I have bisulfite sequencing reads for 6 samples. At first I used fastQC to check the quality of the reads. I found lots of Illumina universal adapters in my reads. Then I used trimmomatic paired-end to trim these overrepresented sequences. After trimming with the trimmomatic, the quality of the forward reads improved but reverse reads still are terrible. Actually, there were overrepresented sequences in both forward and reverse reads before trimming. After trimming, the forward reads do not have those over-represented sequences, but the reverse reads still have them.
java -jar trimmomatic-0.39.jar PE -threads 12 PV6N-3_S35_L004_R1_001.fastq.gz PV6N-3_S35_L004_R2_001.fastq.gz pv6n_3_output_forward_paired3.fq.gz pv6n_3_output_forward_unpaired3.fq.gz pv6n_3_output_reverse_paired3.fq.gz pv6n_3_output_reverse_unpaired3.fq.gz ILLUMINACLIP:sequence_adaptors.fasta:2:30:10:2:keepBothReads
This is my command. Almost 20 percents of the reverse reads contain over-represented sequences.