#Manifests & Pipeline Configuration

#Overview

The pipeline uses several sources of configuration

• Main manifest: contains per-plate metadata used by the image-processing steps (channels, which channels to use for background prediction, which channels to deconvolve, which PSF files to use, etc.). See example manifest: TBD
• Registration manifest: describes reference/query plate pairs for registering images across cycles. See example: TBD
• Blacklist: list of plate/well combinations to exclude from a run. See example: TBD
• nextflow.confi Contains system wide parameter and defualts, and lives in the repo folder
• <my_run>.config Lives in your instance folder and provides settings for your specific run

#Main manifest — columns & purpose

Below are the columns present in the example main manifest and what each field controls in the pipeline.

• plate: Plate identifier used throughout the pipeline (string).
• index_xml: path to an index/XML file for the plate (or NA if not used).
• channels: comma-separated list of channel indices available for this plate (e.g. 1,2,3,4).
• bp_channels: channel indices used for background/flatfield estimation (background prediction).
• cp_nucl_channel: channel index used as the nucleus channel for Cellpose segmentation (or none).
• cp_cell_channel: channel index used as the whole-cell channel for Cellpose (or none).
• dc_channels: channel indices to feed into deconvolution (if dc_run=true).
• dc_psfs: comma-separated list of PSF file paths corresponding to dc_channels (order should match channels list).
• mask_channels: channel indices used to demultiplex nucelear vs cytoplasmic signals (if applicable).

Example (single-row, simplified):

Notes and tips:

• Channel indexes here are 1-indexed (as is the Phenix metadata), but the pipeline actually runs 0-indexed. I may harmonize these in future to avoid confusion.
• none should be used to indicate you want to skip something or its not available.
• All plates should be defined (cycle one (reference) and subsequent cycles). You can set cp_ columns to none for these.
• The dc_psfs field must match the dc_channels order-by-order. If a plate has no PSFs for a channel, use none for that plate or omit deconvolution for that plate.
• Use none in cellpose fields if Cellpose should be skipped for that plate, but cellmasks are needed to progress the pipeline beyond deconvolution.

#Registration manifest — columns & purpose

• reference_plate: plate id used as the registration reference.
• reference_channel: channel index on the reference plate used to compute transforms.
• query_plates: plate id(s) to be registered to the reference.
• query_channels: channel index(s) on the query plate(s) used for registration.

Example rows in the registration file map a reference plate/channel to a query plate/channel used when aligning cycles.

#Blacklist file — format & purpose

• Format: TSV with two columns plate<TAB>well. Each row excludes the given well from processing for that plate.
• Use this to exclude wells with acquisition failures or severe artefacts.

#How to point the pipeline to these file

• Option A — set params in your instance config (recommended for repeatable runs):

params {
  rn_manifest = 'inputs/manifest.tsv'
  rn_manifest_registration = 'inputs/manifest_registration.tsv'
  rn_blacklist = null   // or 'inputs/blacklist.tsv'
}

• Option B — pass on the command line when launching Nextflow (overrides config):

nextflow run </path/to/main.nf> -profile local --rn_manifest inputs/manifest.tsv --rn_manifest_registration inputs/manifest_registration.tsv --rn_blacklist inputs/blacklist.tsv

#Practical notes

• The pipeline expects wells listed in the registration manifest to be present in all cycles referenced; missing wells will be ignored for registration.
• If you prefer per-plate control, you can remove wells directly from the per-plate manifest in the rn_image_dir instead of using a global blacklist.

References

• See the run/config guide for full parameter names and examples: Configuring the pipeline