Likelihood-Based Inference of B Cell Clonal Families

Area of Science:

• Immunology
• Computational Biology
• Bioinformatics

Background:

• B cell receptor (BCR) sequence diversity is crucial for the human immune system.
• Diversity arises from gene rearrangement and affinity maturation, creating distinct B cell lineages.
• High-throughput sequencing generates large datasets of BCR sequences for research and diagnostics.

Purpose of the Study:

• To develop and validate a computational framework for inferring clonal family membership from BCR sequence data.
• To address the challenge of clustering diverse BCR sequences into their respective lineages.
• To improve the accuracy and efficiency of clonal family inference in immune repertoire analysis.

Main Methods:

• A likelihood-based framework utilizing a multi-hidden Markov Model (multi-HMM) for BCR sequences.
• Development of an agglomerative algorithm for maximum likelihood clustering.
• Implementation of two approximate algorithms balancing speed and accuracy, and a fast algorithm for specific lineage identification.

Main Results:

• The proposed multi-HMM framework significantly outperforms existing clonal family inference methods in simulations.
• The algorithms provide distinct clustering results compared to previous methods when applied to real immune repertoire datasets.
• Validated computational approaches for reconstructing B cell lineage relationships.

Conclusions:

• The multi-HMM based framework offers a powerful new tool for B cell repertoire analysis.
• Accurate clonal family inference is essential for understanding immune responses and developing diagnostics.
• This work advances the field of computational immunology by providing improved methods for lineage reconstruction.