RNATracker

RNATracker predicts mRNA subcellular localization from nucleotide sequences to infer spatial aspects of post-transcriptional gene regulation.

Key Features:

• Deep learning architecture: Implements a deep neural network integrating Convolutional Neural Networks (CNNs), Long Short-Term Memory networks (LSTMs), and attention layers to capture complex sequence and structural patterns.
• Sequence and structure integration: Accepts primary nucleotide sequence and optional secondary structure annotations as inputs to improve localization prediction.
• Predictive distribution: Infers the distribution of mRNA transcripts across a predefined set of subcellular compartments.
• Mechanistic insight generation: Enables isolation of model components to propose testable mechanistic hypotheses and candidate cis-regulatory "zipcode" sequences related to RNA-binding protein interactions.

Scientific Applications:

• mRNA trafficking analysis: Predicts localization patterns to support studies of mRNA transport and compartmentalization mechanisms.
• RNA-binding protein and zipcode discovery: Aids identification of candidate cis-elements and their potential RNA-binding protein partners that direct localization.
• Post-transcriptional regulation studies: Links sequence and secondary-structure features to functional outcomes of gene expression at subcellular locations.

Methodology:

Uses a deep neural network combining CNNs, LSTMs, and attention layers with inputs of primary sequence and optional secondary structure annotations to predict distributions across predefined subcellular compartments and to enable isolation of model components for hypothesis generation.