GIAB NA12878 dataset

[NielQC]

Hi all,
first of all must say I'm a newbie in NGS and a completely inept in forums like this, so dont blame me too much. In fact its my first post...

My objective is to create a pipeline for variant calling using the well known reference genome NA12878 from Genome In a Bottle (GIAB) consortium for validating my variant calls. I want to use a HiSeq 300x dataset (ftp://ftp.ncbi.nlm.nih.gov/giab/ftp/...01_HiSeq_300x/), and here my doubts started. In this directory you can see folders like that:

- 131219_D00360_005_BH814YADXX
- 131219_D00360_006_AH81VLADXX
- 131223_D00360_007_BH88WKADXX
- 131223_D00360_008_AH88U0ADXX

... and so on until 14 folders. I understand every folder its a run, so I went to the newest run, "131219_D00360_005_BH814YADXX", that contains 6 samples. Cant understand how and why that samples were generated. I think they were obtained from the same library (right?), so theoretically in each sample are covered the same regions. can I merge all R1 and all R2 of all samples together in an unique R1 and R2, or should I use just one sample?

The principal problem here it's i don't understand the "sample " concept. If it's the same individual, why making 6 samples when you could just sequencing one.

I hope I have explained clearly enough my doubts, thank you in advice

NielQC

[tristan dubos]

Hi NielQC,
You should read ftp://ftp.ncbi.nlm.nih.gov/giab/ftp/...nd_NA12878.txt that will help you to understand
It seems like to be a sequencing of trio then if it's not the same sample you can t merge files from different samples
Best regards

[Macspider]

As a side note, the reason why you sequence one sample more than once is that you usually do want to have some replicates, i.e. statistical power in the subsequent analyses. This doesn't apply to every case in NGS science, but it usually does.
For example, if you were to calculate the differentially expressed genes and had just 1 sequencing run per condition, your p-values of the differentially expressed genes would be very high. With 2 replicates per sample, they would be a bit lower and you might have some statistical confirmation that those genes are indeed differentially expressed.

This was just an example, but the concept of replicates is very important in NGS (as it is in many science aspects), because sequencing errors happen!