nfcore/rnaseq vs nfcore/differentialabundance
nfcore/rnaseq is used for pre-processing, specifically generating a count table from the FASTQ files. nfcore/differentialabundance is then used for the differential expression/abundance analysis, that is comparing expression of genes across different conditions.
Count file contains floating point numbers
The count file is generated by nfcore/rnaseq using the default path (aligner: STAR, quantification by: Salmon). Salmon “infers” (aggregates) gene counts (actually count estiamtes) from transcript-level count (estimates) and takes gene lengths into account (the gene lengths might differ across samples since not all isoforms could be expressed in a given samples). All in all the resulting count matrix contains floating point numbers as these are just estimates. One can simply round() if feeding the count table to DESeq2, for example nfcore/differentialabundance rounds the count matrix.
DESeq2 normalization
DESeq2 uses median of ratios for normalization.