fricktobias / dbs-pro Goto Github PK

The UMI sequences can be used to identify chimeric sequences by looking for UMI:s linked to several different ABC or DBS sequences.

Just and idea I had about how we might want to change the order in our pipeline.

I have found the following issue. For UMIs we cluster them for each ABC target but do not separate on DBS. This could mean that we are merging UMIs that should in fact be separate. My proposal would be to separate all UMIs by ABC and DBS before clustering. This would better represent the actual conditions in the experiment.

I am however unsure about the benefits in the end, possibly this would only be a lot of work for nothing, but I wanted to raise the idea anyway to set what you think.

Current pipeline

START. Input = Fastq file

1. Separate for DBS
   1.1 Extract DBS
   1.2. Cluster DBS
   1.3 Correct DBS fastq

2. Separate for ABCs
   2.1. Extract ABC-UMI
   2.2 Split ABC-UMI by ABC
   2.3 Cluster ABCs independently
   2.4 Correct ABC fastqs.

3. Analysis of corrected DBS and ABC files.

END.

Purposed outline pipeline

START. Fastq file

1. Extract DBS
2. Extract ABC-UMI
3. Cluster DBS
4. Correct DBS fastq
5. Split/Tag ABC-UMI by DBS //This represent separated dropletts
6. Split/Tag ABC-UMI by ABC // This represents spliting within dropletts for different targets.
7. Cluster for each DBS-ABC pair indepentently
8. Correct DBS-ABC pairs
9. Analysis

END.

Include options to pass to snakefile. Used config in snakemake module?

Make dict with arguments in run.py --> pass to config.

These modules are currently not included in the environment file

• pandas
• seaborn
• dnaio
• snakemake
• umi_tools

It is too slow...

Fastq is not needed.

Currently snakemake is using all inputs/output as a globbed list rather than looping over them.

The wrapper DBS-Pro_automation.sh script should be & it's functionality should be moved to the installed package DBS-Pro.

Instead of individual rules for each case a general case should be developed

We should have a config file containing all relevant parameters for running the pipeline that is prone to change. For example:

• Handle sequences
• ABC sequences

Currently it is seemingly needed to give the path to the construct relative to the output directory rather than your working directory.

Provide example of how to edit and run the pipeline in a linked publically available google colab jupyter noteboook

Use pytest and run tests settings and running from remote directories.

Currently the entire DBS FASTQ is read into memory.

Currently the script only runs from the DBS-Pro folder

Currently the script takes the name from a inputed tsv file. Instead we want the names to be given from the file supplied in the config file.

The ABC file system is currently not that adaptable and it would be nice to be able to have several ABC-sequence files for different setups.

I'd suggest adding functionality for adding new construct file using dbspro set and a separate command for changing which is used by adding a dbspro config command (or something like that).