CRISPR-GEM: A Novel Machine Learning Model for CRISPR Genetic Target Discovery and Evaluation.

Area of Science:

• Genomics and Bioinformatics
• Gene Editing Technologies
• Machine Learning in Biology

Background:

• CRISPR gene editing offers precise control over gene expression for cell therapies.
• Effective CRISPR strategies depend on selecting target genes within complex gene regulatory networks (GRNs).
• Existing GRN reconstruction methods are limited to single cell types and transcription factors, overlooking many potential CRISPR targets.

Purpose of the Study:

• To develop a novel machine learning model, CRISPR-GEM, for predicting the downstream effects of CRISPR gene editing.
• To overcome limitations of current GRN models by incorporating diverse gene types beyond transcription factors.
• To identify optimal CRISPR target genes that effectively modulate GRNs towards a desired cell phenotype.

Main Methods:

• Developed a multi-layer perceptron (MLP)-based synthetic GRN model (CRISPR-GEM).
• Input and output nodes identified as differentially expressed genes between experimental and target cell types.
• Trained MLP to learn regulatory relationships and predict gene expression changes following CRISPR-mimetic perturbations.

Main Results:

• CRISPR-GEM accurately predicts downstream effects of CRISPR gene editing.
• The model identifies top-scoring genes that best modulate GRNs to achieve target transcriptomic shifts.
• This approach enables the assessment of a wider range of potential CRISPR target genes.

Conclusions:

• CRISPR-GEM is the first machine learning model designed to predict optimal CRISPR target genes.
• This tool enhances CRISPR strategies by providing informed gene selection for cell therapies.
• The model facilitates improved safety and efficacy in therapeutic applications by optimizing gene editing targets.