Bodenhofer, U. | Search bioRxiv

bioRxiv2018

Machine learning identifies the immunological signature of Juvenile Idiopathic Arthritis

Juvenile idiopathic arthritis (JIA) is the most common childhood rheumatic disease, with a strongly debated pathophysiological origin. Both adaptive and innate immune processes have been proposed as primary drivers, which may account for the observed clinical heterogeneity, but few high-depth studies have been performed. Here we profiled the adaptive immune system of 85 JIA patients and 43 age-matched controls, identifying immunological changes unique to JIA and others common across a broad spectrum of childhood inflammatory diseases. The JIA immune signature was shared between clinically distinct subsets, but was accentuated in the systemic JIA patients and those patients with active disease. Despite the extensive overlap in the immunological spectrum exhibited by healthy children and JIA patients, machine learning analysis of the dataset proved capable of diagnosis of JIA patients with ~90% accuracy. These results pave the way for large-scale longitudinal studies of JIA, where machine learning could be used to predict immune signatures that correspond to treatment response group.

immunology↗

bioRxiv2017

Learning The High-Dimensional Immunogenomic Features That Predict Public And Private Antibody Repertoires

Recent studies have revealed that immune repertoires contain a substantial fraction of public clones, which are defined as antibody or T-cell receptor (TCR) clonal sequences shared across individuals. As of yet, it has remained unclear whether public clones possess predictable sequence features that separate them from private clones, which are believed to be generated largely stochastically. This knowledge gap represents a lack of insight into the shaping of immune repertoire diversity. Leveraging a machine learning approach capable of capturing the high-dimensional compositional information of each clonal sequence (defined by the complementarity determining region 3, CDR3), we detected predictive public- and private-clone-specific immunogenomic differences concentrated in the CDR3s N1-D-N2 region, which allowed the prediction of public and private status with 80% accuracy in both humans and mice. Our results unexpectedly demonstrate that not only public but also private clones possess predictable high-dimensional immunogenomic features. Our support vector machine model could be trained effectively on large published datasets (3 million clonal sequences) and was sufficiently robust for public clone prediction across studies prepared with different library preparation and high-throughput sequencing protocols. In summary, we have uncovered the existence of high-dimensional immunogenomic rules that shape immune repertoire diversity in a predictable fashion. Our approach may pave the way towards the construction of a comprehensive atlas of public clones in immune repertoires, which may have applications in rational vaccine design and immunotherapeutics.

systems biology↗