DLpTCR: an ensemble deep learning framework for predicting immunogenic peptide recognized by T cell receptor.

Area of Science:

• Immunology
• Computational Biology
• Bioinformatics

Background:

• Predicting immunogenic peptides for T cell receptor (TCR) recognition is vital for vaccine development and cancer immunotherapy.
• Current in silico methods often neglect TCR interactions, leading to costly experimental validation.
• Developing precise computational tools for TCR-peptide interaction prediction is essential.

Purpose of the Study:

• To introduce DLpTCR, a multimodal ensemble deep learning framework for predicting TCR-peptide interactions.
• To evaluate the model's performance and generalizability using COVID-19 and IEDB datasets.
• To provide a computational tool for accelerating immunogenic peptide identification.

Main Methods:

• Developed DLpTCR, a multimodal ensemble deep learning framework.
• Utilized COVID-19 and IEDB datasets for independent model evaluation.
• Assessed prediction accuracy for single/paired TCR chain and peptide interactions.

Main Results:

• DLpTCR achieved an area under the curve of 0.91 for predicting peptide-single TCR chain interactions on COVID-19 data.
• The model demonstrated an overall accuracy of 81.03% on IEDB data for peptide-paired TCR chain interactions.
• Results indicate DLpTCR can learn generalizable interaction rules for TCR-antigen recognition.

Conclusions:

• DLpTCR offers a robust and accurate computational approach for predicting immunogenic peptide-TCR interactions.
• The model's generalizability suggests its utility in diverse immunological applications.
• Accessible webserver and software package facilitate broader adoption and research.