Biology subjects

Anagnostou, V.

Publications and source records attributed to Anagnostou, V..

8 recordsLinked to original sources

bioRxiv2026

A generative reference grammar of healthy TCR repertoires reveals cancer-associated immune remodeling

T-cell receptor (TCR) repertoires record how adaptive immunity is organized and how cancer and therapy reshape it, but this signal is hard to read: treatment-associated change is entangled with the V(D)J recombination constraints that shape every repertoire. We present CRAFT (Cancer Repertoire Anomaly Finding Transformer), a conditional sequence-to-sequence transformer that learns a nucleotide-level generative grammar of productive TCR-beta CDR3 sequences from healthy donors, conditioned on germline V(D)J assignments. A dual-head decoder mirrors the independence of V-D and D-J recombination, and curriculum training produces embeddings that define a healthy-reference coordinate system in which cancer-associated change appears as structured, measurable deviation. In proof-of-concept applications to a neoadjuvant checkpoint-blockade cohort sampled longitudinally across blood, and to serial single-cell profiling of T-cell subsets during oncolytic immunotherapy, CRAFT geometric metrics capture response-associated remodeling, including shifts in repertoire organization over time. On antigen-labeled benchmarks, CRAFT organizes specificity classes coherently, recovering structure that reflects shared antigen recognition.

cancer biology↗

bioRxiv2023

Elucidating the heterogeneity of immunotherapy response and immune-related toxicities by longitudinal ctDNA and immune cell compartment tracking in lung cancer

PurposeAlthough immunotherapy is the mainstay of therapy for advanced non-small cell lung cancer (NSCLC), robust biomarkers of clinical response are lacking. The heterogeneity of clinical responses together with the limited value of radiographic response assessments to timely and accurately predict therapeutic effect -especially in the setting of stable disease-call for the development of molecularly-informed real-time minimally invasive predictive biomarkers. In addition to capturing tumor regression, liquid biopsies may be informative in evaluating immune-related adverse events (irAEs). Experimental designWe investigated longitudinal changes in circulating tumor DNA (ctDNA) in patients with metastatic NSCLC who received immunotherapy-based regimens. Using ctDNA targeted error-correction sequencing together with matched sequencing of white blood cells and tumor tissue, we tracked serial changes in cell-free tumor load (cfTL) and determined molecular response for each patient. Peripheral T-cell repertoire dynamics were serially assessed and evaluated together with plasma protein expression profiles. ResultsMolecular response, defined as complete clearance of cfTL, was significantly associated with progression-free (log-rank p=0.0003) and overall survival (log-rank p=0.01) and was particularly informative in capturing differential survival outcomes among patients with radiographically stable disease. For patients who developed irAEs, peripheral blood T-cell repertoire reshaping, assessed by significant TCR clonotypic expansions and regressions were noted on-treatment. ConclusionsMolecular responses assist with interpretation of heterogeneous clinical responses especially for patients with stable disease. Our complementary assessment of the tumor and immune compartments by liquid biopsies provides an approach for monitoring of clinical benefit and immune-related toxicities for patients with NSCLC receiving immunotherapy. Statement of translational relevanceLongitudinal dynamic changes in cell-free tumor load and reshaping of the peripheral T-cell repertoire capture clinical outcomes and immune-related toxicities during immunotherapy for patients with non-small cell lung cancer.

genomics↗

bioRxiv2023

SpliceMutr enables pan-cancer analysis of splicing-derived neoantigen burden in tumors

Aberrant alternative splicing can generate neoantigens, which can themselves stimulate immune responses and surveillance. Previous methods for quantifying splicing-derived neoantigens are limited by independent references and potential batch effects. Here, we introduce SpliceMutr, a bioinformatics approach and pipeline for identifying splicing derived neoantigens from paired tumor normal data. SpliceMutr facilitates the identification of tumor-specific antigenic splice variants, predicts MHC-binding affinity, and estimates splicing antigenicity scores per gene. By applying this tool to genomic data from The Cancer Genome Atlas (TCGA), we generate splicing-derived neoantigens and neoantigenicity scores per sample and across all cancer types and find numerous correlations between splicing antigenicity and well-established biomarkers of anti-tumor immunity. Notably, carriers of mutations within splicing machinery genes have higher splicing antigenicity, which provides support for our approach. Further analysis of splicing antigenicity in cohorts of melanoma patients treated with mono- or combined immune checkpoint inhibition suggest that the abundance of splicing antigens is reduced post-treatment from baseline in patients who progress, likely because of an immunoediting process. We also observe increased splicing antigenicity in responders to immunotherapy, which may relate to an increased capacity to mount an immune response to splicing-derived antigens. This new computational tool provides novel analytical capabilities for splicing antigenicity and is openly available for further immuno-oncologic analysis.

cancer biology↗

bioRxiv2022

Lung tumor-infiltrating Treg have divergent transcriptional profiles and function linked to checkpoint blockade response

Regulatory T cells (Treg) are conventionally viewed to suppress endogenous and therapyinduced anti-tumor immunity; however, their role in modulating responses to immune checkpoint blockade (ICB) is unclear. In this study, we integrated single-cell RNAseq/TCRseq of >73,000 tumor-infiltrating Treg (TIL-Treg) from anti-PD-1-treated and treatment naive non-small cell lung cancers (NSCLC) with single cell analysis of tumor-associated antigen (TAA)-specific Treg derived from a murine tumor model. We identified 10 subsets of human TIL-Treg, most of which have high concordance with murine TIL-Treg subsets. Notably, one subset selectively expresses high levels of OX40 and GITR, whose engangement by cognate ligand mediated proliferative programs and NF-kB activation, as well as multiple genes involved in Treg suppression, in particular LAG3. Functionally, the OX40hiGITRhi subset in the most highly suppressive ex vivo and Treg expression of OX40, GITR and LAG3, correlated with resistance to PD-1 blockade. Surprisingly, in the murine tumor model, we found that virtually all TIL-Treg expressing T cell receptors that are specific for TAA fully develop a distinct Th1-like signature over a two-week period after entry into the tumor, down-regulating FoxP3 and up-regulating expression of TBX21 (Tbet), IFN{gamma} and certain pro-inflammatory granzymes. Application of a gene score from the murine TAA-specific Th1-like Treg subset to the human single-cell dataset revealed a highly analogous subcluster that was enriched in anti-PD-1 responding tumors. These findings demonstrate that TIL-Treg partition into multiple distinct transcriptionally-defined subsets with potentially opposing effects on ICB-induced anti-tumor immunity and suggest that TAA-specific TIL-Treg may positively contribute to anti-tumor responses. One-Sentence SummaryWe define 10 subsets of lung cancer-infiltrating regulatory T cells, one of which is highly suppressive and enriched in anti-PD-1 non-responders and the other is Th1-like and is enriched in PD-1 responders.

immunology↗

bioRxiv2022

Deep Neural Networks Predict MHC-I Epitope Presentation and Transfer Learn Neoepitope Immunogenicity

Identifying neoepitopes that elicit an adaptive immune response is a major bottleneck to developing personalized cancer vaccines. Experimental validation of candidate neoepitopes is extremely resource intensive, and the vast majority of candidates are non-immunogenic, making their identification a needle-in-a-haystack problem. To address this challenge, we present computational methods for predicting MHC-I epitopes and identifying immunogenic neoepitopes with improved precision. The BigMHC method comprises an ensemble of seven pan-allelic deep neural networks trained on peptide-MHC eluted ligand data from mass spectrometry assays and transfer learned on data from assays of antigen-specific immune response. Compared with four state-of-the-art classifiers, BigMHC significantly improves the prediction of epitope presentation on a test set of 45,409 MHC ligands among 900,592 random negatives (AUROC=0.9733, AUPRC=0.8779). After transfer learning on immunogenicity data, BigMHC yields significantly higher precision than seven state-of-the-art models in identifying immunogenic neoepitopes, making BigMHC effective in clinical settings. All data and code are freely available at https://github.com/KarchinLab/bigmhc.

bioinformatics↗

bioRxiv2022

Integrative Analysis of Checkpoint Blockade Response in Advanced Non-Small Cell Lung Cancer

Anti-PD-1/PD-L1 agents have transformed the treatment landscape of advanced non-small cell lung cancer (NSCLC). While our understanding of the biology underlying immune checkpoint blockade in NSCLC is still incomplete, studies to date have established predictive roles for PD-L1 tumor expression and tumor mutational burden (TMB). To expand our understanding of the molecular features underlying response to checkpoint inhibitors in NSCLC, we describe here the first joint analysis of the Stand Up 2 Cancer - Mark Foundation (SU2C-MARK) Cohort, a resource of whole exome and/or RNA sequencing from 393 patients with NSCLC treated with anti-PD-(L)1 therapy, along with matched clinical response annotation. We identify a number of associations between molecular features and outcome, including: 1) favorable (e.g., ATM altered), and unfavorable (e.g., TERT amplified) genomic subgroups, 2) distinct immune infiltration signatures associated with wound healing (unfavorable) and immune activation (favorable), and 3) a novel de-differentiated tumor-intrinsic subtype characterized by expression of endodermal lineage genes, immune activation, and enhanced response rate. Taken together, results from this cohort extend our understanding of NSCLC-specific predictors, providing a rich set of molecular and immunologic hypotheses with which to further our understanding of the biology of checkpoint blockade in NSCLC.

genomics↗

bioRxiv2019

High-throughput prediction of MHC Class I and Class II neoantigens with MHCnuggets

Computational prediction of binding between neoantigen peptides and major histocompatibility complex (MHC) proteins is an emerging biomarker for predicting patient response to cancer immunotherapy. Current neoantigen predictors focus on in silico estimation of MHC binding affinity and are limited by low positive predictive value for actual peptide presentation, inadequate support for rare MHC alleles and poor scalability to high-throughput data sets. To address these limitations, we developed MHCnuggets, a deep neural network method to predict peptide-MHC binding. MHCnuggets is the only method to handle binding prediction for common or rare alleles of MHC Class I or II, with a single neural network architecture. Using a long short-term memory network (LSTM), MHCnuggets accepts peptides of variable length and is capable of faster performance than other methods. When compared to methods that integrate binding affinity and HLAp data from mass spectrometry, MHCnuggets yields a fourfold increase in positive predictive value on independent MHC-bound peptide (HLAp) data. We applied MHCnuggets to 26 cancer types in TCGA, processing 26.3 million allele-peptide comparisons in under 2.3 hours, yielding 101,326 unique candidate immunogenic missense mutations (IMMs). Predicted-IMM hotspots occurred in 38 genes, including 24 driver genes. Predicted-IMM load was significantly associated with increased immune cell infiltration (p<2e-16) including CD8+ T cells. Notably, only 0.16% of predicted immunogenic missense mutations were observed in >2 patients, with 61.7% of these derived from driver mutations. Our results provide a new method for neoantigen prediction with high performance characteristics and demonstrate its utility in large data sets across human cancers.\n\nSynopsisWe developed a new in silico predictor of Major Histocompatibility Complex (MHC) ligand binding and demonstrated its utility to assess potential neoantigens and immunogenic missense mutations (IMMs) in 6613 TCGA patients.

immunology↗

bioRxiv2017

Prediction of peptide binding to MHC Class I proteins in the age of deep learning

Binding of peptides to Major Histocompatibility Complex (MHC) proteins is a critical step in immune response. Peptides bound to MHCs are recognized by CD8+ (MHC Class I) and CD4+ (MHC Class II) T-cells. Successful prediction of which peptides will bind to specific MHC alleles would benefit many cancer immunotherapy appications. Currently, supervised machine learning is the leading computational approach to predict peptide-MHC binding, and a number of methods, trained using results of binding assays, have been published. Many clinical researchers are dissatisfied with the sensitivity and specificity of currently available methods and the limited number of alleles for which they can be applied. We evaluated several recent methods to predict peptide-MHC Class I binding affinities and a new method of our own design (MHCnuggets). We used a high-quality benchmark set of 51 alleles, which has been applied previously. The neural network methods NetMHC, NetMHCpan, MHCflurry, and MHCnuggets achieved similar best-in-class prediction performance in our testing, and of these methods MHCnuggets was significantly faster. MHCnuggets is a gated recurrent neural network, and the only method to our knowledge which can handle peptides of any length, without artificial lengthening and shortening. Seventeen alleles were problematic for all tested methods. Prediction difficulties could be explained by deficiencies in the training and testing examples in the benchmark, suggesting that biological differences in allele-specific binding properties are not as important as previously claimed. Advances in accuracy and speed of computational methods to predict peptide-MHC affinity are urgently needed. These methods will be at the core of pipelines to identify patients who will benefit from immunotherapy, based on tumor-derived somatic mutations. Machine learning methods, such as MHCnuggets, which efficiently handle peptides of any length will be increasingly important for the challenges of predicting immunogenic response for MHC Class II alleles.\n\nAuthor SummaryMachine learning methods are a popular approach for predicting whether a peptide will bind to Major Histocompatibility Complex (MHC) proteins, a critical step in activation of cytotoxic T-cells. The input to these methods is a peptide sequence and an MHC allele of interest, and the output is the predicted binding affinity. MHC Class I and II proteins bind peptides of 8-11 amino acids and 16-26 amino acids respectively. This has been an obstacle for machine learning, because the methods used to date can only handle fixed-length inputs. We show that a recently developed technique known as gated recurrent neural networks can handle peptides of variable length and predict peptide-MHC binding as well or better than existing methods, at substantially faster speeds. Our results have implications for the hundreds of MHC alleles that cannot be predicted with current methods.

bioinformatics↗