Benchmarking the performance of human antibody gene alignment utilities using a 454 sequence dataset

Area of Science:

• Immunogenetics
• Bioinformatics

Background:

• Immunoglobulin heavy chain (IGH) genes are assembled through VDJ recombination of IGHV, IGHD, and IGHJ gene segments.
• Evaluating the accuracy of computational tools for identifying these gene segments in VDJ rearrangements is challenging without known rearrangement datasets.
• The Stanford_S22 dataset, derived from an individual with a known IGHV, IGHD, and IGHJ genotype, addresses this evaluation gap.

Purpose of the Study:

• To establish a benchmark dataset for evaluating the performance of IGH gene alignment utilities.
• To assess the accuracy of existing computational tools in identifying VDJ rearrangements.

Main Methods:

• Analysis of thousands of VDJ rearrangements from the Stanford_S22 individual.
• Inference of the individual's IGHV, IGHD, and IGHJ genotype from the rearrangement data.
• Evaluation of seven different IGH alignment utilities using the Stanford_S22 dataset.

Main Results:

• The Stanford_S22 dataset was utilized to benchmark seven IGH alignment utilities.
• Error rates, defined as the failure to correctly partition sequences into known S22 genome genes, ranged from 7.1% to 13.7% across the evaluated utilities.

Conclusions:

• The Stanford_S22 dataset provides a valuable resource for benchmarking IGH alignment tools.
• The study highlights variability in the performance of current IGH utility tools, with error rates indicating room for improvement.