Salmon skips genome alignment entirely. It quantifies transcript abundance directly from reads using a selective alignment approach, producing transcript-level counts significantly faster than BAM-based methods. GenXflo connects the full workflow on one canvas.
Salmon's workflow is structurally different from alignment-based RNA-seq. There is no HISAT2 or STAR step and no BAM file. Reads go directly to Salmon after trimming. FastQC, BBDuk, Salmon, and MultiQC are all pre-configured in the GenXflo library.
Raw read quality check
Adapter trimming
Pseudoalignment and quantification
Aggregated QC report
No BAM file is produced. Salmon quantifies directly from reads using a transcript index.
FeatureCounts and StringTie require a sorted BAM file from HISAT2 or STAR. They count reads that have been explicitly aligned to the genome. Salmon uses a different model: it maps reads directly to a transcript index using a quasi-mapping approach, estimates transcript abundances using an expectation-maximisation algorithm, and produces TPM and count estimates without ever creating a BAM file.
This makes Salmon substantially faster and less storage-intensive for large cohorts. The trade-off is that you get transcript-level quantification without the genome-aligned BAM that other downstream analyses might need. If your goal is differential expression at the gene or transcript level and you do not need the BAM for any other purpose, Salmon is the more efficient choice. If you need the BAM for variant calling, peak calling, or visualisation in IGV, an alignment-based pipeline is required.
Use Salmon when your goal is differential expression analysis at the gene or transcript level and you do not need a genome-aligned BAM for any other purpose. Salmon is faster and produces less intermediate data. Use HISAT2 plus FeatureCounts when you need the BAM file for variant calling, visualisation in IGV, or any other downstream step that requires genome-level alignment.
Yes. Salmon quant.sf files are compatible with the tximeta and tximport R packages, which import Salmon output into a format DESeq2 and edgeR read natively. This approach also propagates transcript-to-gene mapping and uncertainty estimates into the differential expression analysis.
Yes. The exported Nextflow DSL2 code runs on local machines, HPC clusters, and cloud platforms including AWS, Azure, and Google Cloud.