Natural product profiling training set
Description
A collection of morphological fingerprints generated using the Cell Painting protocol with 4 different fluorophores. Probes used were Hoechst (DNA), concanavalin A (golgi), mitotracker (active mitochondria), and phalloidin/WGA (membrane and ER). Cells were treated with a library of ~12,500 different bioactive small molecules, and incubated for 48 hours (full methods in manuscript document). Plates were imaged using an EVOS M7000 microscope, and Cell Profiler (v4.2.1) was used to quantify morphological characteristics/features. Training set fingerprints are compiled here as raw data (.Rdata files for R import), and are used in the GitHub repository to create a database and normalize/QC/process data and train models.
Files
Steps to reproduce
From the manuscript: ---------------------------------- All computer vision used in this work was done using CellProfiler (v4.2.1), the pipeline is available in the GitHub repository. This pipeline is heavily based on the earlier pioneering works, and has been modified for the images acquired in this work. Median summaries for each feature, in each imaging field, were taken to generate phenotypic fingerprints. Fields containing >200 cells were omitted to reduce density-based phenotypic artefacts in the data, and the feature means of the remaining fields, for each well, are used as phenotypic fingerprints for each sample. Positive and negative control compounds for each plate are manually spot-checked, to confirm compound delivery by the laboratory automation, and to ensure that there are no biological concerns present (contamination, seeding density effects, stressed cells, etc).
Institutions
- McMaster UniversityOntario, Hamilton