Lnclocator-imb: An Imbalance-tolerant Ensemble Deep Learning Framework for Predicting Long Non-coding RNA Subcellular...

Area of Science:

• Bioinformatics
• Genetics
• Molecular Biology

Background:

• Long non-coding RNAs (lncRNAs) play crucial roles in gene regulation, including dosage compensation and cell differentiation.
• Determining lncRNA subcellular localization is key to understanding their functions, interactions, and regulatory mechanisms.
• Current computational methods for predicting lncRNA localization need improvement, especially with imbalanced datasets.

Purpose of the Study:

• To develop a novel ensemble deep learning framework, lncLocator-imb, for accurate prediction of lncRNA subcellular localization.
• To enhance the performance of prediction models when dealing with imbalanced biological data.
• To provide a versatile tool for bioinformatics and genetics research.

Main Methods:

• Proposed lncLocator-imb, an ensemble deep learning framework integrating Convolutional Neural Networks (CNN) and Gated Recurrent Units (GRU).
• Incorporated physicochemical pattern features and distributed nucleic acid representation features.
• Utilized the label-distribution-aware margin (LDAM) loss function to address category imbalance during training.

Main Results:

• lncLocator-imb demonstrated robust tolerance to category imbalance, outperforming traditional machine learning models and existing predictors.
• The framework effectively leverages lncRNA sequence information through integrated base classifiers and diverse feature types.
• The proposed approach offers a novel strategy for feature management and handling unbalanced datasets in sequence-based predictions.

Conclusions:

• lncLocator-imb provides a significant advancement in predicting lncRNA subcellular localization, particularly for imbalanced datasets.
• The framework's design offers a versatile resource for various sequence-based prediction tasks in bioinformatics and genetics.
• This study highlights the potential of ensemble deep learning and specialized loss functions for complex biological data analysis.