spaCR

spaCR links pooled CRISPR-Cas9 perturbations to single-cell spatial phenotypes in high-content microscopy images.

Key Features:

• Deep-learning image classification: Classifies single-cell spatial phenotypes using deep learning.
• Cell segmentation: Segments cells, nuclei, pathogens and organelles with Cellpose.
• Feature extraction and storage: Measures morphology, intensity, texture and colocalization features into a SQLite project.
• Classifier training: Trains image and measurement classifiers.
• Genotyping: Maps sequencing reads to wells and gRNAs for well-level barcode genotyping.
• Effect estimation: Estimates per-guide and per-gene effects by multiple linear regression.

Scientific Applications:

• Pooled CRISPR-Cas9 screening: Links CRISPR perturbations to spatial phenotypes for functional genomics studies.
• Host-pathogen interaction studies: Identifies pathogen effectors that modulate host cellular machinery, as demonstrated for Toxoplasma gondii ESCRT subversion.
• Organelle and protein recruitment analysis: Detects effectors promoting recruitment of host proteins to pathogen-derived membranes such as the parasitophorous vacuole membrane.

Methodology:

spaCR combines deep-learning classification of single-cell spatial phenotypes with well-level barcode genotyping and regression-based deconvolution of gene effects. Cell segmentation uses Cellpose. Features are stored in SQLite. Effect estimation uses multiple linear regression.