Analysis-ready VCF at Biobank scale using Zarr.

Background: Variant Call Format (VCF) is the standard file format for interchanging genetic variation data and associated quality control metrics. The usual row-wise encoding of the VCF data model (either as text or packed binary) emphasises efficient retrieval of all data for a given variant, but accessing data on a field or sample basis is inefficient. Biobank scale datasets currently available consist of hundreds of thousands of whole genomes and hundreds of terabytes of compressed VCF. Row-wise data storage is fundamentally unsuitable and a more scalable approach is needed. Results: We present the VCF Zarr specification, an encoding of the VCF data model using Zarr which makes retrieving subsets of the data much more efficient. Zarr is a cloud-native format for storing multi-dimensional data, widely used in scientific computing. We show how this format is far more efficient than standard VCF based approaches, and competitive with specialised methods for storing genotype data in terms of compression ratios and calculation performance. We demonstrate the VCF Zarr format (and the vcf2zarr conversion utility) on a subset of the Genomics England aggV2 dataset comprising 78,195 samples and 59,880,903 variants, with a 5X reduction in storage and greater than 300X reduction in CPU usage in some representative benchmarks. Conclusions: Large row-encoded VCF files are a major bottleneck for current research, and storing and processing these files incurs a substantial cost. The VCF Zarr specification, building on widely-used, open-source technologies has the potential to greatly reduce these costs, and may enable a diverse ecosystem of next-generation tools for analysing genetic variation data directly from cloud-based object stores.

Massively parallel quantification of phenotypic heterogeneity in single-cell drug responses.

Single-cell analysis tools have made substantial advances in characterizing genomic heterogeneity; however, tools for measuring phenotypic heterogeneity have lagged due to the increased difficulty of handling live biology. Here, we report a single-cell phenotyping tool capable of measuring image-based clonal properties at scales approaching 100,000 clones per experiment. These advances are achieved by exploiting a previously unidentified flow regime in ladder microfluidic networks that, under appropriate conditions, yield a mathematically perfect cell trap. Machine learning and computer vision tools are used to control the imaging hardware and analyze the cellular phenotypic parameters within these images. Using this platform, we quantified the responses of tens of thousands of single cell­derived acute myeloid leukemia (AML) clones to targeted therapy, identifying rare resistance and morphological phenotypes at frequencies down to 0.05%. This approach can be extended to higher-level cellular architectures such as cell pairs and organoids and on-chip live-cell fluorescence assays.

Defining HLA-II Ligand Processing and Binding Rules with Mass Spectrometry Enhances Cancer Epitope Prediction.

Increasing evidence indicates CD4+ T cells can recognize cancer-specific antigens and control tumor growth. However, it remains difficult to predict the antigens that will be presented by human leukocyte antigen class II molecules (HLA-II), hindering efforts to optimally target them therapeutically. Obstacles include inaccurate peptide-binding prediction and unsolved complexities of the HLA-II pathway. To address these challenges, we developed an improved technology for discovering HLA-II binding motifs and conducted a comprehensive analysis of tumor ligandomes to learn processing rules relevant in the tumor microenvironment. We profiled >40 HLA-II alleles and showed that binding motifs were highly sensitive to HLA-DM, a peptide-loading chaperone. We also revealed that intratumoral HLA-II presentation was dominated by professional antigen-presenting cells (APCs) rather than cancer cells. Integrating these observations, we developed algorithms that accurately predicted APC ligandomes, including peptides from phagocytosed cancer cells. These tools and biological insights will enable improved HLA-II-directed cancer therapies.

Cytokit: a single-cell analysis toolkit for high dimensional fluorescent microscopy imaging.

BACKGROUND: Multiplexed in-situ fluorescent imaging offers several advantages over single-cell assays that do not preserve the spatial characteristics of biological samples. This spatial information, in addition to morphological properties and extensive intracellular or surface marker profiling, comprise promising avenues for rapid advancements in the understanding of disease progression and diagnosis. As protocols for conducting such imaging experiments continue to improve, it is the intent of this study to provide and validate software for processing the large quantity of associated data in kind. RESULTS: Cytokit offers (i) an end-to-end, GPU-accelerated image processing pipeline; (ii) efficient input/output (I/O) strategies for operations specific to high dimensional microscopy; and (iii) an interactive user interface for cross filtering of spatial, graphical, expression, and morphological cell properties within the 100+ GB image datasets common to multiplexed immunofluorescence. Image processing operations supported in Cytokit are generally sourced from existing deep learning models or are at least in part adapted from open source packages to run in a single or multi-GPU environment. The efficacy of these operations is demonstrated through several imaging experiments that pair Cytokit results with those from an independent but comparable assay. A further validation also demonstrates that previously published results can be reproduced from a publicly available multiplexed image dataset. CONCLUSION: Cytokit is a collection of open source tools for quantifying and analyzing properties of individual cells in large fluorescent microscopy datasets that are often, but not necessarily, generated from multiplexed antibody labeling protocols over many fields of view or time periods. This project is best suited to bioinformaticians or other technical users that wish to analyze such data in a batch-oriented, high-throughput setting. All source code, documentation, and data generated for this article are available under the Apache License 2.0 at https://github.com/hammerlab/cytokit .

Anti-CTLA-4 Activates Intratumoral NK Cells and Combined with IL15/IL15Rα Complexes Enhances Tumor Control.

Antibodies targeting CTLA-4 induce durable responses in some patients with melanoma and are being tested in a variety of human cancers. However, these therapies are ineffective for a majority of patients across tumor types. Further understanding the immune alterations induced by these therapies may enable the development of novel strategies to enhance tumor control and biomarkers to identify patients most likely to respond. In several murine models, including colon26, MC38, CT26, and B16 tumors cotreated with GVAX, anti-CTLA-4 efficacy depends on interactions between the Fc region of CTLA-4 antibodies and Fc receptors (FcR). Anti-CTLA-4 binding to FcRs has been linked to depletion of intratumoral T regulatory cells (Treg). In agreement with previous studies, we found that Tregs infiltrating CT26, B16-F1, and autochthonous Braf V600E Pten -/- melanoma tumors had higher expression of surface CTLA-4 (sCTLA-4) than other T-cell subsets, and anti-CTLA-4 treatment led to FcR-dependent depletion of Tregs infiltrating CT26 tumors. This Treg depletion coincided with activation and degranulation of intratumoral natural killer cells. Similarly, in non-small cell lung cancer (NSCLC) and melanoma patient-derived tumor tissue, Tregs had higher sCTLA-4 expression than other intratumoral T-cell subsets, and Tregs infiltrating NSCLC expressed more sCTLA-4 than circulating Tregs. Patients with cutaneous melanoma who benefited from ipilimumab, a mAb targeting CTLA-4, had higher intratumoral CD56 expression, compared with patients who received little to no benefit from this therapy. Furthermore, using the murine CT26 model we found that combination therapy with anti-CTLA-4 plus IL15/IL15Rα complexes enhanced tumor control compared with either monotherapy.

Joint longitudinal and time-to-event models for multilevel hierarchical data.

Joint modelling of longitudinal and time-to-event data has received much attention recently. Increasingly, extensions to standard joint modelling approaches are being proposed to handle complex data structures commonly encountered in applied research. In this paper, we propose a joint model for hierarchical longitudinal and time-to-event data. Our motivating application explores the association between tumor burden and progression-free survival in non-small cell lung cancer patients. We define tumor burden as a function of the sizes of target lesions clustered within a patient. Since a patient may have more than one lesion, and each lesion is tracked over time, the data have a three-level hierarchical structure: repeated measurements taken at time points (level 1) clustered within lesions (level 2) within patients (level 3). We jointly model the lesion-specific longitudinal trajectories and patient-specific risk of death or disease progression by specifying novel association structures that combine information across lower level clusters (e.g. lesions) into patient-level summaries (e.g. tumor burden). We provide user-friendly software for fitting the model under a Bayesian framework. Lastly, we discuss alternative situations in which additional clustering factor(s) occur at a level higher in the hierarchy than the patient-level, since this has implications for the model formulation.

Mutations in an Innate Immunity Pathway Are Associated with Poor Overall Survival Outcomes and Hypoxic Signaling in Cancer.

Complement-mediated cytotoxicity may act as a selective pressure for tumor overexpression of complement regulators. We hypothesize that the same selective pressure could lead to complement alterations at the genetic level. We find that, when analyzed as a pathway, mutations in complement genes occur at a relatively high frequency and are associated with changes in overall survival across a number of cancer types. Analysis of pathways expressed in patients with complement mutations that are associated with poor overall survival reveals crosstalk between complement and hypoxia in colorectal cancer. The importance of this crosstalk is highlighted by two key findings: hypoxic signaling is increased in tumors harboring complement mutations, and hypoxic tumor cells are resistant to complement-mediated cytotoxicity due, in part, to hypoxia-induced expression of complement regulator CD55. The range of strategies employed by tumors to dysregulate the complement system testifies to the importance of this pathway in tumor progression.

Coral: Clear and Customizable Visualization of Human Kinome Data.

Protein kinases represent one of the largest gene families in eukaryotes and play roles in a wide range of cell signaling processes and human diseases. Current tools for visualizing kinase data in the context of the human kinome superfamily are limited to encoding data through the addition of nodes to a low-resolution image of the kinome tree. We present Coral, a user-friendly interactive web application for visualizing both quantitative and qualitative data. Unlike previous tools, Coral can encode data in three features (node color, node size, and branch color), allows three modes of kinome visualization (the traditional kinome tree as well as radial and dynamic force networks), and generates high-resolution scalable vector graphics files suitable for publication without the need for refinement using graphics editing software. Due to its user-friendly, interactive, and highly customizable design, Coral is broadly applicable to high-throughput studies of the human kinome. The source code and web application are available at github.com/dphansti/CORAL and phanstiel-lab.med.unc.edu/Coral, respectively.

MHCflurry: Open-Source Class I MHC Binding Affinity Prediction.

Predicting the binding affinity of major histocompatibility complex I (MHC I) proteins and their peptide ligands is important for vaccine design. We introduce an open-source package for MHC I binding prediction, MHCflurry. The software implements allele-specific neural networks that use a novel architecture and peptide encoding scheme. When trained on affinity measurements, MHCflurry outperformed the standard predictors NetMHC 4.0 and NetMHCpan 3.0 overall and particularly on non-9-mer peptides in a benchmark of ligands identified by mass spectrometry. The released predictor, MHCflurry 1.2.0, uses mass spectrometry datasets for model selection and showed competitive accuracy with standard tools, including the recently released NetMHCpan 4.0, on a small benchmark of affinity measurements. MHCflurry's prediction speed exceeded 7,000 predictions per second, 396 times faster than NetMHCpan 4.0. MHCflurry is freely available to use, retrain, or extend, includes Python library and command line interfaces, may be installed using package managers, and applies software development best practices.

Genomic Features of Response to Combination Immunotherapy in Patients with Advanced Non-Small-Cell Lung Cancer.

Combination immune checkpoint blockade has demonstrated promising benefit in lung cancer, but predictors of response to combination therapy are unknown. Using whole-exome sequencing to examine non-small-cell lung cancer (NSCLC) treated with PD-1 plus CTLA-4 blockade, we found that high tumor mutation burden (TMB) predicted improved objective response, durable benefit, and progression-free survival. TMB was independent of PD-L1 expression and the strongest feature associated with efficacy in multivariable analysis. The low response rate in TMB low NSCLCs demonstrates that combination immunotherapy does not overcome the negative predictive impact of low TMB. This study demonstrates the association between TMB and benefit to combination immunotherapy in NSCLC. TMB should be incorporated in future trials examining PD-(L)1 with CTLA-4 blockade in NSCLC.

Chemotherapy weakly contributes to predicted neoantigen expression in ovarian cancer.

BACKGROUND: Patients with highly mutated tumors, such as melanoma or smoking-related lung cancer, have higher rates of response to immune checkpoint blockade therapy, perhaps due to increased neoantigen expression. Many chemotherapies including platinum compounds are known to be mutagenic, but the impact of standard treatment protocols on mutational burden and resulting neoantigen expression in most human cancers is unknown. METHODS: We sought to quantify the effect of chemotherapy treatment on computationally predicted neoantigen expression for high grade serous ovarian carcinoma patients enrolled in the Australian Ovarian Cancer Study. In this series, 35 of 114 samples were collected after exposure to chemotherapy; 14 are matched with an untreated sample from the same patient. Our approach integrates whole genome and RNA sequencing of bulk tumor samples with class I MHC binding prediction and mutational signatures extracted from studies of chemotherapy-exposed Caenorhabditis elegans and Gallus gallus cells. We additionally investigated the relationship between neoantigens, tumor infiltrating immune cells estimated from RNA-seq with CIBERSORT, and patient survival. RESULTS: Greater neoantigen burden and CD8+ T cell infiltration in primary, pre-treatment samples were independently associated with improved survival. Relapse samples collected after chemotherapy harbored a median of 78% more expressed neoantigens than untreated primary samples, a figure that combines the effects of chemotherapy and other processes operative during relapse. The contribution from chemotherapy-associated signatures was small, accounting for a mean of 5% (range 0-16) of the expressed neoantigen burden in relapse samples. In both treated and untreated samples, most neoantigens were attributed to COSMIC Signature (3), associated with BRCA disruption, Signature (1), associated with a slow mutagenic process active in healthy tissue, and Signature (8), of unknown etiology. CONCLUSION: Relapsed ovarian cancers harbor more predicted neoantigens than primary tumors, but the increase is due to pre-existing mutational processes, not mutagenesis from chemotherapy.

Contribution of systemic and somatic factors to clinical response and resistance to PD-L1 blockade in urothelial cancer: An exploratory multi-omic analysis.

BACKGROUND: Inhibition of programmed death-ligand 1 (PD-L1) with atezolizumab can induce durable clinical benefit (DCB) in patients with metastatic urothelial cancers, including complete remissions in patients with chemotherapy refractory disease. Although mutation load and PD-L1 immune cell (IC) staining have been associated with response, they lack sufficient sensitivity and specificity for clinical use. Thus, there is a need to evaluate the peripheral blood immune environment and to conduct detailed analyses of mutation load, predicted neoantigens, and immune cellular infiltration in tumors to enhance our understanding of the biologic underpinnings of response and resistance. METHODS AND FINDINGS: The goals of this study were to (1) evaluate the association of mutation load and predicted neoantigen load with therapeutic benefit and (2) determine whether intratumoral and peripheral blood T cell receptor (TCR) clonality inform clinical outcomes in urothelial carcinoma treated with atezolizumab. We hypothesized that an elevated mutation load in combination with T cell clonal dominance among intratumoral lymphocytes prior to treatment or among peripheral T cells after treatment would be associated with effective tumor control upon treatment with anti-PD-L1 therapy. We performed whole exome sequencing (WES), RNA sequencing (RNA-seq), and T cell receptor sequencing (TCR-seq) of pretreatment tumor samples as well as TCR-seq of matched, serially collected peripheral blood, collected before and after treatment with atezolizumab. These parameters were assessed for correlation with DCB (defined as progression-free survival [PFS] >6 months), PFS, and overall survival (OS), both alone and in the context of clinical and intratumoral parameters known to be predictive of survival in this disease state. Patients with DCB displayed a higher proportion of tumor-infiltrating T lymphocytes (TIL) (n = 24, Mann-Whitney p = 0.047). Pretreatment peripheral blood TCR clonality below the median was associated with improved PFS (n = 29, log-rank p = 0.048) and OS (n = 29, log-rank p = 0.011). Patients with DCB also demonstrated more substantial expansion of tumor-associated TCR clones in the peripheral blood 3 weeks after starting treatment (n = 22, Mann-Whitney p = 0.022). The combination of high pretreatment peripheral blood TCR clonality with elevated PD-L1 IC staining in tumor tissue was strongly associated with poor clinical outcomes (n = 10, hazard ratio (HR) (mean) = 89.88, HR (median) = 23.41, 95% CI [2.43, 506.94], p(HR > 1) = 0.0014). Marked variations in mutation loads were seen with different somatic variant calling methodologies, which, in turn, impacted associations with clinical outcomes. Missense mutation load, predicted neoantigen load, and expressed neoantigen load did not demonstrate significant association with DCB (n = 25, Mann-Whitney p = 0.22, n = 25, Mann-Whitney p = 0.55, and n = 25, Mann-Whitney p = 0.29, respectively). Instead, we found evidence of time-varying effects of somatic mutation load on PFS in this cohort (n = 25, p = 0.044). A limitation of our study is its small sample size (n = 29), a subset of the patients treated on IMvigor 210 (NCT02108652). Given the number of exploratory analyses performed, we intend for these results to be hypothesis-generating. CONCLUSIONS: These results demonstrate the complex nature of immune response to checkpoint blockade and the compelling need for greater interrogation and data integration of both host and tumor factors. Incorporating these variables in prospective studies will facilitate identification and treatment of resistant patients.

Somatic Mutations and Neoepitope Homology in Melanomas Treated with CTLA-4 Blockade.

Immune checkpoint inhibitors are promising treatments for patients with a variety of malignancies. Toward understanding the determinants of response to immune checkpoint inhibitors, it was previously demonstrated that the presence of somatic mutations is associated with benefit from checkpoint inhibition. A hypothesis was posited that neoantigen homology to pathogens may in part explain the link between somatic mutations and response. To further examine this hypothesis, we reanalyzed cancer exome data obtained from our previously published study of 64 melanoma patients treated with CTLA-4 blockade and a new dataset of RNA-Seq data from 24 of these patients. We found that the ability to accurately predict patient benefit did not increase as the analysis narrowed from somatic mutation burden, to inclusion of only those mutations predicted to be MHC class I neoantigens, to only including those neoantigens that were expressed or that had homology to pathogens. The only association between somatic mutation burden and response was found when examining samples obtained prior to treatment. Neoantigen and expressed neoantigen burden were also associated with response, but neither was more predictive than somatic mutation burden. Neither the previously described tetrapeptide signature nor an updated method to evaluate neoepitope homology to pathogens was more predictive than mutation burden. Cancer Immunol Res; 5(1); 84-91. ©2016 AACR.

pileup.js: a JavaScript library for interactive and in-browser visualization of genomic data.

UNLABELLED: P: ileup.js is a new browser-based genome viewer. It is designed to facilitate the investigation of evidence for genomic variants within larger web applications. It takes advantage of recent developments in the JavaScript ecosystem to provide a modular, reliable and easily embedded library. AVAILABILITY AND IMPLEMENTATION: The code and documentation for pileup.js is publicly available at https://github.com/hammerlab/pileup.js under the Apache 2.0 license. CONTACT: correspondence@hammerlab.org.