mpusp / snakemake-crispr-guides Goto Github PK

• currently report shows only guides for first N 1000 bp windows
• much more useful would be broad overview of coding and non-coding regions and the guide location
• e.g. spanning larger regions of chromosomes

• relative path when sourcing utilities.R can fail with changed wd --> more robust path from snakemake object to source other scripts
• legend.position = 0 fails creation of report.html

• currently only csv output available

• an arbitrary number of control guides should be designed automatically
• random sequence, low off-target binding

• different Cas enzymes
• different scores --> test how prediction changes
• include warnings for scores that are appropriate for certain applications or not (e.g. Cas9 derived scores vs Cas12 enzyme choice)
• do online tools find similar guides?

• right now a fixed window of e.g. 500 nt downstream and upstream of TSS is used
• this sometimes overlaps with upstream genes leading to removal of guides with >1 TSS window
• ideal is behavior where overlap is reduced by shrinking allowed 5' UTR to the terminus of upstream gene
• ... and limit downstream TSS window to length of gene if shorter

• change license to MIT
• add explicit statement that third party tools have their own license terms

• this issue is to test and find out what is required to be listed as "standardized usage"
• apparently a README.md in config/ with description how to run the wf, will be section 3 on renderd usage page
• what else is needed?

• no trivial question as A) bacteria don't have annotated transcripts
• and B) how to deal with operon structure? Treat every CDS as potentially independent transcript?

• currently names of chromosomes are hardcoded to NC.XXXX.[12], typical RefSeq numbers
• however it should be possible to add custom chromosomes or plasmids with arbitrary initial name

• BSgenome generation fails when ID is missing in gff files
• could be added automatically from locus_tag

• right now biostrings is used to search off-targets as it has no dependencies
• bowtie requires a BSgenome object
• can be made with BSgenome R package

• CRISPRoff score was one of the best features to predict guide efficacy in our recent CRISPRi lib paper (https://www.biorxiv.org/content/10.1101/2023.02.13.528328v1)
• can include it as extra module (available on conda?)

• errors out with get_genome complaining about not finding output files
• works locally, must be a problem specific to the testing archictecture
• try simplified workflow on extra branch

• we found in a recent study (Miao et al., 2023) that high G but no other base provided better guide efficacy
• G was enriched in high efficacy guides in a stretch of -4 to -14 bases upstream of PAM site (NNNNNNGGGGGGGGGGNNNN-NGG)

When running the pipeline for Synechococcus PCC 11901, the number of CDS considered by the pipeline was less than the number of CDS annotated in the genome.

This was due to some CDS annotations using GeneMarkS-2+ rather than RefSeq or protein homology. By including GeneMarkS-2+ in the gff_source_type dictionary in the get_genome script, I was able to include the previously missing CDS.

Are there any other methods of annotating CDS when assembling genomes that should be included here?

• currently guides are designed based on annotated features, mostly the CDS of genes
• intergenic regions etc are not targeted by default, but could be, using tiling
• tiling is littering a genomic regions with guides in regular intervals
• even for tiling guides, off and on-target check should be performed and top candidates selected (how many?)

• right now all "off-targets" meaning all guides that target more than one locus are removed
• this is not always wanted: duplicated genes can be targeted at multiple sites with the same guide
• need to add functionality to give users the choice to filter out such "good" off-targets or not
• multi target guides should be flagged in output tables and in report

• pipeline overview showing all steps
• CRISPR guide RNA design scheme, similar to CRISPRi paper Larson et al. 2013

When using "GCF_005577135.1", species name is parsed as "Picosynechococcus sp. PCC 11901 chromosome". The correct and valid name for the organism's name would be “Synechococcus sp. PCC 11901” or "Picosynechococcus sp. PCC 11901" (https://www.ncbi.nlm.nih.gov/Taxonomy/Browser/wwwtax.cgi?id=2579791).
This causes (in create_bsgenome.R)

txdb <- quiet_txdb( file = genome_gff, organism = genome_name, chrominfo = seqinfo_genome )$result

to break, which can be fixed by hardcoding the taxonomyId:

txdb <- quiet_txdb( file = genome_gff, organism = genome_name, chrominfo = seqinfo_genome, taxonomyId = 2579791 )$result

The same happens in design_guides.R in:

txdb <- makeTxDbFromGFF( file = genome_gff, organism = unname(genome(seqinfo_genome)[1]), chrominfo = seqinfo_genome )

• space that acts as an insulator between intergenic and CDS targeting guides
• should be appended on the CDS/target's 5' and 3' end