NN-RNALoc

NN-RNALoc predicts mRNA sub-cellular localization using a neural network that combines distance-based sub-sequence profiles and protein-protein interaction data to improve localization inference.

Key Features:

• Distance-Based Sub-Sequence Profiles: Represents RNA sequences with distance-based sub-sequence profiles that are more memory- and time-efficient than traditional k-mer frequency methods as k increases.
• Protein-Protein Interaction Integration: Incorporates protein-protein interaction (PPI) information as features to enhance localization prediction accuracy.
• Computational Efficiency: Demonstrates reduced computing time compared to prior models such as mRNALoc and RNATracker.
• Feature Extraction: Extracts features from mRNA sequences and associated protein information for model input.
• Neural Network Architecture: Uses a neural network to process sequence-based and interaction-based inputs to predict cellular location.
• Benchmarking: Evaluated on CeFra-seq and RNALocate benchmark datasets and shows improved performance relative to mRNALoc and RNATracker.

Scientific Applications:

• Gene Regulation Studies: Supports investigation of how mRNA localization contributes to protein targeting and post-transcriptional regulation.
• Developmental Biology Research: Assists studies of mRNA placement relevant to embryonic development and neural dendrite formation.
• Disease Mechanism Exploration: Facilitates research into diseases associated with mislocalized RNAs, including neuromuscular disorders and cancer.

Methodology:

Compute distance-based sub-sequence profiles from mRNA sequences, extract associated protein and PPI features, and input these features to a neural network to predict sub-cellular mRNA localization.