Deseqdataset Construction From Quantification

DESeqDataSet Construction from Quantification

Summary

Convert transcript-level quantification files (from Salmon, kallisto, or RSEM) into a DESeqDataSet object suitable for differential expression testing. This skill bridges quantification tools and DESeq2 by aggregating transcript abundance to gene level and wrapping the result in a metadata-aware R object.

When to use

You have transcript-level quantification files (quant.sf, kallisto abundance.h5, or RSEM output) from one or more RNA-seq samples and need to construct a count matrix for differential expression analysis. Use this when you want un-normalized estimated gene-level counts paired with sample metadata and a statistical design formula.

When NOT to use

• You already have a gene-level count matrix from HTSeq, featureCounts, or similar; use DESeqDataSetFromMatrix() instead.
• You have quantification data imported via tximeta (which produces a SummarizedExperiment); construct DESeqDataSet directly from the SummarizedExperiment with design formula.
• Your input files are alignment BAM files rather than quantification files; use featureCounts or HTSeq to produce counts first.

Inputs

• transcript-level quantification files (Salmon quant.sf.gz, kallisto abundance.h5, or RSEM .genes.results)
• tx2gene mapping table (data frame with columns: transcript ID, gene ID)
• sample metadata table (data frame with sample names as row names, columns for condition/covariates)

Outputs

• DESeqDataSet object with count matrix, sample metadata (colData), and design formula
• assay matrix containing un-normalized estimated gene-level counts

How to apply

First, load quantification file paths and sample metadata (with condition labels and run identifiers), then construct or load a tx2gene mapping table linking transcript IDs to gene IDs. Pass quantification files and the mapping to tximport() with type='salmon' (or 'kallisto'/'rsem') to aggregate transcript abundance estimates to gene level, producing a list containing estimated counts, lengths, and offsets. Finally, construct a DESeqDataSet using DESeqDataSetFromTximport(), specifying the tximport list, sample metadata table (colData), and design formula (e.g., ~condition). Verify the result by examining the assay matrix to confirm values are un-normalized integer estimates without library-size scaling.

Related tools

• tximport (Imports transcript abundance quantifications and aggregates to gene level using tx2gene mapping)
• DESeq2 (Provides DESeqDataSetFromTximport() constructor and differential expression testing framework) — https://github.com/thelovelab/DESeq2
• Salmon (Upstream quantification tool producing transcript-level abundance estimates (quant.sf files))
• kallisto (Alternative upstream quantification tool producing transcript abundance estimates)
• RSEM (Alternative upstream quantification tool producing transcript abundance estimates)
• tximportData (Provides example Salmon quantification files and tx2gene mapping for tutorial workflows)

Examples

txi <- tximport(files, type="salmon", tx2gene=tx2gene); ddsTxi <- DESeqDataSetFromTximport(txi, colData = samples, design = ~ condition)

Evaluation signals

• The resulting DESeqDataSet contains a count assay with integer (or quasi-integer) values without library-size normalization or log transformation.
• Row names of the count matrix match gene IDs from the tx2gene mapping; dimensions are (genes × samples).
• colData(dds) contains all sample metadata columns and matches the order of quantification files.
• The design formula is correctly stored and includes condition or covariate variables present in colData.
• Calling assay(dds) reveals un-scaled estimated counts aggregated across all transcripts per gene.

Limitations

• tximport produces estimated counts based on transcript-level abundance; these are not raw read counts and should not be treated as such downstream.
• The tx2gene mapping must be accurate and complete; missing or duplicate mappings will result in missing genes or gene aggregation errors.
• Design formula must reference columns in the sample metadata table; misnamed variables will cause construction or analysis to fail.
• Pre-filtering of low-count genes is recommended after DESeqDataSet construction to reduce memory use and improve statistical power, but is not performed automatically.

Evidence

• [other] tximport ingests transcript-level quantification files and produces a list object that contains un-normalized counts aggregated to the gene level, which serves as input to DESeqDataSetFromTximport for constructing a DESeqDataSet.: "tximport ingests transcript-level quantification files and produces a list object that contains un-normalized counts aggregated to the gene level, which serves as input to DESeqDataSetFromTximport"
• [other] you could import the data with tximport, which produces a list, and then you can use DESeqDataSetFromTximport(): "you could import the data with tximport, which produces a list, and then you can use DESeqDataSetFromTximport()"
• [other] If you have performed transcript quantification (with Salmon, kallisto, RSEM, etc.) you could import the data with tximport: "If you have performed transcript quantification (with Salmon, kallisto, RSEM, etc.) you could import the data with tximport"
• [other] ddsTxi <- DESeqDataSetFromTximport(txi, colData = samples, design = ~ condition): "ddsTxi <- DESeqDataSetFromTximport(txi, colData = samples, design = ~ condition)"
• [other] txi <- tximport(files, type="salmon", tx2gene=tx2gene): "txi <- tximport(files, type="salmon", tx2gene=tx2gene)"