Hi everyone! I’m running nf-core/crisprseq (2.4.0dev) in screening mode and I’m getting a consistent failure at the MAGECK_COUNT step with my paired end sequencing. The error is

ERROR ~ Error executing process > 
NFCORE_CRISPRSEQ:CRISPRSEQ_SCREENING:MAGECK_COUNT

ERROR: Cannot automatically determine the --trim-5 length.
Only 0.0439 % of the reads can be identified.
(exit status 255)

From the MAGeCK logs: • R1: mapping looks fine (~93% in trim test) • R2: almost no reads are identified during trim-5 auto-detection I also tried with cutadapter like in below

nextflow run crisprseq \                       
  --analysis screening \
  --input /Users/metinyazar/Desktop/nf_core/bond_group/crisprseq_results/MS1943_vs_UNC6852/samplesheet.csv \
  --library /Users/metinyazar/Desktop/nf_core/bond_group/crisprseq_results/MS1943_vs_UNC6852/library_fixed.tsv \
  --outdir /Users/metinyazar/Desktop/nf_core/bond_group/crisprseq_results/MS1943_vs_UNC6852/MS1943_vs_UNC6852_results \
  --contrasts /Users/metinyazar/Desktop/nf_core/bond_group/crisprseq_results/MS1943_vs_UNC6852/rra_contrasts.txt \
  --rra \
  --drugz \
  --bagel2 \
  --mle \
  --hitselection \
  --drugz_remove_genes "luciferase,LacZ,EGFP,Non_Targeting_Control" \
  --three_prime_adapter AGATCGGAAGAGCACACGTCTGAACTCCAGTCAC \
  --extra_cutadapt_three_prime_args "-A AGATCGGAAGAGCGTCGTGTAGGGAAAGAGTGT" \
  -profile docker \
  -c /Users/metinyazar/Desktop/nf_core/bond_group/crisprseq_results/MS1943_vs_UNC6852/user.config

It is still giving similiar error. Never done paired one before. How should I fix this? Thank you!

Hi Metin, sorry but I have never processed paired-end screening data, so I can be of very little help. As far as I understand, Mageck in paired end mode is supposed to search for the guide in R2 when it is not found in R1, is that correct? Could it be that actually no guides are found on R2? and in that case, wouldn't it make sense to treat the data as single-end just using R1 alone?

Yes, Mageck insists that there is no guides in R2 which is reverse one. I am now manually investigating this issue. I am also wondering that is treating data as single-end biologically correct? I am seeing enough coverage in the QC report and already did the analysis by treating single end. I really need someone to reassure me that I’m doing the right thing by treating it as single-end. 🙂

Hi, if your experiment is a standard CRISPR screen with no unusual design (such as constructs containing two guides), it makes sense that the guide sequence appears only on one read, typically closer to the 5′ end. The fact that you see good mapping on R1 supports this, I think.

When I manually investigate R1 and R2, there are more repeat sequences in R2 than R1. Probably R2 does not contain much guides or this is some kind of special adapter that mageck or cutadapter cannot trim.