LSA-ac4C

LSA-ac4C predicts N4-acetylcytidine (ac4C) sites in human mRNA to enable accurate computational identification of epitranscriptomic ac4C loci for studies of mRNA regulation and cancer.

Key Features:

• Target modification: Predicts N4-acetylcytidine (ac4C) modification sites in human mRNA.
• Model architecture: Implements a deep hybrid neural network integrating double-layer Long Short-Term Memory (LSTM) networks and a self-attention mechanism.
• Automated baseline: Leverages automated machine learning technology to establish a reliable baseline.
• Sequence dependency capture: Combines LSTM and self-attention to capture complex patterns and dependencies within mRNA sequences.
• Benchmark performance: Outperforms state-of-the-art methods with improvements of ACC by 2.89%, MCC by 5.96%, and AUROC by 1.53% on an independent test set.

Scientific Applications:

• ac4C site identification: Facilitates computational identification of ac4C sites in mRNA sequences.
• Epitranscriptomics research: Supports studies of mRNA regulation mediated by N4-acetylcytidine.
• Cancer research: Assists investigations into associations between ac4C and various cancers and potential therapeutic targets.
• RNA modification impact: Aids exploration of the implications of RNA modifications in health and disease.

Methodology:

Uses automated machine learning and a deep hybrid neural network that integrates double-layer LSTM networks with a self-attention mechanism, and was evaluated on an independent test set using ACC, MCC, and AUROC.