Trimmomatic QC report shows drop in the reads and presence of overrepresented sequences

Question

This question was also asked on Biostars

I am performing a de novo genome assembly using Illumina paired-end short reads, sequenced on a NovaSeq X by our collaborator at UCLA.

At present, I am in the stage of trimming the adapters. Here, you can have a look at the basic statistics and information on the adapter content obtained from the Fast QC report, for R1.

FOR RAW READS

I used Trimmomatic for trimming the adapter. The following is the Trimmomatic Settings

ILLUMINACLIP:~/adapters/TruSeq3-PE.fa:2:30:10 MINLEN:36

Here, you can see the basic statistics and adapter content of the Trimmed reads.

Here, the output was:

Both surviving: 566832403 Forward only surviving: 39244376 Reverse only surving: 0.00

Dropped reads: <1%

Here you can go through the full FastQC report for:

Raw Reads

Trimmed Reads

Now I have two questions:

Question 1

Can I go ahead with the assembly process, for there is zero-adapter-presence in the reads? Should I mind the loss of reads?

Question 2

I see that there are over-represented sequences, both in read 1 and read 2. I doubt if I can leave them be, or I should trim them too. I would like to know if the over-represented sequences need to be tossed, or if they can be left untouched and proceed with the assembly. If I need to remove these sequences, can Trimmomatic do it?

If I were to trim the over-represented sequences, can I use an adapter "fasta" file as follows?

The following are the over-represented sequences for R1.

The following are the over-represented sequences for R2.

Can I have my fasta file as follows, to remove the over-represented sequences:

>PrefixPE/1
-------sequence1--------------
-------sequence2--------------
-------sequence3--------------
-------sequence4--------------
>PrefixPE/2
-------sequence1--------------

java -jar trimmomatic-0.30.jar PE s_1_1_sequence.txt.gz s_1_2_sequence.txt.gz 
lane1_forward_paired.fq.gz lane1_forward_unpaired.fq.gz lane1_reverse_paired.fq.gz 
lane1_reverse_unpaired.fq.gz ILLUMINACLIP:overRepresentedSeqs.fa:2:30:10  MINLEN:36

NB: Sorry about the lengthy post. Since the file sizes are large, it takes over a day for the Trimmomatic to finish the task, so that it is not feasible to go with multiple rounds of trial and error check.

Answer 1

Given that it's a NovaSeq, these are almost certainly "no signal" reads as a result of the two-colour chemistry and should probably be excluded (or at least ignored).

A long polyG sequence is unlikely to affect mapping by much, because they're fairly rare in most genomes. Unfortunately you can't exclude them using Trimmomatic alone.

PolyG tail trimming can be done using fastp. I don't have any experience using fastp, so you'll need to read the documentation yourself.