Hello there, I tried to run bowtie2 mapping by specifying the --fasta parameter with a library file in fasta format. However, I got the error message "Process NFCORE_CRISPRSEQCRISPRSEQ SCREENINGBOWTIE2_ALIGN declares 5 input channels but 4 were specified". How should I fix this? Many thanks!

could you try running the dev version of the pipeline? If I am not wrong this is a bug that is fixed on dev

Thanks! I'll try it.

maybe @Matteo Bonfanti can help more than me with this?

Hi Patrick, maybe you are facing this issue here: https://github.com/nf-core/crisprseq/issues/227 When aligning with bowtie you need to make sure that the sequences are trimmed to remove the adapters and constructs that were inserted by the crispr, and leave only the sequence of the guide. Is this the case with your data? In case, the dev version now includes also some cutadapt parameter to set up the appropriate options for training.

Hi Matteo, this is exactly what I got, zero mapping rate, using the bowtie2 alignment option. I guess one has to trim very accurately before mapping with bowtie. On the other hand, specifying the 5' adapter parameter dramatically improved mageck count's own alignment algorithm. So bowtie2 is not necessary anymore.

In general, yes, to align with bowtie you need to accurately trim your sgRNA sequences. That's why in the dev branch we also included options to tailor the behaviour of cutadapt prior to mapping (extra_cutadapt_five_prime_args and extra_cutadapt_three_prime_args). Mapping with mageck is simpler, because mageck takes care of everything, but the alternative of the bowtie mapping is given because bowtie gives you the possibility of fine-tuning the alignment (e.g. by allowing a certain degree of mismatch between the sequence and the reference) You can find some explanations and a tutorial in the mageck manual, here: https://sourceforge.net/p/mageck/wiki/advanced_tutorial/#h-tutorial-1-allow-mismatches-for-read-mapping