Hi everyone, I am using the chipseq pipeline to align DAP-seq results. The method is pretty similar to chipseq but is in vitro. In short, we extracted the genomic DNA from our model organism then we sonicated it and added TruSeq adapters. Afterward, we put these libraries in contact with our proteins of interest that were produced in vitro, washed, and eluted the fragments that were bound to them. We then performed a PCR to add the multiplex indices to the eluted fragments and sent them for sequencing. We then aligned the sequencing results with the pipeline with the following command, everything went smoothly:

nextflow run nf-core/chipseq -profile docker --input /mnt/data-bioinfo/Users/maxlaurent/nextflow/projet2/data/samplesheet_pe.csv --outdir /mnt/data-bioinfo/Users/maxlaurent/nextflow/projet2/output --fasta /mnt/data-bioinfo/Users/maxlaurent/nextflow/projet2/genome/Zm-B73-REFERENCE-GRAMENE-4.0.fa.gz --gff /mnt/data-bioinfo/Users/maxlaurent/nextflow/projet2/genome/Zm-B73-REFERENCE-GRAMENE-4.0_Zm00001d.2.gff.gz --read_length 150 --save_reference true --aligner bowtie2 --narrow_peak true --keep_multi_map false

And the input samplesheet was the following:

sample,fastq_1,fastq_2,antibody,control
TF1_REP1_T1,/mnt/data-bioinfo/Users/maxlaurent/nextflow/projet2/data/TF1_LIB1_1.fastq.gz,/mnt/data-bioinfo/Users/maxlaurent/nextflow/projet2/data/TF1_LIB1_2.fastq.gz,HaloTag,GFP_REP1
TF1_REP4_T1,/mnt/data-bioinfo/Users/maxlaurent/nextflow/projet2/data/TF1_LIB4_1.fastq.gz,/mnt/data-bioinfo/Users/maxlaurent/nextflow/projet2/data/TF1_LIB4_2.fastq.gz,HaloTag,GFP_REP4
TF1_REP5_T1,/mnt/data-bioinfo/Users/maxlaurent/nextflow/projet2/data/TF1_LIB5_1.fastq.gz,/mnt/data-bioinfo/Users/maxlaurent/nextflow/projet2/data/TF1_LIB5_2.fastq.gz,HaloTag,GFP_REP5
GFP_REP1_T1,/mnt/data-bioinfo/Users/maxlaurent/nextflow/projet2/data/GFP_LIB1_1.fastq.gz,/mnt/data-bioinfo/Users/maxlaurent/nextflow/projet2/data/GFP_LIB1_2.fastq.gz,,
GFP_REP4_T1,/mnt/data-bioinfo/Users/maxlaurent/nextflow/projet2/data/GFP_LIB4_1.fastq.gz,/mnt/data-bioinfo/Users/maxlaurent/nextflow/projet2/data/GFP_LIB4_2.fastq.gz,,
GFP_REP5_T1,/mnt/data-bioinfo/Users/maxlaurent/nextflow/projet2/data/GFP_LIB5_1.fastq.gz,/mnt/data-bioinfo/Users/maxlaurent/nextflow/projet2/data/GFP_LIB5_2.fastq.gz,,
TF2_REP1_T1,/mnt/data-bioinfo/Users/maxlaurent/nextflow/projet2/data/TF2_LIB1_1.fastq.gz,/mnt/data-bioinfo/Users/maxlaurent/nextflow/projet2/data/TF2_LIB1_2.fastq.gz,HaloTag,GFP_REP1
TF2_REP4_T1,/mnt/data-bioinfo/Users/maxlaurent/nextflow/projet2/data/TF2_LIB4_1.fastq.gz,/mnt/data-bioinfo/Users/maxlaurent/nextflow/projet2/data/TF2_LIB4_2.fastq.gz,HaloTag,GFP_REP4
TF2_REP5_T1,/mnt/data-bioinfo/Users/maxlaurent/nextflow/projet2/data/TF2_LIB5_1.fastq.gz,/mnt/data-bioinfo/Users/maxlaurent/nextflow/projet2/data/TF2_LIB5_2.fastq.gz,HaloTag,GFP_REP5
TF3_REP1_T1,/mnt/data-bioinfo/Users/maxlaurent/nextflow/projet2/data/TF3_LIB1_1.fastq.gz,/mnt/data-bioinfo/Users/maxlaurent/nextflow/projet2/data/TF3_LIB1_2.fastq.gz,HaloTag,GFP_REP1
TF3_REP4_T1,/mnt/data-bioinfo/Users/maxlaurent/nextflow/projet2/data/TF3_LIB4_1.fastq.gz,/mnt/data-bioinfo/Users/maxlaurent/nextflow/projet2/data/TF3_LIB4_2.fastq.gz,HaloTag,GFP_REP4
TF3_REP5_T1,/mnt/data-bioinfo/Users/maxlaurent/nextflow/projet2/data/TF3_LIB5_1.fastq.gz,/mnt/data-bioinfo/Users/maxlaurent/nextflow/projet2/data/TF3_LIB5_2.fastq.gz,HaloTag,GFP_REP5

In the quality control of the reads, we observed that there is an unexpectedly high percentage of duplicated reads for all our samples. This resulted in very low read numbers for alignment after filtering (7/12 samples with less than 1 million reads and the rest between 1 and 7 million reads starting with an average of 30 million reads per sample). The sample similarity analysis done with DESeq2 was quite weird as none of our replicates correlated. This makes us think that our experiment failed. By checking Picard documentation, we believe that before the PCR to add the indices, we had very low DNA material (after the elution from the protein). As we performed a PCR with 30 cycles, we think that we enriched our samples for the same fragments and that it resulted in a high number of duplicates. We are now redoing the sample preparation and planning to reduce the number of cycles to 18 so that we don’t change the library complexity too much. I post this question to see whether someone had the same problem before and also if our hypothesis is plausible and to see if we maybe missed other potential causes. Thank you in advance for your help

I'm not a biologist, but isn't 30 cycles of PCR really high? I think we use ~10-12 cycles in our circulating cell-free DNA samples, which start with 5ng of DNA. There is a tool (haven't used it myself) called EstimateLibraryComplexity that may be useful on the current samples