DM3Loc: multi-label mRNA subcellular localization prediction and analysis based on multi-head self-attention...

Area of Science:

• Molecular Biology
• Bioinformatics
• Computational Biology

Background:

• Subcellular localization of messenger RNAs (mRNAs) is a key regulatory mechanism controlling gene expression.
• This process is crucial for various cellular events, but computational prediction methods are limited, especially for multi-localized mRNAs.

Purpose of the Study:

• To develop an advanced computational method for predicting subcellular localization of mRNAs, particularly for cases with multiple annotations.
• To improve the accuracy and interpretability of mRNA subcellular localization prediction.

Main Methods:

• A multi-head self-attention model, named DM3Loc, was proposed for multi-label mRNA subcellular localization prediction.
• The method was evaluated against existing computational tools and approaches.

Main Results:

• DM3Loc demonstrated superior performance compared to current methods and tools.
• The model provides interpretability, enabling analysis of RNA-binding protein motifs and critical mRNA signals influencing localization.
• Analysis revealed numerous mRNA isoform-specific localizations and significantly enriched gene functions associated with different subcellular locations.

Conclusions:

• DM3Loc offers a powerful and interpretable solution for multi-label mRNA subcellular localization prediction.
• The findings highlight the importance of mRNA localization in cellular function and provide a valuable tool for further research.