Sequence to Location: Protein Subcellular Localization Driven by Deep Pretrained Language Model.

Area of Science:

• Computational biology
• Bioinformatics
• Molecular biology

Background:

• Protein subcellular localization is crucial for cellular function, and mislocalization is linked to various diseases.
• Traditional methods for determining protein localization are laborious, time-consuming, and complex.
• There is a need for efficient and accurate computational methods for protein subcellular localization prediction.

Purpose of the Study:

• To develop a novel deep learning-based algorithm, SubLoc, for predicting protein subcellular localization.
• To integrate protein sequence and structural information for enhanced prediction accuracy.
• To provide a more efficient and accurate alternative to traditional experimental methods.

Main Methods:

• Utilized the ProtT5 protein language model to generate protein embedding vectors.
• Constructed a 3D protein structure graph using amino acid contact information and processed it with a graph convolutional network.
• Employed a bidirectional gated recurrent unit with a multi-head attention mechanism to analyze sequence features and integrate data.

Main Results:

• SubLoc demonstrated exceptional performance in predicting protein localization across 10 subcellular compartments.
• The algorithm outperformed existing comparative methods in precision, recall, and MCC values.
• SubLoc showed particularly high accuracy in identifying Cytoplasm and Mitochondrion locations.

Conclusions:

• SubLoc offers a powerful and accurate deep learning approach for protein subcellular localization prediction.
• The integration of sequence and structural data significantly enhances prediction capabilities.
• SubLoc has the potential to accelerate research in cell biology and disease mechanisms.