CRISPR-Cas9 off-target predictions using CNN and double CNN: Comparative analysis

CRISPR-Cas9, a revolutionary gene-editing technology, faces significant challenges due to off-target effects. These unintended edits can have detrimental consequences, necessitating accurate prediction methods. This research explores the efficacy of Convolutional Neural Networks (CNNs) and Double CNNs in predicting off-target sites, comparing their performance against traditional feed-forward neural networks (FNNs). CRISPR-Cas9 has emerged as a transformative tool for precise gene editing, yet its off-target effects remain a critical concern, potentially causing unintended genetic modifications. Accurate prediction of these off-target sites is essential to enhance the safety and efficacy of CRISPR-Cas9 applications. This study investigates the use of deep learning models, specifically a four-layer feed-forward neural network (FNN), Convolutional Neural Networks (CNNs), and Double CNNs, to predict CRISPR-Cas9 off-target effects. By encoding DNA and guide RNA sequences into numerical vectors, these models can detect subtle mismatches and patterns indicative of off-target activity. The performance of each model is evaluated using the Area Under the Curve (AUC) metric. Results show that CNNs and Double CNNs significantly outperform the FNN, with the Double CNN model achieving the highest AUC score of 0.98. These findings highlight the potential of deep learning approaches to improve the precision of CRISPR-Cas9 off-target predictions, paving the way for safer genetic editing practices. Data Availability at https://github.com.