Prediction of mRNA subcellular localization using deep recurrent neural networks

Area of Science:

• Molecular Biology
• Computational Biology
• Genomics

Background:

• Messenger RNA (mRNA) subcellular localization is vital for post-transcriptional gene regulation.
• RNA trafficking relies on RNA-binding proteins and cis-regulatory elements (zipcodes).
• Current high-throughput methods identify localized RNAs but lack mechanistic detail.

Purpose of the Study:

• To develop a deep learning model for predicting mRNA subcellular localization from sequence.
• To provide insights into mRNA trafficking mechanisms and zipcode identification.

Main Methods:

• Introduced RNATracker, a novel deep neural network.
• Integrated Convolutional Neural Networks (CNN), Long Short-Term Memory (LSTM), and attention layers.
• Utilized both RNA sequence and secondary structure information for predictions.

Main Results:

• RNATracker accurately predicts mRNA distributions across subcellular compartments.
• Demonstrated significantly superior predictive power compared to baseline methods.
• Identified model components for generating mechanistic hypotheses and candidate zipcode sequences.

Conclusions:

• RNATracker offers a powerful computational approach to study mRNA localization.
• The model facilitates the discovery of regulatory elements governing RNA trafficking.
• Provides a foundation for further research into post-transcriptional gene regulation.