multiplex samples

[Wei-HD]

Dear all,

I have a question to ask about multiplexing with Illumina Hiseq (I can have 150-200 Million reads per lane). I have 12 samples from 3 different conditions (each has 4 biological replicates, mouse cell culture):

control samples (C1, C2, C3, C4)
shRNA-1 samples (A1, A2, A3, A4)
shRNA-2 samples (B1, B2, B3, B4)

I plan to put 4 samples in each lane, then I have 3 lanes (save some money). I am not sure how should I choose the samples in each lane, either put 4 replicates in the same lane? or do a combination of different groups like:

Lane 1: C1, C2, A1, B1
Lane 2: C3, A2, A3, B2
Lane 3: C4, A4, B3, B4

And another question about single end and paired end sequencing? I am mostly interesting in the differential gene expression, not that into splicing or SNP. The core facility provides singe end 50 bp/100 bp and paired end 50 bp/100 bp, would single end 50 bp be enough for my purpose?

Thanks a bunch for your advice!
Wei

[dpryan]

If you're just interested in doing DE analysis, then single-end 50bp reads will work just fine (at least if you have a decently annotated species like human or mouse). Also, for DE that's actually a bit overkill in terms of lanes. You might be able to get by with just 1, though I'd do 2 to be safe (the multiplexing is simple in those cases).

Regarding your multiplexing strategy, your core (and most other providers) would probably recommend just multiplexing all the samples on each of the lanes. You'll then have technical replicates that can simply be merged either before or after alignment (it doesn't matter which in this case).

[Wei-HD]

Hi dpryan,

Thanks a lot for your advice! I will ask the core facility tomorrow about the lane numbers.

Regarding your multiplexing strategy, your core (and most other providers) would probably recommend just multiplexing all the samples on each of the lanes. You'll then have technical replicates that can simply be merged either before or after alignment (it doesn't matter which in this case).[/QUOTE]

Here you mean merge 2 fastq files (from technical replicates) together before alignment or merge 2 SAM files after alignment. I've never done either of them, can you give me some hint? using Samtools?

Again thanks a lot!
Wei

[dpryan]

For merging fastq files:

cat sample1_lane1.fastq sample1_lane2.fastq > sample1.fastq

This also works for gzipped files, which saves space:

cat sample1_lane1.fastq.gz sample1_lane2.fastq.gz > sample1.fastq.gz

Samtools has a "merge" command to merge BAM files:

samtools merge sample1.bam sample1_lane1.bam sample1_lane2.bam

You can merge SAM files too. I think picard mergeSam can do that, though you could also do that with samtools and just append one file to another (i.e., use ">>").

It's usually easiest to just "cat" the fastq files.

[Wei-HD]

Thanks dpryan! That is really helpful

[Wei-HD]

Hi dpryan,

I have more questions for you

Do you think these two ways will make a difference in terms of total reads number of each sample? is that possible that one sample has way more reads than other samples?

1. multiplex 12 samples in one lane, 2 lanes as technical replicates
2. 6 samples one lane, 2 lanes in total

[GenoMax]

Do you think these two ways will make a difference in terms of total reads number of each sample? is that possible that one sample has way more reads than other samples?

1. multiplex 12 samples in one lane, 2 lanes as technical replicates
2. 6 samples one lane, 2 lanes in total

It is certainly possible to get more reads for some samples if the libraries are not mixed at equal concentration. Search the forum for ways of concentration estimation.

Pooling following #1 would give you some data for all samples (if one lane fails). In case of #2, you would lose all data for 6 samples in case a lane fails.

When multiplexing samples libraries with uniform quality are important.