A hybrid deep learning framework for WT or mutant peptide prediction using p53 mutation data

Area of Science:

• Molecular Biology
• Genomics
• Bioinformatics

Background:

• p53 protein is a crucial tumor suppressor that maintains genome stability.
• Single amino acid variations (SAVs), particularly hotspot mutations, can lead to oncogenic transformation by impairing p53's DNA-binding and regulatory functions.
• Understanding the impact of these SAVs is vital for elucidating disease mechanisms and cellular processes like apoptosis, DNA repair, and cell cycle regulation.

Purpose of the Study:

• To develop a computational framework for identifying deleterious SAVs in the p53 protein.
• To analyze the structural, sequential, and biochemical effects of p53 variants.
• To provide an interpretable and efficient tool for prioritizing high-risk p53 variants.

Main Methods:

• Utilized quantitative biochemical and microbial descriptors of p53 peptide sequences.
• Employed a deep learning architecture combining a convolutional neural network (CNN) and a 2-layer bidirectional long short-term memory (Bi-LSTM) network with attention mechanisms and ESM-2 embeddings.
• Performed systematic analysis of mutation datasets using features like molecular weight, instability index, hydrophobicity, motif enrichment, and amino acid substitution patterns.
• Validated the model using stratified 5-fold cross-validation and SHAP-based interpretation for feature importance.

Main Results:

• The proposed deep learning model achieved high performance metrics: 0.98 accuracy, 0.98 area under the ROC curve (AUROC), and 0.99 precision-recall AUC.
• Identified key amino acid residues and biochemical factors contributing to the model's predictive power through SHAP analysis.
• Demonstrated the model's ability to classify p53 sequences as wild-type or mutant with high precision.

Conclusions:

• The developed approach offers a robust and interpretable framework for evaluating the functional consequences of p53 variants, including hotspot mutations.
• Provides insights into the complex interplay of structural, sequential, and biochemical factors influencing p53 variant pathogenicity.
• Presents a computationally efficient tool for researchers and clinicians to prioritize high-risk p53 variants for further investigation.