Nucleotide context models outperform protein language models for predicting antibody affinity maturation.

Area of Science:

• Immunology
• Computational Biology
• Bioinformatics

Background:

• Antibodies are key to adaptive immunity, developing as B cell receptors (BCRs) on B cells.
• BCRs undergo affinity maturation, a process involving somatic hypermutation (SHM) and selection, to enhance antigen binding.
• Computational models for affinity maturation have emerged from molecular evolution and language modeling perspectives.

Purpose of the Study:

• To compare the predictive power of molecular evolution and language modeling approaches for antibody affinity maturation.
• To evaluate models of SHM, SHM with selection, and protein language models using BCR repertoire data.
• To introduce EPAM, a framework for benchmarking and advancing antibody evolution models.

Main Methods:

• Compared nucleotide-based molecular evolution models with protein language models.
• Utilized large human BCR repertoire datasets and an antigen-specific mouse experiment.
• Developed and applied a nucleotide-based convolutional neural network for SHM modeling.

Main Results:

• Precise modeling of SHM, incorporating nucleotide context, significantly enhances prediction of affinity maturation.
• A nucleotide-based convolutional neural network modeling SHM outperformed state-of-the-art protein language models.
• Incorporating selection estimates provided only modest improvements in predictive power.

Conclusions:

• Nucleotide context is crucial for accurate modeling of somatic hypermutation in antibody affinity maturation.
• Nucleotide-based models offer superior predictive capabilities compared to current protein language models for BCR evolution.
• The EPAM framework facilitates integrated research into antibody evolution and predictive modeling.