CRISPRpred(SEQ)

CRISPRpred(SEQ) predicts single-guide RNA (sgRNA) on-target activity for CRISPR genome editing applications using sequence-derived features to estimate editing efficacy.

Key Features:

• Traditional machine learning: Employs traditional machine learning techniques rather than deep learning models.
• Hand-crafted sequence features: Uses hand-crafted features derived directly from sgRNA sequences.
• Sequence-based focus: Emphasizes sequence-derived features for on-target activity prediction.
• Explainability and reproducibility: Prioritizes explainable and reproducible predictions over opaque model architectures.
• Benchmark evaluation: Evaluated on benchmark datasets across four different cell lines.
• Comparative performance: Outperformed DeepCRISPR in three of four cell lines with improvements of 2.174%, 6.905%, and 8.119% in on-target prediction accuracy.

Scientific Applications:

• On-target activity prediction: Predicting sgRNA on-target activity to inform CRISPR genome editing experiments.
• Method comparison and benchmarking: Comparative benchmarking of predictive performance across cell lines against methods such as DeepCRISPR.

Methodology:

Uses traditional machine learning models trained on hand-crafted, sequence-derived features extracted from sgRNA sequences and evaluated on benchmark datasets across four cell lines with comparisons to DeepCRISPR.