An ensemble learning method combined with multiple feature representation strategies to predict lncRNA subcellular...

Area of Science:

• Molecular Biology
• Bioinformatics
• Computational Biology

Background:

• Long non-coding RNAs (lncRNAs) play critical roles in cellular functions and disease pathogenesis.
• Understanding lncRNA subcellular localization is key to elucidating molecular mechanisms and disease pathways.
• Current experimental methods for determining lncRNA localization are often laborious and time-consuming.

Purpose of the Study:

• To develop an accurate and efficient computational method for predicting lncRNA subcellular localization.
• To leverage heterogeneous multi-source features for improved prediction.
• To address challenges posed by imbalanced datasets in lncRNA research.

Main Methods:

• Developed an ensemble prediction model, lncSLPre.
• Integrated diverse features including sequence composition, physicochemical properties, and structural information.
• Employed the Synthetic Minority Over-sampling Technique (SMOTE) to handle class imbalance.
• Utilized feature selection to refine predictive power.

Main Results:

• The ensemble approach combining heterogeneous features significantly enhanced prediction accuracy.
• SMOTE effectively addressed the imbalanced dataset issue.
• Feature selection proved crucial for removing redundant information.
• lncSLPre demonstrated superior performance over existing methods, with accuracy improvements of up to 13.13%.

Conclusions:

• lncSLPre offers a robust and accurate computational solution for predicting lncRNA subcellular localization.
• The integration of multi-source features and SMOTE is effective for lncRNA localization prediction.
• Accurate lncRNA localization prediction aids in understanding disease mechanisms and developing therapeutic strategies.