DeepTCR is a deep learning framework for revealing sequence concepts within T-cell repertoires.

Area of Science:

• Immunology
• Bioinformatics
• Computational Biology

Background:

• Deep learning excels at pattern recognition, with significant potential in immunogenomics.
• T-cell receptor (TCR) sequencing is crucial for assessing adaptive immune diversity and identifying antigen determinants.
• Existing methods may struggle with the complexity of TCR sequencing data.

Purpose of the Study:

• To introduce DeepTCR, a deep learning framework for analyzing complex TCR sequencing data.
• To demonstrate the utility of deep learning for improved TCR 'featurization'.
• To enhance the classification and extraction of antigen-specific TCRs.

Main Methods:

• Development of a suite of unsupervised and supervised deep learning models.
• Learning a joint representation of TCRs based on CDR3 sequences and V/D/J gene usage.
• Application to multiple human and murine immunogenomic datasets, including single-cell RNA-Seq and T-cell cultures.

Main Results:

• DeepTCR provides an improved 'featurization' of TCRs across diverse datasets.
• Enhanced classification accuracy for antigen-specific TCRs.
• Successful extraction of antigen-specific TCRs from noisy data.

Conclusions:

• Deep learning models, like DeepTCR, are flexible and powerful tools for extracting meaningful information from complex immunogenomic data.
• DeepTCR facilitates both descriptive and predictive analyses in immunogenomics.
• This approach advances the study of immune responses and antigen recognition.