Benchmarking data-driven filtering for denoising of TCRpMHC single-cell data.

Area of Science:

• Immunology
• Bioinformatics
• Computational Biology

Background:

• T cell receptor (TCR)-peptide-MHC (pMHC) interactions are crucial for T cell-mediated immunity.
• High-throughput single-cell sequencing with DNA-barcoded MHC multimers allows T cell specificity studies.
• Data variability and low signal-to-noise ratios challenge current TCR-pMHC specificity analysis.

Purpose of the Study:

• To benchmark and evaluate two denoising methods, ICON and ITRAP, for single-cell TCR-pMHC specificity data.
• To assess the effectiveness of these methods in improving data quality and signal-to-noise ratio.
• To determine which method offers superior performance for analyzing T cell specificities.

Main Methods:

• Application and evaluation of ICON and ITRAP denoising methods on publicly available 10x Genomics immune profiling data.
• Analysis using internal metrics and machine learning models trained on raw and denoised data.
• Comparison of signal-to-noise ratios and data consistency between raw and denoised datasets.

Main Results:

• Both ICON and ITRAP identified approximately 75% of the raw data as noise.
• Denoising with both methods increased the signal-to-noise ratio compared to raw data.
• The ITRAP method showed superior performance in terms of data consistency and overall effectiveness.

Conclusions:

• Denoising is essential for improving data quality in high-throughput TCR-pMHC specificity studies.
• The ITRAP method offers a more robust approach for analyzing T cell specificities.
• Enhanced data quality through denoising is paramount for advancing our understanding of T cell-mediated immunity.