Imports and core functionality: "Imports transcript-level abundance, estimated counts and transcript lengths, and summarizes into matrices for use with downstream statistical analysis packages such as edgeR, DESeq2, limma-voom."
• [readme] Length matrix as offset: "Average transcript length, weighted by sample-specific transcript abundance estimates, is provided as a matrix which can be used as an offset for different expression of gene-level counts."
• [intro] Correction for differential isoform usage: "this approach corrects for potential changes in gene length across samples (e.g. from differential isoform usage)"
• [readme] Supported quantifiers: "tximport as of version 1.3.9 will import inferential replicates (Gibbs samples or bootstrap samples) from Salmon, Sailfish or kallisto."
• [intro] Gene-level summarization workflow: "Transcripts need to be associated with gene IDs for gene-level summarization. If that information is present in the files, we can skip this step... We first make a data.frame called tx2gene with two"
• [readme] Transcript-level output option: "The argument txOut=TRUE can be used to generate transcript-level matrices."
• [intro] readr performance benefit: "While tximport works without any dependencies, it is significantly faster to read in files using the readr package"