Same sequencing sample in multiple lanes. How to analyse it?

Question

[Federico]

Hi,

I have the following samples:

Lane 1:

• 184631_S1_L001_trimmed_R1.fastq
• 184631_S1_L001_trimmed_R2.fastq

Lane 2:

• 184631_S1_L001_trimmed_R1.fastq
• 184631_S1_L001_trimmed_R2.fastq

I would like to align these fastq files to the reference library. Bowtie2 is the tool I am using. As you can see, the same sample was spread across 2 lanes in order to obtain adequate sequencing depth. The expected output is a count matrix that will be further analysed using DESeq2.

How (and when) to proper treat and merge the same sample from two different lanes?

[kashiff007]

Are they technical replicates? If not, you can merge them before mapping with the simple cat command.

[Federico]

They are not. The shown samples are from the same multiplex, and the multiplex was then sequenced on 2 lanes. What are you suggesting is to merge them before the alignment? R1.fastq from both lanes and same for R2.fastq?

[kashiff007]

Yes, merge all forward and reverse read files separately.

cat Lane1/184631_S1_L001_trimmed_R1.fastq Lane2/184631_S1_L001_trimmed_R1.fastq > final_R1.fastq
cat Lane1/184631_S1_L001_trimmed_R2.fastq Lane2/184631_S1_L001_trimmed_R2.fastq > final_R2.fastq

[swbarnes2]

If they same from the same library prep, they absolutely are technical replicates, no matter how many different lanes they are split over.

[Federico]

Thanks a lot for your suggestion!

Answer 1

[Istvan Albert]

You can merge the files before and after alignment.

Merging the FASTQ files perhaps makes life a little simpler in that fewer files need to be managed.

Merging after alignments gives you a bit of a better understanding of whether the two lanes exhibit different characteristics during alignment. You also get somewhat simpler parallel processing when generating alignments (in the sense that the files are already split).

[ektavpathak]

Hi Istvan! Would it be fine to add the counts of samples (same sample from different lanes) from count matrix instead of merging aligned files? Logically it should not make any difference.

Best Ekta

[Istvan Albert]

yes, I think adding up the counts is also a valid approach

[ST]

I guess this could really depend on the "counting" strategy, the library complexity and possible lane biases. If e.g., EM is used to guide multi-mapping read assignment based on robust evidence from uniquely mapped ones, any biases/differences in unique mappers in BAMs from individual lanes would guide the assignment of multimappers differently in each count file.