after FASTQC, what should we do?

[szy0931]

After FASTQC, just like the post in the link (http://rustbeltscientist.tumblr.com/...-first-rna-seq), I feel I do not have too many things to do since every action on data may drive the data from the real fact.

[GenoMax]

FastQC is meant to be a guide and should be used as such. The "limits" in it for various tests (which are configurable BTW, if the red "X" bothers you) were set for re-sequencing data and are not always applicable to myriad of other data types (e.g. RNAseq).

Dr. Simon Andrews has a lot of informative posts about FastQC observations at this site: https://sequencing.qcfail.com/software/fastqc/

Having something "fail" in FastQC does not mean that you can't move forward with analysis. I suggest that you pass your data through a trimming program (to ensure that there is no adapter contamination) and then move on to alignments. I recommend BBMap suite for both these tasks.

[szy0931]

I have the following red rights.
1. Per base sequence content. The first 15 baes have big imbalance and the last base (the 51st) always has much less A. I think removing the last base should be fine, but removing the 15 bases may affect mapping. I have many overrepresented short reads (<15 bases). they may cause the imbalance. I removed them. The imbalance got some correction, but they are still imbalanced.
2. Sequence Duplication Levels. It is caused by overoverrepresented reads. What else can cause this?
3. Overrepresented sequences. in this, adapter content is green while Kmer is red. I did blast and found there are three kinds of overrepresented reads. 1) short reads ( most <15, but some <21 bases). They do have hits on the genome. Should I remove them? Or I can leave them because they do have hits on the genome. 2) ssrA. Are they high copy or highly expressed in nature. I have not figured out. Are they suppose to be Should I remove them? Or I can leave them because I do not care their expression. Or, I can filter them out by adding a step of aligning with ssrA. 3) genomic DNA. Leave them or remove them? I want to leave them because they may be the genes I care about. But, I do not know if their overrepresentation here is caused by technical issues or they are really highly expressed in cells.

[GenoMax]

Forget the red lights for now :-)

What kind of dataset is this? My guess based on your description .. RNAseq.

Sounds like you did not read Simon's (Author of FastQC) posts from the page I linked above.

For #1 : https://sequencing.qcfail.com/articl...med-libraries/

For #2 and #3: https://sequencing.qcfail.com/articl...l-duplication/

If you have not scanned your dataset for illumina adapters then you need to do that now. Only thing you should care about removing is any adapter contamination.

[szy0931]

Thank you very much！
Just back from the experiment.
I am reading the links.

[VC87]

Hi everyone!
I am new to NGS analysis, so there are a couple of things i don't understand very well.One is the per base sequence content plot of FASTQC. Here it is shown the percenatge of each nucleotide for each position in the read.My question is: shouldn't each position of the read have only one nucleotide?Why is there always a percentegae of each nucleotide?Hope this makes some sense.Thanks in advance

[dpryan]

Hi everyone!
I am new to NGS analysis, so there are a couple of things i don't understand very well.One is the per base sequence content plot of FASTQC. Here it is shown the percenatge of each nucleotide for each position in the read.My question is: shouldn't each position of the read have only one nucleotide?Why is there always a percentegae of each nucleotide?Hope this makes some sense.Thanks in advance

Normally you have millions, possibly hundreds of millions, of reads, with each likely originating from a different region of the genome. As such, one would ideally expect the percentages of A/C/T/G at each position to match whatever the average level is in the genome. Often, however, you'll see some sort of sequence bias, especially at the 5' end.

[GenoMax]

Hi everyone!
I am new to NGS analysis, so there are a couple of things i don't understand very well.One is the per base sequence content plot of FASTQC. Here it is shown the percenatge of each nucleotide for each position in the read.My question is: shouldn't each position of the read have only one nucleotide?Why is there always a percentegae of each nucleotide?Hope this makes some sense.Thanks in advance

@Devon is referring to positions in the sequencing read. Every insert starts getting sequenced at base #1 and as such can be expected to have a random base at that position and for the remainder of the read. This is generally true for genomic re-sequencing.

If you are sequencing amplicons, 16S etc then all reads may have exactly the same nucleotide at many positions (case you are thinking about). People have to devise strategies to include random bases at the start as spacers to prevent low nucleotide diversity which can cause run failure.

Illumina sequencing assumes a random distribution of nucleotides at any given position in the read and works best when that condition is satisfied. That is the reason you will find that a neutral DNA (like phiX) is used as a spike-in to increase diversity (if a spacer strategy was not used for amplicons).

[VC87]

Thank you both dpryan and genomax!I think i undrstood both of your answers