LETI 2024 Summer Project: Base Editing Prediction

Overview

This project focuses on leveraging advanced machine learning techniques to predict outcomes for base editing. The primary objectives include:

1. Feature extraction: Identifying the most relevant features for base editing.
2. Model development: Creating and evaluating predictive models to forecast base editing efficiency.

Methods

This project employs a combination of cutting-edge deep learning and feature extraction techniques, including:

• Convolutional Neural Networks (CNNs):

  • Used to analyze sequence data and detect spatial patterns related to base editing.
  • Enabled efficient feature extraction from complex nucleotide sequences.

• Autoencoders:

  • Applied for dimensionality reduction and unsupervised feature learning.
  • Extracted latent features that highlight key properties of input data while reducing noise.

• Graph Neural Networks (GNNs):

  • Designed to model sequence-structure relationships, capturing interactions in RNA or DNA secondary structure.
  • Provided insights into connectivity and dependencies within biological datasets.

Dataset

• Nucleotide sequences annotated with base editing efficiency labels.
• Features include PAM compatibility, editing positions, and nucleotide context.
• Both raw and encoded sequence data were used for training and validation.