Kmers remain even after trimmomatic trimming

[modi2020]

Hi Guys,

I have Illumina NGS DNA 150 BP paired end reads.
My initial Fasqc report indicated the presence of Kmers towards the end of the reads at 145-147.
I used trimmomatic to trim them off. I trimmed 5 BP from the start of the read and 5 BP at the end of the read which makes my read length 140 bp (and should remove the kmers). However, when I looked at the Fastqc report post filtering, it showed that the kmers still exist but are now in the position 135-136. I have attached the pre and post filtering Fastqc reports if it helps to visualize them.


My trimmomatic trimming command was as follows:
java -Xmx15g -classpath trimmomatic-0.22.jar org.usadellab.trimmomatic.TrimmomaticPE -threads 4 -phred33 -trimlog trimmlog_log.txt input_R1.fastq input_R2.fastq output_R1.fq unpaired_output1.fq output_R2.fq unpaired_output2.fq HEADCROP:5 CROP:140 LEADING:20 TRAILING:20 SLIDINGWINDOW:4:20 MINLEN:60

I would really appreciate it if any one can guide me through this issue as I couldn't figure it out.

[mastal]

You should do some sequence-specific adapter trimming using the ILLUMINACLIP command.

[modi2020]

Dear Mastal,

Thank you for your prompt reply.
I honestly thought that ILLUMINACLIP command was only for Overrepresented sequences e.g Illumina adapters. I will try using it and see what I get.
Thank you

You should do some sequence-specific adapter trimming using the ILLUMINACLIP command.

[mastal]

It is only for the Illumina adapters.

The adapters are not in the same position in each read, and they are not present in every read, only when the DNA insert is shorter than the length of one read, so that you read through into some of the adapter sequence.

[luc]

The pre-filter FASTQC kmer plot shows the presence of these enriched kmers are about 50-fold enriched even at base 110 of your reads.
As mentioned better try a normal adapter trimming and (depending on the purpose of teh experiments) the FASTQC reports also indicate that the reads could use some quality trimming or filtering.

[modi2020]

I have used the ILLUMINACLIP command to remove the kmers. My command was as follows:
java -Xmx15g -classpath trimmomatic-0.22.jar org.usadellab.trimmomatic.TrimmomaticPE -threads 4 -phred33 -trimlog trimmlog_log.txt input_R1.fastq input_R2.fastq output_R1.fq unpaired_output1.fq output_R2.fq unpaired_output2.fq ILLUMINACLIP:adapters3.fasta:2:30:10 HEADCROP:5 CROP:140 LEADING:20 TRAILING:20 SLIDINGWINDOW:4:20 MINLEN:60

I also attached the adapters3.fasta file in which I specified the kmers.
The program ran at first for a few minutes but its progress has stopped for six hours now even though its running on 4 cores.

Do you think that my command and adapters format correct?

It is only for the Illumina adapters.

The adapters are not in the same position in each read, and they are not present in every read, only when the DNA insert is shorter than the length of one read, so that you read through into some of the adapter sequence.

[mastal]

Your trimmomatic command looks OK, but it looks like you are using Illumina barcodes instead of adapter sequences.

Just use the adapters.fasta file that comes with trimmomatic.

Do you know what type of Illumina kit was used for library prep?

[modi2020]

Dear Mastal,

After following your advise I am glad to say that it worked! I have attached the output of Fastqc herein.
I have basically used the file TruSeq3-PE-2.fa as adapters sequences and it worked wonderfully. I have asked the fellows who did the library prep about the kit but haven't gotten a reply yet. If they did I will let you know. I am just glad it worked now

Again, thank you so much for your follow up with me

Your trimmomatic command looks OK, but it looks like you are using Illumina barcodes instead of adapter sequences.

Just use the adapters.fasta file that comes with trimmomatic.

Do you know what type of Illumina kit was used for library prep?

[modi2020]

Dear Luc,

Thank you for your advise. I have actually trimmed up to 110 previously but the presence of kmers in the Fastqc report persisted. When I used the adapters set supplied by trimmomatic for TruSeq3-PE-2.fa, it helped remove them completely. I have attached the result of that run on my previous reply to mastal.

Again thank you so much

The pre-filter FASTQC kmer plot shows the presence of these enriched kmers are about 50-fold enriched even at base 110 of your reads.
As mentioned better try a normal adapter trimming and (depending on the purpose of teh experiments) the FASTQC reports also indicate that the reads could use some quality trimming or filtering.

[vromanr_2015]

it didnt worked for me

I tried to trim with Truseq3 fasta file and even after that I got the first 9bp with kmers. I tried to crop that but after that I got overrepresented reads (which I didn't have before).

any help highly appreciated.

[Brian Bushnell]

Perhaps you could post your fastqc results? The first 9 bp are unlikely to be related to adapters, but some systems (like Nextera) will have highly biased initial bases.

[vromanr_2015]

It didnt worked for me

sure! sorry about that.

[Brian Bushnell]

You need to ask what kind of library this was. The spikiness at the beginning is similar to Nextera libraries (in which case you should not trim the first 9bp), but I'm not quite sure.

If it is a Nextera library, then it makes sense that trimming Truseq adapters will not change anything. I don't understand fastqc's relative enrichment graph, though, so I won't comment on that.

The Nextera adapter sequences are included with BBTools in the resources directory (nextera.fa.gz). But I recommend that you consult the source of the library to find out what adapters were used before simply trying that to see if it works.

[vromanr_2015]

This is RNA-seq data from wheat, Could be an adapter that I don't have?

[Brian Bushnell]

Yes. The organism and data-source are generally independent of the adapter type. Although, there is an RNA-specific set of TruSeq adapters, which is also included with BBTools as truseq_rna.fa.gz.