Posters - Schedules

Human Q fever is a worldwide zoonotic disease caused by the bacterium Coxiella burnetii with high infectious rate. However, there is no FDA-approved vaccine. A previous study found that the formalin-inactivated PI vaccine (PIV) was more protective than the PII vaccine (PIIV) in guinea pigs. Our research objective is to explore the mechanisms of vaccine-induced protective immunity against C. burnetii infection. For this purpose, we utilized 10x Genomics immune profiling technology to simultaneously assay T cell, B cell receptor sequences, transcriptional profiles, and surface protein expression at the single-cell level. Our bioinformatics analysis leveraged the Seurat Weighted Nearest Neighbor algorithm to integrate transcriptional profiles and surface protein expressions to perform cell clustering. After annotating cell types, we focused on B cells and CD4+ T cells to study adaptive immunity in different vaccination groups. To identify specific vaccine-responding clonotypes, we performed B cells trajectory inference, TCR and BCR V(D)J usage, and clonotype analysis. This study helps understand the mechanism of vaccine-induced innate and adaptive immunity against C. burnetii infection and facilitates the discovery of safe and effective vaccines for Q fever.

Skin tissue repair and recovery after injury is important to the healing process and organ homeostasis. The outermost skin layer, the epidermis, is comprised of stem cells which are key players in the recovery process. This repair process is hallmarked by multiple complex biological pathways that involve coordination between growth factors, cytokines, and different cell types including resident immune cells, fibroblasts, keratinocytes, and melanocytes. However, there is limited understanding of the interplay of these cells and secreted products and their relationships dictating the recovery process. The recent development of single-cell RNA-sequencing has provided the power to extract transcriptomic, spatial, and lineage information for each individual cell and affords a unique opportunity to study the interactions occurring during skin wound healing. In this study, we elucidate the precise cell populations and their changes in gene expression caused by injury to the skin using non-injured tissue as a comparator. Using a Seurat workflow, we integrated and analyzed injured and non-injured samples for differentially expressed genes across both conditions using the tweediverse R package. We also investigated cell-cell communications occurring during injury and identified key genes marking biological changes due to injury.

In immuno-oncology, cancer vaccines train the immune system to recognize proteins presented by tumors (neoantigens). The search-space for neoantigens is immensely large and the cost of developing personalized vaccines correspondingly high, therefore systematic and reliable ways of finding promising candidates are needed. We further developed our novel software ImmunoPepper for identifying cancer-specific antigens from complex splicing graphs with patient-specific mutations. It allows for the generation of mutation and splicing-derived peptides. Moreover, ImmunoPepper enables several filtering steps. Finally, ImmunoPepper supports in-silico prediction of binding to the MHC-I complex. We test the ImmunoPepper software on five BRCA samples, and make recommendations for best proteomic validation of the predicted neoantigens. We first build splicing graphs of BRCA and GTEX cohorts and generate splicing-derived neoantigens from the samples. Then, we filter the candidates against a cohort of normal samples, and we set a threshold of RNA expression support in the sample, as well as a threshold for recurrence in a cohort of 1102 BRCA samples. We combine the latter filters, leading to experimental sets of candidates. We search the candidates against MS spectra and conclude about the validation rate of the putative neoantigens generated by ImmunoPepper as a function of the filtering strategy.

Untreated lyme disease can lead to neurological, cardiac, and dermatological complications. Rapid diagnosis of the erythema migrans (EM) rash, a characteristic symptom of Lyme disease, is therefore crucial to early diagnosis and treatment. In this study, we aim to utilize deep learning frameworks including Tensorflow and Keras to create deep convolutional neural networks (DCNN) to detect images of acute Lyme Disease from images of erythema migrans. This study uses a custom database of erythema migrans images of varying quality to train a DCNN capable of classifying images of EM rashes vs non-EM rashes. Images from publicly available sources were mined to create an initial database. Machine based removal of duplicate images was then performed, followed by a thorough examination of all images by a clinician. The resulting database was combined with images of confounding rashes and regular skin, resulting in a total of 683 images. This database was then used to create a DCNN with an accuracy of 93% when classifying images of rashes as EM vs non EM. Finally, this model was converted into a web and mobile application to allow for rapid diagnosis of EM rashes by both patients and clinicians.

The causal cell types in many autoimmune diseases, including atopic dermatitis (AD), are still unclear. Identifying these disease-relevant cell types is critical for the choice of proper experimental model system, understanding the disease etiology and facilitating the development of targeted therapies.

To answer this question, we utilize GWAS SNP enrichment analysis which connects GWAS signals with cell-type specific functional annotations to prioritize the cell types. To be more specific, we applied LDSC-SEG with epigenetic annotations of immune cells isolated from healthy donors, on GWAS summary statistics to identify the disease relevant cells in AD. The results indicate variants associated with AD are mainly enriched with T cells including Th2 cells.

Moreover, we constructed cell-type programs and disease-progression programs for each cell type using scRNA-seq data from the skin samples of healthy controls and AD patients, and developed MAGMA_sc, a methodology built on MAGMA that is tailored for scRNA-seq data, to combine these RNA expression-based programs and GWAS summary statistics. The results revealed additional enrichment of keratinocytes during disease progression in AD.

Together these results demonstrate GWAS SNP enrichment analysis is a powerful bioinformatic strategy to leverage epigenetic and transcriptomic data for interpreting GWAS signals and identifying disease-relevant cell types.

Rheumatoid arthritis (RA) is a chronic, inflammatory and autoimmune disease affecting 1% of the worldwide population. It is characterized by symptomatic flares during which significant inflammation and destruction of the joints appear. The pathophysiology of the disease remains poorly understood and the available treatments can only alleviate its symptoms and rarely induce long-term remission. We currently lack effective tools to predict the course and response of RA to treatments.

RA is a heterogeneous entity resulting in not all patients responding to the same treatments. We posit that this difference is due to a plurality of immune alterations causing distinct immune cell expression profiles namely the RA endophenotypes.

To conduct this study, single-cell RNA sequencing data have been generated from blood mononuclear cell samples of patients presenting with RA, prior to the initiation of treatments. Using bioinformatics methods we aim to discriminate and define specific RA endophenotypes.

The long-term objectives of the project will be to study the possible correlation between immune endophenotypes and response to treatment. Developing a tool capable of linking a new expression profile to a characterized endophenotype (e.i. aggressiveness or response to treatment) would pave the way for personalized medicine in RA.

The T cell receptor (TCR) provides an early, high-resolution metric to ascertain the immune response to vaccinations, with the complete TCR repertoire reflecting the T cells present in an individual. To this end, we developed Tseek, an unbiased and accurate method for profiling the TCR repertoire and used it to explore the T cell response to both the COVID-19 mRNA vaccine (n=9) and the seasonal inactivated Influenza vaccine (n=5) at several time points. The COVID-19 vaccine elicited a broad T cell response involving multiple expanded clones, whereas the Influenza vaccine elicited a narrower response involving fewer clones. Many distinct T cell clones responded at each sample, providing temporal details lacking in the antibody measurements. Neutralizing antibody titers were also measured in the covid vaccine samples, and the TCR responses broadly presaged the antibody responses. The physical symptoms following vaccinations did not reflect the TCR/antibody responses. The TCR repertoire is an individual fingerprint: donors of blood samples taken years apart could be identified solely based upon their TCR repertoire. These results demonstrate the promise of non-invasive (PBMCs) monitoring of the TCR repertoire, as an early, sensitive measure of the adaptive immune response to vaccination, to help various aspects of vaccine design.

Single-cell RNA sequencing (scRNA-seq) allows us to distinguish clusters/subsets of cells with distinct functions. Given the growing size and complexity of the produced datasets, powerful and robust models are needed to identify clusters across different platforms and experimental settings. A new deep learning architecture named Transformers are particularly powerful in learning meaningful and interpretable representations (encodings) of data. Here, we present scPerformer that utilizes Transformers for classifying cell types from scRNA-seq data. To evaluate our method, we first showed that it has comparable performance to state-of-the-art models on smaller datasets but tends to outperform them as the number of cells and/or cell types increases. We then tested our model on a large dataset (~800k cells) of activated CD4+ T cells pre-sorted into different memory subsets. scPerformer distinguished fine-resolution T-cell subsets accurately. Next, we applied scPerformer on two large-scale CITE-seq datasets of circulating immune cells from COVID-19 patients showing that it recapitulates clusters obtained using surface protein expression using only gene expression. Lastly, by attributing the prediction of the scPerformer model to its input features (genes) using the Integrated Gradients, we were able to extract the important genes that drive the correct classification and distinction between different cell subsets.

Similarity in T-cell receptor (TCR) sequences implies shared antigen specificity between receptors, and could be used to discover novel therapeutic targets. However, existing methods cluster clustering T-cell receptor sequences by similarity are computationally inefficient, making them impractical to use on the ever-expanding datasets of the immune repertoire. Here, we developed GIANA (Geometric Isometry based TCR AligNment Algorithm) a computationally efficient tool for this task that provides the same level of clustering specificity as TCRdist at 600 times its speed, and without sacrificing accuracy. GIANA also allows the rapid query of large reference cohorts within minutes. Using GIANA to cluster large-scale TCR datasets provides candidate disease-specific receptors, and provides a new solution to repertoire classification. Querying unseen TCR-seq samples against an existing reference differentiates samples from patients across various cohorts associated with cancer, infectious and autoimmune disease. Our results demonstrate how GIANA could be used as the basis for a TCR-based non-invasive multi-disease diagnostic platform.

Alternative splicing (AS) plays a major role in the differentiation of immune cells during an immune response as 29% of AS genes are specific to the immune system. Although the role of AS is extensively investigated in T cells, its role in B cell activation is less characterised. We sought to develop a long-read technology ONT workflow to understand post-transcriptional regulation at both gene and isoform levels of human germinal centre (GC) B cells. As one of the challenges of ONT is the accurate computational analysis of isoforms, we developed ‘Nexons’ pipeline to identify the differentially spliced transcript variants using long-read sequencing. An in-depth analysis of splicing regulators with the Nexons revealed the differential regulation of the poison exon (PE) in splicing regulators (e.g. SRSF3) in GC B cells. In GC B cells, PEs of the splicing factors were preferentially spliced out whereas naïve B cells expressed isoforms carrying PE, leading to nonsense-mediated mRNA decay. Moreover, we identified novel spliced variants of these genes, which were undetectable due to the limitations of short-read data. Altogether, our findings validate the combination of Nexons with Smart-seq2 adapted ONT RNA-sequencing workflow as a suitable method for the identification and quantification of complex isoforms.

T cells are a pivotal part of the adaptive immunity to control infectious diseases and cancer through the recognition of pathogenic peptides by the T-cell receptor (TCR). Yet, their phenotypic and cellular differentiation in the context of the cell’s TCR, and therefore the fate of the T cell is still poorly understood. Rapidly developing single-cell technologies enable the measurement of multiple biological modalities, including transcriptome, and TCR sequences at the same time offering unprecedented insights into the molecular behavior of T cells. Even though both modalities have been shown to be interlinked, they are often analyzed separately, potentially missing the interdependency between function and cell state. Therefore, we developed mvTCR - a multiview Variational Autoencoder capable of fusing information of both modalities from paired single-cell data. mvTCR captures T-cell function and state better than unimodal models, while seamlessly scaling to atlas-level dataset sizes. We showcased mvTCR on a dataset containing SARS-CoV-2 patients finding separated clusters of expanded clonotypes with similar receptors and gene profiles, which were unobservable at the transcriptomic level. Based on these results, we envision that mvTCR will greatly benefit the analysis of large-scale T-cell datasets and thereby help to deepen our understanding of adaptive immunity.

Cytokines of the common gamma chain (CGC) family play fundamental roles in the life cycle and the signaling of immunocytes. To dissect the CGC cytokine signaling cascade across immunocytes, we systematically measured the primary transcriptional responses of fourteen immunocytes to the six main cytokines. It is well known that CGC cytokine signaling is routed through the common g-chain receptor, as well as specific receptor subunits which, as a result, activate transcription of specific and common target genes in different cell types. Unexpectedly though, we also observed extensive transcript down-regulation for all combinations of cytokines and immocytes in our data.
To determine the underlying factors that orchestrate these contrasting responses to cytokine signaling, we trained a multi-task convolutional neural network (CNN) on sequences from different genomic regions (upstream-flanking, 5’UTR, exons, introns, 3’UTR, downstream-flanking). Interestingly, coding sequences and the 3’UTR contained the most predictive signals across all models, and adding other regions only improved the performance slightly, suggesting that changes in gene expression upon cytokine stimuli are largely regulated on the post-transcriptional level. Importance scoring of the convolutional kernels identified commonly learned kernels across different folds, initializations, and frequently also for cytokines across different cell types, consistent with their shared regulatory cascades.

Various candidates have been suggested for peripheral monitoring of the tumor microenvironment (TME). Examples include circulating tumor DNA and circulating tumor cells. Paired transcriptome and TCRαβ repertoire sequencing enable characterization of peripheral T cells alongside tumor infiltrating lymphocytes (TILs). A recent study in human metastatic melanoma demonstrated that effector and cytotoxic signatures from blood T cells mirror concurrently sequenced TILs, suggesting tumor-matched clonotypes as a candidate for peripheral monitoring of the TME. We hypothesize that viability of this cell subpopulation for monitoring cancers depends on consistent characterization and association with dynamics of the TME. Here, we analyze a transcriptome and TCRαβ repertoire dataset from a novel metastatic melanoma cohort. We identify differentially abundant cell populations in matched clonotypes, exposing phenotypically separable subpopulations. Next, we examine splicing profiles of tumor-matched blood T cells (TMB-Ts), uncovering upregulated and downregulated genes dynamically associated with clonal expansion of TMB-Ts. We investigate cell surface markers expressed within TMB-Ts, leveraging gene regulatory pathways within co-regulatory networks of cell surface markers. We further these aims using novel methods in differential abundance analysis. These results clarify the immunological relatedness between the tumor and systemic immune environments.

T cell exhaustion in the tumor immune microenvironment presents a challenge to the efficacy of immune checkpoint inhibitor therapies used in various late-stage cancers. These treatments are highly effective in a subset of patients but fail for others, via mechanisms that are not fully understood. Identifying gene expression and regulation patterns in immune cell subpopulations within the tumor immune microenvironment that contribute to T cell exhaustion poses an analytical challenge, even with the availability of high-resolution expression profiles from scRNA-seq data. We have constructed an analysis framework to order samples by degree of exhaustion in tumor-infiltrating CD8 T cells and identify immune cell type-specific gene regulatory network patterns associated with this trajectory. In this ongoing work, we apply this framework to scRNA-seq data from human skin tumor samples. Furthermore, we will compare the T cell exhaustion trajectory in tumors to an analogous trajectory in chronic viral infection to identify shared and distinct immune cell activity associated with T cell exhaustion under these two conditions. Considering the behavior of other immune cell types, in addition to tumor cells and the CD8 T cells themselves, will expand our current understanding of this biological process that underpins the effective treatment of late-stage cancers.

Identifying T-cell epitopes is important for improving our understanding of cellular immunity and assisting in the development of peptide-based vaccines, therapeutics, and diagnostics. Human T-cell immunogenicity requires the binding of peptides derived from antigens to HLA molecules, and numerous data-driven prediction tools, have been developed for binding prediction. However, it remains challenging due to the polymorphic nature of HLA class II molecules and variations in peptide length.
The presented work tests the performance of support vector machine (SVM) models trained for allele-specific binding prediction with a comprehensive dataset downloaded from the IEDB database. The SVM models aim to classify binders and non-binders based on their peptide sequences and derived features. In developing the SVM model, we take advantage of underlying periodicities in physicochemical properties along the sequence of a peptide that have been shown to be predictive. Once the physicochemical descriptors are generated, Fourier transforms are then applied to be able to encode peptide sequences of varying lengths. In training and testing the model, cross validation and grid search are applied across multiple train and test datasets, and a feature selection algorithm is also incorporated into the model development to identify an essential set of predictive features.

Advances in immunotherapy and immune checkpoint inhibitors (ICIs) have transformed the standard of care in many cancer types including melanoma. While ICIs have improved the overall outcome of a melanoma patient, a proportion develop primary or secondary tumor resistance. Therefore, a need exists to develop predictive biomarkers to better select patients for ICI therapy. Biomarkers based on genomics and transcriptomics have been investigated, however, most have not been systematically evaluated across different cohorts to determine their reproducibility in metastatic melanoma. We evaluated 28 published biomarkers of ICIs based on gene expression signatures in 8 published studies with available RNA-sequencing data. We found that gene expression-based signatures developed from IFN-γ-responsive genes and T-cell markers in the tumor immune microenvironment are generally predictive of ICIs responders in these patients. Additionally, we identified that these biomarkers have higher predictive values in on-treatment samples as compared to pre-treatment samples in metastatic melanoma. From gene set enrichment analysis, the responders were enriched with inflammation, immune activation, and infiltrated cytotoxic T-cells. Conversely, non-responders were enriched with stromal related cell types and T helper 17 cells. In summary, future biomarker development in ICIs could benefit from integrating multi-omics data to realize personalized therapeutic approach for melanoma patients.

Malaria, mainly caused by Plasmodium falciparum (Pf) species, is a major public health burden with over 400,000 deaths in 2019. The Pf circumsporozoite protein (PfCSP) is the most prevalent sporozoite
antigen, contains a multiply repeated ‘NANP’-tetrapeptide, and uses both sequence degeneracy and structural diversity to evade the immune response. Infusion of PfCSP-directed antibody, CIS43, is effective at preventing malaria infection for up to 9 months. Improved CIS43 are eagerly sought, but the molecular basis mechanism to improve the repeat-targeting antibodies has been unclear. Here, a humanized mouse model where B cells expressed inferred germline CIS43 (iGL-CIS43) antibody was created and used to immunize and to gain informatic insights leading to antibody variants with improved protective efficacy. Based upon informatic sieving of over a hundred mouse derived variants, variants were designed that incorporated somatic hypermutations from the best mouse derived variants. One such improved antibody, iGL-CIS43.D13, exhibited ~7-fold improvement over CIS43 and appears to be the current best in class antibody. Structural and in Silico analysis of antibodies elicited by junctional peptide immunization revealed specific details of improved protective efficacy. Thus, informatics combined antibodies from this mouse model can design highly potent anti-malarial antibodies with improved therapeutic potential.

Rheumatoid Arthritis (RA) is an autoimmune disease causing synovial tissue inflammation. Recent work by the AMP-RA consortium (Zhang et al BioRxiv) has defined 77 RA tissue fine-grain cell states using transcriptional and proteomic data. Many states have been shown to play a pathogenic role in mediating inflammation with implications for treatment response and prognosis. The epigenetic relationship between and regulation of these states remains undefined. We used 12 single nucleus ATAC/RNA multiome and 18 scATAC-seq datasets from disaggregated RA/OA patient synovial tissue to characterize their epigenetic cell states. Our study suggested that broad cell types are generally concordant between modalities, but fine-grain transcriptional cell states are hard to distinguish at the epigenetic level. For example, while 9 fibroblast cell states have been observed transcriptionally, only 4 epigenetically defined cell states were seen. We defined combinations of TFs associated with these broader cell states that may explain the regulation underlying the discrepancy between modalities. Within the AMP-RA known transcriptional cell states, we see instances of both 1-1 and 1-many relationships with epigenetic cell states. These relationships can help explain variable RA inflammation between patients and may suggest treatment plans that target either cell states or exogenous factors (e.g., TFs, cytokines).

Intracellular Cytokine Staining (ICS) is used to assess the nature and magnitude of the immune system’s response to antigens. Most ICS experiments are still analyzed manually.
We used Full Annotation Using Shape-constrained Trees (FAUST) (https://doi.org/10.1016/j.patter.2021.100372) to perform automated gating of protein markers in an ICS study of human T cell response to Mycobacterium tuberculosis antigens. This approach accounted for batch effects in this study by identifying antigen-specific cell subsets on a per-sample basis and also used information from the different stimulation conditions to obtain appropriate cytokine gates. We then used FAUST’s cell subset discovery feature to identify candidate antigen-specific combinations of markers which appear consistently across samples in the experiment. As a parallel approach, we ran Combinatorial Polyfunctionality analysis of Antigen-Specific T-cell Subsets (COMPASS) (https://doi.org/10.1038/nbt.3187) on the FAUST-gated results.
We compare our findings to those of the original publication which used manual gating, and we show that we are able to validate and extend the published findings. Thus, we provide an example of using FAUST as an automated analytical tool for identifying meaningful cell populations in ICS data.