#Analysing features in R

The R package tglow-r provides downstream analysis of the pipeline's output: reading the CellProfiler features into a single object, quality control, transformation and scaling, dimensionality reduction, clustering, and modelling. Its vignettes are written for the output structure of this pipeline.

• Documentation: tglow-r
• Repository: TrynkaLab/tglow-r

Useful pipeline outputs to combine with tglow-r:

• rr__features/cellprofiler/<plate>_cells.parquet and <plate>_image.parquet: the CellProfiler features, aggregated per plate.
• rr__features/measurements/<unscaled or scaled>/: the pipeline's own intensity features, per well.
• rr__features/cellprofiler/features/: the per-well CellProfiler zip archives. Since v0.2.0 these are only published with cpr_publish_features_per_well = true.
• rr__qc/well_qc.tsv: the per-well QC verdicts, for filtering wells. See QC report.
• rr__cellcrops/: the cell crops index, for linking cells to their images.

See Understanding output for details on each folder.

#Loading the features

Both feature sets load into a TglowDataset:

library(tglowr)

# CellProfiler features: reads every <plate>_cells.parquet / <plate>_image.parquet pair
cp <- read_cellprofiler_parquet("output/rr__features/cellprofiler")

# The pipeline's own intensity features: one folder per plate
px <- read_pipeline_parquet("output/rr__features/measurements/unscaled", assay.out="unscaled")

# Combine the objects on cell position
merged <- match_objects_xy_nn(cp, px)

Both take a plates argument to read only some plates, for example plates = c("plate1", "plate2").