christopherbarrington / scamp Goto Github PK

like software versions yaml file it may be useful to explicitly record these parameters so they can be more-easily reported.

and same for stages

assumes an inputs directory - could just use the absolute path

NXF_HOME isn't clear - not everyone has it set

Unused values will invalidate the cache?

when merging gif files to create an index, the names should be checked for consistency

could

SEURAT:
    8 weeks replicate 1:

be used to indicate that this should be run with all default parameters

Ambiguous method overloading for method java.util.LinkedHashMap#plus.
Cannot resolve which method to invoke for [null] due to overlapping prototypes between:
        [interface java.util.Collection]
        [interface java.util.Map]

• build index
• build sample sheet
• count

each main.nf file should have an accompanying meta.yaml file to include easy to access information on the modules inputs/outputs etc

try to avoid using conflicting keys

for example: https://github.com/nf-core/modules/blob/master/modules/nf-core/bwa/mem/meta.yml

method to download and format as granges, if not provided in an index

• blacklist
• genes
• promoters
• transcripts/exons/introns/centromeres
• repeats?

remove the \ in scripts so they are properly formatted in work directories

write a module (?) that will present the information produced/data analysed/processes/steps for a (sub)workflow.

these should be collectable into a single document/webpage

may be better to have this option in the module so that it can be set as false for multimodal assays where the barcodes are checked

formats differ between R and shell scripts. and versions files need deciding

completed workbook/webpage reports could be pushed to a (or multiple) GitHub repositories

maybe specified with:

    github:
        remotes:
          - ChristopherBarrington/scamp-push-test:babs/docs

is it worth having multiple remotes? maybe would have a public/private version of reports to different remotes?

these should be in the analysis stanza groups either _default or in an analysis

should be tidier!

• publishing config
• resources config
• ext arguments config
• profiles
• crick config ... nfcore configs?

use --scamp-params param.yaml for example so that --params.stuff can also be used

an output could be a parameters yaml file with additional fields added - eg seurat path so that the pipeline can be restarted from here rather than the cache in the longer-term

reports just titled 'output'

once a seurat object is created the workflow should be the same independent of technology, so different quantification methods can use the same subworkflow

https://www.nextflow.io/docs/edge/process.html#fair

make an assay for ensembl and another for gene names

the make granges and biomart (for example) may be duplicated, can they be moved to a genome processing subworkflow?

• gtf to granges (may be index-dependent?)
• identification of mitochondrial features (from biomart?)
• identification of cell cycle genes (from cc.genes?)

should only be needed in one place

include hive datasets for the guess parameters script, need indexes etc

no asf label for it, sample sheet lists 'None'

wit #54

match up with asf lims

better organisation of the utilities functions module into directories (particularly for the bigger functions)

move script blocks to templates/main.sh to keep better organised and easier to read scripts

steal from nf-core!

write the module that uses an R script to create the initial (basic) Seurat object

these could be value channels instead of mapped?

need to include:

• ^GEX_
• ^ATAC_

• script
• documentations

validate parameters before running pipeline

https://github.com/nextflow-io/nf-validation

Using js.build instead to pass parameters:

https://discourse.gohugo.io/t/variables-inside-javascript/39086/3

• can use markdown in json strings

• move utilities out of modules
• make a utility archetype
• add documentation yaml for each utility
• add creation command to populate documentation script

currently uses two function calls but seems it could be streamlined

merge_process_emissions(make_object, ['opt', 'seurat'])
  .map{merge_metadata_and_process_output(it)}
  .dump(tag: 'seurat:prepare:cell_ranger:objects', pretty: true)
  .set{objects}

could this be in the {scampr} onLoad?

rather than one big master library "csv" make each file separately for each dataset. the library and feature type information is not in the dataset config.

processes should have their own environments, not expect software to be available for them