Disease diagnostics using machine learning of immune receptors

Expanding the diagnostic toolbox for complex genetic immune disorders

Laboratory-based immunology evaluation is essential to the diagnostic workup of patients with complex immune disorders, and is as essential, if not more so, depending on the context, as genetic testing, because it enables identification of aberrant pathways amenable to therapeutic intervention and clarifies variants of uncertain significance. There have been considerable advances in techniques and instrumentation in the clinical laboratory in the past 2 decades, although there are still "miles to go." One of the goals of the clinical laboratory is to ensure advanced diagnostic testing is widely accessible to physicians and thus patients, through reference laboratories, particularly in the context of academic medical centers. This ensures a greater likelihood of translating research discoveries into the diagnostic laboratory, on the basis of patient care needs rather than a sole emphasis on commercial utility. However, these advances are under threat from burdensome regulatory oversight that can compromise, at best, and curtail, at worst, the ability to rapidly diagnose rare immune disorders and ensure delivery of precision medicine. This review discusses the clinical utility of diagnostic immunology tools, beyond cellular immunophenotyping of lymphocyte subsets, which can be used in conjunction with clinical and other laboratory data for diagnosis as well as monitoring of therapeutic response in patients with genetic immunologic diseases.

The Type 1 Diabetes T Cell Receptor and B Cell Receptor Repository in the AIRR Data Commons: a practical guide for access, use and contributions through the Type 1 Diabetes AIRR Consortium

Human molecular genetics has brought incredible insights into the variants that confer risk for the development of tissue-specific autoimmune diseases, including type 1 diabetes. The hallmark cell-mediated immune destruction that is characteristic of type 1 diabetes is closely linked with risk conferred by the HLA class II gene locus, in combination with a broad array of additional candidate genes influencing islet-resident beta cells within the pancreas, as well as function, phenotype and trafficking of immune cells to tissues. In addition to the well-studied germline SNP variants, there are critical contributions conferred by T cell receptor (TCR) and B cell receptor (BCR) genes that undergo somatic recombination to yield the Adaptive Immune Receptor Repertoire (AIRR) responsible for autoimmunity in type 1 diabetes. We therefore created the T1D TCR/BCR Repository (The Type 1 Diabetes T Cell Receptor and B Cell Receptor Repository) to study these highly variable and dynamic gene rearrangements. In addition to processed TCR and BCR sequences, the T1D TCR/BCR Repository includes detailed metadata (e.g. participant demographics, disease-associated parameters and tissue type). We introduce the Type 1 Diabetes AIRR Consortium goals and outline methods to use and deposit data to this comprehensive repository. Our ultimate goal is to facilitate research community access to rich, carefully annotated immune AIRR datasets to enable new scientific inquiry and insight into the natural history and pathogenesis of type 1 diabetes.

Learning antibody sequence constraints from allelic inclusion

Antibodies and B-cell receptors (BCRs) are produced by B cells, and are built of a heavy chain and a light chain. Although each B cell could express two different heavy chains and four different light chains, usually only a unique pair of heavy chain and light chain is expressed-a phenomenon known as allelic exclusion. However, a small fraction of naive-B cells violate allelic exclusion by expressing two productive light chains, one of which has impaired function; this has been called allelic inclusion. We demonstrate that these B cells can be used to learn constraints on antibody sequence. Using large-scale single-cell sequencing data from humans, we find examples of light chain allelic inclusion in thousands of naive-B cells, which is an order of magnitude larger than existing datasets. We train machine learning models to identify the abnormal sequences in these cells. The resulting models correlate with antibody properties that they were not trained on, including polyreactivity, surface expression, and mutation usage in affinity maturation. These correlations are larger than what is achieved by existing antibody modeling approaches, indicating that allelic inclusion data contains useful new information. We also investigate the impact of similar selection forces on the heavy chain in mouse, and observe that pairing with the surrogate light chain significantly restricts heavy chain diversity.

Supervised fine-tuning of pre-trained antibody language models improves antigen specificity prediction

Antibodies play a crucial role in adaptive immune responses by determining B cell specificity to antigens and focusing immune function on target pathogens. Accurate prediction of antibody-antigen specificity directly from antibody sequencing data would be a great aid in understanding immune responses, guiding vaccine design, and developing antibody-based therapeutics. In this study, we present a method of supervised fine-tuning for antibody language models, which improves on previous results in binding specificity prediction to SARS-CoV-2 spike protein and influenza hemagglutinin. We perform supervised fine-tuning on four pre-trained antibody language models to predict specificity to these antigens and demonstrate that fine-tuned language model classifiers exhibit enhanced predictive accuracy compared to classifiers trained on pre-trained model embeddings. The change of model attention activations after supervised fine-tuning suggested that this performance was driven by an increased model focus on the complementarity determining regions (CDRs). Application of the supervised fine-tuned models to BCR repertoire data demonstrated that these models could recognize the specific responses elicited by influenza and SARS-CoV-2 vaccination. Overall, our study highlights the benefits of supervised fine-tuning on pre-trained antibody language models as a mechanism to improve antigen specificity prediction.

Building Representation Learning Models for Antibody Comprehension

Antibodies are versatile proteins with both the capacity to bind a broad range of targets and a proven track record as some of the most successful therapeutics. However, the development of novel antibody therapeutics is a lengthy and costly process. It is challenging to predict the functional and biophysical properties of antibodies from their amino acid sequence alone, requiring numerous experiments for full characterization. Machine learning, specifically deep representation learning, has emerged as a family of methods that can complement wet lab approaches and accelerate the overall discovery and engineering process. Here, we review advances in antibody sequence representation learning, and how this has improved antibody structure prediction and facilitated antibody optimization. We discuss challenges in the development and implementation of such models, such as the lack of publicly available, well-curated antibody function data and highlight opportunities for improvement. These and future advances in machine learning for antibody sequences have the potential to increase the success rate in developing new therapeutics, resulting in broader access to transformative medicines and improved patient outcomes.

Ultrasensitive allele inference from immune repertoire sequencing data with MiXCR

Allelic variability in the adaptive immune receptor loci, which harbor the gene segments that encode B cell and T cell receptors (BCR/TCR), has been shown to be of critical importance for immune responses to pathogens and vaccines. In recent years, B cell and T cell receptor repertoire sequencing (Rep-Seq) has become widespread in immunology research making it the most readily available source of information about allelic diversity in immunoglobulin (IG) and T cell receptor (TR) loci in different populations. Here we present a novel algorithm for extra-sensitive and specific variable (V) and joining (J) gene allele inference and genotyping allowing reconstruction of individual high-quality gene segment libraries. The approach can be applied for inferring allelic variants from peripheral blood lymphocyte BCR and TCR repertoire sequencing data, including hypermutated isotype-switched BCR sequences, thus allowing high-throughput genotyping and novel allele discovery from a wide variety of existing datasets. The developed algorithm is a part of the MiXCR software ( https://mixcr.com ) and can be incorporated into any pipeline utilizing upstream processing with MiXCR. We demonstrate the accuracy of this approach using Rep-Seq paired with long-read genomic sequencing data, comparing it to a widely used algorithm, TIgGER. We applied the algorithm to a large set of IG heavy chain (IGH) Rep-Seq data from 450 donors of ancestrally diverse population groups, and to the largest reported full-length TCR alpha and beta chain (TRA; TRB) Rep-Seq dataset, representing 134 individuals. This allowed us to assess the genetic diversity of genes within the IGH, TRA and TRB loci in different populations and demonstrate the connection between antibody repertoire gene usage and the number of allelic variants present in the population. Finally we established a database of allelic variants of V and J genes inferred from Rep-Seq data and their population frequencies with free public access at https://vdj.online .

tidytcells: standardizer for TR/MH nomenclature

T cell receptors (TR) underpin the diversity and specificity of T cell activity. As such, TR repertoire data is valuable both as an adaptive immune biomarker, and as a way to identify candidate therapeutic TR. Analysis of TR repertoires relies heavily on computational analysis, and therefore it is of vital importance that the data is standardized and computer-readable. However in practice, the usage of different abbreviations and non-standard nomenclature in different datasets makes this data pre-processing non-trivial. tidytcells is a lightweight, platform-independent Python package that provides easy-to-use standardization tools specifically designed for TR nomenclature. The software is open-sourced under the MIT license and is available to install from the Python Package Index (PyPI). At the time of publishing, tidytcells is on version 2.0.0.

DeepAIR: A deep learning framework for effective integration of sequence and 3D structure to enable adaptive immune receptor analysis

Structural docking between the adaptive immune receptors (AIRs), including T cell receptors (TCRs) and B cell receptors (BCRs), and their cognate antigens are one of the most fundamental processes in adaptive immunity. However, current methods for predicting AIR-antigen binding largely rely on sequence-derived features of AIRs, omitting the structure features that are essential for binding affinity. In this study, we present a deep learning framework, termed DeepAIR, for the accurate prediction of AIR-antigen binding by integrating both sequence and structure features of AIRs. DeepAIR achieves a Pearson's correlation of 0.813 in predicting the binding affinity of TCR, and a median area under the receiver-operating characteristic curve (AUC) of 0.904 and 0.942 in predicting the binding reactivity of TCR and BCR, respectively. Meanwhile, using TCR and BCR repertoire, DeepAIR correctly identifies every patient with nasopharyngeal carcinoma and inflammatory bowel disease in test data. Thus, DeepAIR improves the AIR-antigen binding prediction that facilitates the study of adaptive immunity.