How to align the same samples that were sequenced in multiple flow cells and lanes

Question

[Matina]

Hi,

I have a set of FASTQ files that I want to align to the reference genome. The sequencing for each sample has been done on 2 different runs (flow cells) and 2 different lanes so for each sample I have 4 files. I am not sure when I should merge my files, before or after alignment? I read previous posts that suggest to merge the samples after alignment, but I am not sure what is the best in my case. Could I merge the samples using samtools? Do I just simply cat one at the end of the other?

An example for sample1 is shown below (FC = flow cell, L = Lane)

sample1.FC1.L1

sample1.FC1.L2

sample1.FC2.L1

sample1.FC2.L2

Thanks a lot in advance!

Answer 1

[Pierre Lindenbaum]

when aligning, you should specify the lane information in the read group (RG): .e.g see: How to choose the right RG,SM and LB values for alignment

you can align and sort each pair of fastq and merge them later: e.g: Merging Bam Files

[Matina]

thanks a lot for the reply. I was wondering why it is important to specify RG.

[Istvan Albert]

Read groups may be used to indicate which libraries are technical replicates of one another. That will help the variant caller decide how much variability comes from the instrument itself.

Answer 2

[Istvan Albert]

In general it is probably best to keep these separate as they form technical replicates and will help you assess potential biases between runs. You would be able the merge the alignment files later.

If you were to perform a study that works best with maximal data (like a genome assembly) then merging them early on is recommended

[Matina]

OK i got it, align first and then merge. I want to do a simple differential expression analysis. Thanks a lot!

[japizsaab]

Hi! I wanted to follow up on this if you wouldn't mind elaborating a bit more.

When you say " You would be able the merge the alignment files later.", I understand you mean that by just selecting these as replicates for downstream analyses you are essentially merging them and so there's no need to literally merge the files together, right?

I'm mostly curious as to whether this is generally accepted, and whether I can report the data generated like this just the same as if I had generated it by concatenating them from the beginning. I have already done DEG and pathway analyses while keeping the "flowcell1" and "flowcell2" samples as technical replicates, and practically I understand that I wont obtain much different results from concatenating them at the beginning, but still, I wonder if such processing is assumed (and so expected) in published data.

Thanks in advance for your help!

[Istvan Albert]

when we talk about merging what we all mean that one could literally merge the files if that becomes necessary - for example suppose there is an analysis tool can only take a single BAM file.

Combining information from multiple alignment files is also technically a merging, but not of the BAM files, but some subset information from them.