phenoScreen

Functions for analysing phenotypic screening data, designed to fit in dplyr workflows.

Functions related to plotting and manipulating 96/384 well plates have been moved to the platetools package.

Examples

Normalising feature data against negative control values within plates, then scaling features via a z-score, and replace features with 5 principal components.

data %>%
    group_by(Metadata_plate_name) %>%
    normalise(Metadata_compound, neg_control = "DMSO") %>%
    ungroup() %>%
    scale_features() %>%
    pca(n_components = 5)

---

Average single cell data down to an image mean, then remove redundant feature columns.

data %>%
    group_by(Metadata_image_id) %>%
    squash(mean) %>%
    ungroup() %>%
    remove_correlated(threshold = 0.99)

---

Workflows fit nicely for piping data into ggplot2 for plotting.

data %>%
    group_by(Metadata_plate_name) %>%
    normalise(Metadata_compound, neg_control = "DMSO") %>%
    ungroup() %>%
    scale_features() %>%
    pca(n_components = 5) %>%
    ggplot() +
        geom_point(aes(PC1, PC2)) +
        theme_minimal()

Installation

To install from github:

if (!require(devtools)) install.packages("devtools")
devtools::install_github('Swarchal/phenoScreen')

Caveats

PhenoScreen assumes that metadata and featuredata columns are labelled differently, and that all metadata columns share a common prefix. By default this is Metadata (following CellProfiler convention), though this can be changed globally by changing the settings in options at top of your script after loading phenoScreen.
e.g

library(phenoScreen)
library(dplyr)

options("metadata_prefix") = "New_Metadata_prefix"

# do stuff here
# ...

Or, within each function with the metadata_prefix argument.
e.g

data %>%
    group_by(Metadata_image_id, metadata_prefix = "New_Metadata_prefix") %>%
    squash(mean, metadata_prefix = "New_Metadata_prefix") %>%
    ungroup() %>%
    remove_correlated(threshold = 0.99, metadata_prefix = "New_Metadata_prefix")

N.B: It is also assumed that all columns that do not have the metadata prefix are featuredata, and that all featuredata is numeric.