CRISPR-GEM

CRISPR-GEM predicts the downstream effects of CRISPR gene editing by constructing a synthetic gene regulatory network to identify optimal CRISPR target genes that modulate experimental group GRNs to motivate transcriptomic shifts toward a target group phenotype.

Key Features:

• Multilayer perceptron architecture: Uses an MLP-based synthetic gene regulatory network to learn regulatory relationships in a black-box approach for accurate prediction of output gene expression from input gene expression.
• Differential gene identification: Identifies input nodes as differentially expressed genes between defined experimental and target cell/tissue types, with output nodes representing target cell/tissue type genes.
• CRISPR-mimetic perturbation scoring: Applies in silico perturbations to each input gene individually and compares resulting model predictions against target group predictions to score and assess each input gene as a CRISPR candidate.
• Cross-cell-type applicability: Functions across a range of cell types rather than being limited to single cell types.
• Broad target scope: Extends beyond transcription factors to enable general CRISPR target gene discovery.

Scientific Applications:

• CRISPR genetic target discovery: Identifies optimal CRISPR target genes for achieving reliable therapeutic effects in cell therapies.
• Gene regulatory network modulation: Modulates experimental group GRNs to motivate transcriptomic shifts toward a target group phenotype.
• Cell therapy development: Enhances CRISPR strategies across a range of cell therapies through informed target gene selection.

Methodology:

CRISPR-GEM constructs a multilayer perceptron-based synthetic gene regulatory network. Input and output nodes are identified as differentially expressed genes between defined experimental and target cell/tissue types. MLP training learns regulatory relationships in a black-box approach, enabling prediction of output gene expression from input gene expression. CRISPR-mimetic perturbations are applied to each input gene individually, and resulting predictions are compared to target group predictions to score CRISPR candidates.