Some suggestions on improving knock-knock (v0.4.1)

[hukai916]

Hi Jeff,

I am very interested in the way how knock-knock visualizes CRISPR outcomes. Therefore, I scratched my head for quite a few days and finally got knock-knock to work with one of your sample data when not using donor sequence. I am summarizing what I found below to help you further improve the tool.

First, please kindly be advised to update the documentation with the latest information, i.e. input requirements, it can save tons of debugging time for others.

knock-knock build-targets

For this command to work, I must:

• update the column names used in targets.csv, it is clear that some names are no longer accepted by the software, for example, donor_sequence should be donor, amplicon_primers should be primer_names, etc.
• match column names in targets.csv and that in data/pacbio/sample_sheet.csv
• remove donor column from targets.csv and data/pacbio/sample_sheet.csv, otherwise, self.reference_sequences complains about missing donor. I think the reference_sequences parsing modules are to blame.

knock-knock process

I must:

• need to update the default --stage to use "generate_figures" rather than "visualize"
• need to update the following line of target_info.py, otherwise the primer length is not 2 in my test. https://github.com/jeffhussmann/knock-knock/blob/a4d64c800c0b666d77a4c253588805bbadc22d68/knock_knock/target_info.py#L798

others

• again, the documentation should be updated regularly and example data and output should be provided. Also, the documentation should specify which version of the tool it serves.

Last, if you can share the most recent test data and instructions, I am happy to help to further debug the tool, just let me know.

--Kai

[jeffhussmann]

Hi Kai -

Thanks for your interest in knock-knock, and your patience in dealing with its outdated documentation and example data! My primary focus these days is prime editing, and it has been a long time since I have used the non-prime-editing-related parts of knock-knock.

Commit https://github.com/jeffhussmann/knock-knock/commit/1871c414a7090f2d3f0ee52648377f1da4b39c5e updates HDR-related code and brings the example data and related documentation up to date. Note that this involves some changes to formatting of target specification files that are probably easiest to understand to checking the diff of README.md in that commit.

With these changes, I am able to successfully process the provided example data on a fresh install. In particular, I am not able to reproduce the issues you mention above related to missing donors or to primers. Could you provide further context for these?

[hukai916]

Thanks very much for the quick fixes, @jeffhussmann! I am testing the updates out in the next few days and will report back.

Another issue that I found with previous example data is that when knock-knock aligns the primers using the hits library, it failed to detect any match on the reference. I believe this is because STAR only output reads with at least 50 bp mapped, and the primer sequences are not that long. Setting the corresponding parameter to 10 solved the problem. However, this parameter is hard-coded in your hits library, and I need to hack it by creating my own version of the hits library to make knock-knock happy with the example primer. I can provide more info when I am with my Mac. Any input? Thanks!

[jeffhussmann]

I am unable to reproduce any issues with primer alignment for the example data. Please provide all details necessary to reproduce the problem, including the version number/commits of knock-knock, hits, and STAR you are using.

Primer alignment is performed here, where the mode='permissive' argument removes any lower bound on alignment length, as you can see here.