Why so much variability in RNA-seq technical replicates?

Question

[jaqx008]

Hi all, I download an RNA seq data for gene expression analysis from NCBI. When I did counts with bedtools for my gene of interest, the variability was too much. Does that mean this data is unreliable? What could have caused this? and Does that mean I cant discard the bad data and use the ones I feel are closer from the same research group?

See part of data as example before normalization.

     Female_rep1    Female_rep2 Male_rep1   Male_rep2
gene1   149 11  125 30
gene2   108 122 388 68
gene3   18  30  393 44
gene4   170 91  1270    179
gene5   86  3   176 2
gene     254    311 898 215

Thanks

[GenoMax]

You must mean biological replicates. Technical reps would be the same library sequenced multiple times. Almost no one does that since it is not needed for Illumina sequencing

[jaqx008]

Would you advice I continue working with this data? based on the difference in gene expression from the above replicates?

[swbarnes2]

You know that bedtools is really not the software of choice for generating gene counts?

[jaqx008]

I have always used the multicov option. Could you emntion the one you think is best?

[swbarnes2]

Most people use FeatureCounts or RSEM

[Trivas]

I've used bedtools coverage before for gene expression analysis, but it's important to know that these are raw counts and not normalized by sequencing depth (FPM/TPM).

I would look into making a correlation matrix or heat map or PCA plot to see how your samples cluster before moving forward with any differential gene expression analysis.

[jaqx008]

Ok. thanks for your recommendation.