Deciphering the language of antibodies using self-supervised learning.

An individual's B cell receptor (BCR) repertoire encodes information about past immune responses and potential for future disease protection. Deciphering the information stored in BCR sequence datasets will transform our understanding of disease and enable discovery of novel diagnostics and antibody therapeutics. A key challenge of BCR sequence analysis is the prediction of BCR properties from their amino acid sequence alone. Here, we present an antibody-specific language model, Antibody-specific Bidirectional Encoder Representation from Transformers (AntiBERTa), which provides a contextualized representation of BCR sequences. Following pre-training, we show that AntiBERTa embeddings capture biologically relevant information, generalizable to a range of applications. As a case study, we fine-tune AntiBERTa to predict paratope positions from an antibody sequence, outperforming public tools across multiple metrics. To our knowledge, AntiBERTa is the deepest protein-family-specific language model, providing a rich representation of BCRs. AntiBERTa embeddings are primed for multiple downstream tasks and can improve our understanding of the language of antibodies.

Publisher Correction: Whole-genome sequencing of a sporadic primary immunodeficiency cohort.

An amendment to this paper has been published and can be accessed via a link at the top of the paper.

Whole-genome sequencing of a sporadic primary immunodeficiency cohort.

Primary immunodeficiency (PID) is characterized by recurrent and often life-threatening infections, autoimmunity and cancer, and it poses major diagnostic and therapeutic challenges. Although the most severe forms of PID are identified in early childhood, most patients present in adulthood, typically with no apparent family history and a variable clinical phenotype of widespread immune dysregulation: about 25% of patients have autoimmune disease, allergy is prevalent and up to 10% develop lymphoid malignancies1-3. Consequently, in sporadic (or non-familial) PID genetic diagnosis is difficult and the role of genetics is not well defined. Here we address these challenges by performing whole-genome sequencing in a large PID cohort of 1,318 participants. An analysis of the coding regions of the genome in 886 index cases of PID found that disease-causing mutations in known genes that are implicated in monogenic PID occurred in 10.3% of these patients, and a Bayesian approach (BeviMed4) identified multiple new candidate PID-associated genes, including IVNS1ABP. We also examined the noncoding genome, and found deletions in regulatory regions that contribute to disease causation. In addition, we used a genome-wide association study to identify loci that are associated with PID, and found evidence for the colocalization of-and interplay between-novel high-penetrance monogenic variants and common variants (at the PTPN2 and SOCS1 loci). This begins to explain the contribution of common variants to the variable penetrance and phenotypic complexity that are observed in PID. Thus, using a cohort-based whole-genome-sequencing approach in the diagnosis of PID can increase diagnostic yield and further our understanding of the key pathways that influence immune responsiveness in humans.

Publisher Correction: Telomerecat: A ploidy-agnostic method for estimating telomere length from whole genome sequencing data.

A correction to this article has been published and is linked from the HTML and PDF versions of this paper. The error has been fixed in the paper.

Timing the Landmark Events in the Evolution of Clear Cell Renal Cell Cancer: TRACERx Renal.

Clear cell renal cell carcinoma (ccRCC) is characterized by near-universal loss of the short arm of chromosome 3, deleting several tumor suppressor genes. We analyzed whole genomes from 95 biopsies across 33 patients with clear cell renal cell carcinoma. We find hotspots of point mutations in the 5' UTR of TERT, targeting a MYC-MAX-MAD1 repressor associated with telomere lengthening. The most common structural abnormality generates simultaneous 3p loss and 5q gain (36% patients), typically through chromothripsis. This event occurs in childhood or adolescence, generally as the initiating event that precedes emergence of the tumor's most recent common ancestor by years to decades. Similar genomic changes drive inherited ccRCC. Modeling differences in age incidence between inherited and sporadic cancers suggests that the number of cells with 3p loss capable of initiating sporadic tumors is no more than a few hundred. Early development of ccRCC follows well-defined evolutionary trajectories, offering opportunity for early intervention.

Telomerecat: A ploidy-agnostic method for estimating telomere length from whole genome sequencing data.

Telomere length is a risk factor in disease and the dynamics of telomere length are crucial to our understanding of cell replication and vitality. The proliferation of whole genome sequencing represents an unprecedented opportunity to glean new insights into telomere biology on a previously unimaginable scale. To this end, a number of approaches for estimating telomere length from whole-genome sequencing data have been proposed. Here we present Telomerecat, a novel approach to the estimation of telomere length. Previous methods have been dependent on the number of telomeres present in a cell being known, which may be problematic when analysing aneuploid cancer data and non-human samples. Telomerecat is designed to be agnostic to the number of telomeres present, making it suited for the purpose of estimating telomere length in cancer studies. Telomerecat also accounts for interstitial telomeric reads and presents a novel approach to dealing with sequencing errors. We show that Telomerecat performs well at telomere length estimation when compared to leading experimental and computational methods. Furthermore, we show that it detects expected patterns in longitudinal data, repeated measurements, and cross-species comparisons. We also apply the method to a cancer cell data, uncovering an interesting relationship with the underlying telomerase genotype.

Genomic Evolution of Breast Cancer Metastasis and Relapse.

Patterns of genomic evolution between primary and metastatic breast cancer have not been studied in large numbers, despite patients with metastatic breast cancer having dismal survival. We sequenced whole genomes or a panel of 365 genes on 299 samples from 170 patients with locally relapsed or metastatic breast cancer. Several lines of analysis indicate that clones seeding metastasis or relapse disseminate late from primary tumors, but continue to acquire mutations, mostly accessing the same mutational processes active in the primary tumor. Most distant metastases acquired driver mutations not seen in the primary tumor, drawing from a wider repertoire of cancer genes than early drivers. These include a number of clinically actionable alterations and mutations inactivating SWI-SNF and JAK2-STAT3 pathways.