Output from featureCounts() as input to DESeq2

Question

[Elizabeth Sam]

I am new to RNA-seq. I plan to find out the differentially expressed genes from two samples. For that I first downloaded the fastq files and aligned the reads using align(). Later, the gene level expression values were summarized as integer number using featureCounts() . Can I give this as input to DeSeq2.? If yes, then what steps to follow.

[EVR]

Typical RNA-seq data analysis is as follows:

Align the fastq reads onto to Genome(using Hisat2) or Transcriptome(using bowtie2) which will give you output in sam/bam format.

Extract the raw counts for every transcript/gene in every replicate from every sample

Later this count data must provided into DEseq2 for further down stream analysis(see the manual)

If you are totally new to RNA-seq analysis, then kinldy make use of Trinity

[WouterDeCoster]

The work flow OP wants to use is fine, no need for trinity.

Answer 1

[WouterDeCoster]

If you (in R) run the following you can easily transfer the count data to deseq2:

counts <- featureCounts(bams, blablabla_restofcommand)$counts
deseqdata <- DESeqDataSetFromMatrix(countData=counts, colData=sampleInfo, design=~condition)

You obviously need to fill in some more parts or adjust colData and design to your experiment.

[Elizabeth Sam]

Thanks a lot WouterDeCoster.

[Elizabeth Sam]

I have one more doubt. I have two text files as output from featureCounts(), one of normal and other tumor. To find the differentially expressed genes, I plan to merge the two files into one count file and give this as input to DESeq2. Is this approach correct?

[ShirleyDai]

Not appropriate to merge tumor and normal data. You can use this info as a covarite in the design.

[WouterDeCoster]

Did I understood correctly that you are looking for differentially expressed genes between the normal and tumor samples?

[Elizabeth Sam]

Sorry for the late reply. Yes, I want to find the differentially expressed genes between normal and tumor samples.

[WouterDeCoster]

As in my code example above, the counts object will hold all counts generated from the files in the bams object. So it's perfectly fine to have both the normal and tumor samples in there together. You are not merging the data, you are putting it together in one dataframe/object. The comment of ShirleyDai wasn't accurate.

[Elizabeth Sam]

Thanks WouterDeCoster.

[ShirleyDai]

What I mean taking tumor/normal info as a covariate is to make a multi-factorial model. Source: https://support.bioconductor.org/p/58893/

[WouterDeCoster]

Right, that would be design in case of a paired analysis.