different replicate numbers

[pandapiggie]

Dear all,

I would like to know if it is possible to do a differential expression analysis on samples that have different replicate numbers?

I recently received my data using illumina hi-seq for bacteria samples. I had 3 replicates per sample but when I did the alignment on bowtie2 1 sample had a replicate with only 49% alignment rate.

Does anyone have suggestion on how to proceed? I am thinking of just doing a 3 replicate sample A vs 2 replicate sample B. Is this acceptable or should I adjust it to 2 replicates per sample?

Thank you for your answer.

Jason

[ddb]

Dear all,

I would like to know if it is possible to do a differential expression analysis on samples that have different replicate numbers?

Yes it is possible to do this.

I recently received my data using illumina hi-seq for bacteria samples. I had 3 replicates per sample but when I did the alignment on bowtie2 1 sample had a replicate with only 49% alignment rate.

I would not automatically discard this sample based on the mapping rate but would investigate the cause of the low mapping first. Possible causes are adapter sequences left on the reads preventing the mapping or sample contamination, ribosomal contamination etc. Did you do any preprocessing of the reads? You might find FastQC and FastQ Screen useful for investigating this. If you captured the unmapped reads do a blast search of some to see what hits you get.

Does anyone have suggestion on how to proceed? I am thinking of just doing a 3 replicate sample A vs 2 replicate sample B. Is this acceptable or should I adjust it to 2 replicates per sample?

Do not adjust to 2 replicates per sample. Keep as much data as possible.

[GenoMax]

@pandapiggie you may want to use bowtie v.1 (or turn off splice detection in other aligners) to see if that improves alignment (since this is bacterial data).

There is always the possibility that that one replicate may be bad. Have you scanned/trimmed these sequences for presence of adapter?

[pandapiggie]

@pandapiggie you may want to use bowtie v.1 (or turn off splice detection in other aligners) to see if that improves alignment (since this is bacterial data).

There is always the possibility that that one replicate may be bad. Have you scanned/trimmed these sequences for presence of adapter?

Dear GenoMax,

Thanks for the suggestion. May I ask if I were to use bowtie v1 for that sample do I need to redo the alignment for all the other samples?

Sorry for the beginner question.

[pandapiggie]

Sorry another question pop up in this case.

Which pipeline would be more accurate in this case? I was thinking using cufflinks but from what I understand DESeq2 is good for my case as there is uneven number of replicates between samples.

Can anyone help me?

Thanks