FastQC,kmer content, per base sequence content: is this good enough

[mgg]

Hi,

I'd appreciate some advice on processing some Illumina libraries

Initial FastQC runs showed the data as not great. I've used cutadapt to trim off adapters and FastQC shows improvements to all libraries.

One remains of concern, because it still retains kmer and other issues (I've attached files for kmer content & per base sequence content for both the original and the processed data)

My question is simple: is this good enough? (my next step is assembly with velvet) Does this data need some further processing before Velvet? If so, with what? I've considered trimming off the first 10nuc to remove the anomalous per_base_sequence_content trace, but that would do little for the persistent kmers.

If this were your data, what would you do before velvet assembly?

thanks
mgg

for the record my cutadapt commands are below

PHP Code:

# trim reads/2
cutadapt -b AGATCGGAAGAGCGTCGTGTAGGGAAAGAGTGT --minimum-length=10  --overlap=4 --quality-base=64 --quality-cutoff=3 --match-read-wildcards infile_2.fq -o processed/outfile_2.fq --wildcard-file=processed/outfile_2.fq.wildcard

# trim reads/1
cutadapt -b GATCGGAAGAGCACACGTCTGAACTCCAGTCAC --minimum-length=10  --overlap=4 --quality-base=64 --quality-cutoff=3 --match-read-wildcards infile_1.fq -o processed/outfile_1.fq --wildcard-file=processed/outfile_1.fq.wildcard 


[minoru_harvest]

yeah. i got the same question
i have a very similar graph with your prosessed-per-base-sequencecontent

[Wallysb01]

Looks like you have some base pair bias issues going on from bases 1-10 in your reads. You should trim those off.

[Jane M]

Hello everybody,

I come back to this topic which fits well to my interrogation: I would like your point of view on my RNA-Seq data (paired-ends, 100bp) generated by an Illumina HiSeq 2000 machine.
I attached the "Per Base sequence Quality" and "Kmer Content" for 3 examples. In the first one, the library was prepared using polyA method. The 2 next examples were performed by ribodepletion. I would like to know if my data are "good enough" despite these 2 last profiles and if there is an explanation for this increase of A/T sequence along the read?

I have the feeling from these examples and some others that the "Kmer Content profile" depends on the library preparation (ribodepletion vs polyA), the run (samples from a same run show a similar profile) and the sample itself (I observed similar profiles for a same sample ran on 2 different runs). Is this true?

Thank you,
Jane

[fahmida]

Originally Posted by mgg

Hi,

I'd appreciate some advice on processing some Illumina libraries

Initial FastQC runs showed the data as not great. I've used cutadapt to trim off adapters and FastQC shows improvements to all libraries.

One remains of concern, because it still retains kmer and other issues (I've attached files for kmer content & per base sequence content for both the original and the processed data)

My question is simple: is this good enough? (my next step is assembly with velvet) Does this data need some further processing before Velvet? If so, with what? I've considered trimming off the first 10nuc to remove the anomalous per_base_sequence_content trace, but that would do little for the persistent kmers.

If this were your data, what would you do before velvet assembly?

thanks
mgg

for the record my cutadapt commands are below

PHP Code:

# trim reads/2
cutadapt -b AGATCGGAAGAGCGTCGTGTAGGGAAAGAGTGT --minimum-length=10  --overlap=4 --quality-base=64 --quality-cutoff=3 --match-read-wildcards infile_2.fq -o processed/outfile_2.fq --wildcard-file=processed/outfile_2.fq.wildcard

# trim reads/1
cutadapt -b GATCGGAAGAGCACACGTCTGAACTCCAGTCAC --minimum-length=10  --overlap=4 --quality-base=64 --quality-cutoff=3 --match-read-wildcards infile_1.fq -o processed/outfile_1.fq --wildcard-file=processed/outfile_1.fq.wildcard 


Are these reads from mate pair libraries? You may also want to check the read duplication levels in that case.

[Jane M]

I come back to my previous question because I still have doubts concerning the quality of my data. Any feedback would be appreciated

[Wallysb01]

I didn't see the attachment. But from what you describe, it sounds ok.

[Jane M]

Originally Posted by Wallysb01

I didn't see the attachment. But from what you describe, it sounds ok.

Oups, I forgot to attach the file!

[Jane M]

Originally Posted by Jane M

Oups, I forgot to attach the file!

Any comment with the attachment?

[Wallysb01]

Looks good enough for mapping. Might want to see if you have some adapter contamination in the first one. I've often found weird suden spikes of particular kmers are the adapters.

[Jane M]

Thank you Wallysb01.

Isn't it suprising to see an increase of AAAAA and TTTTT all along the read? It shoulb be constant, right?, like in the first case.
Why is there such a difference between polyA and ribodepletion?

Do all the "normal/good profiles" of these 2 methods always differ?