AACR 2022 Proceedings: Part B April 11-13

By CTI Meeting Technology

The AACR Annual Meeting is the focal point of the cancer research community, where scientists, clinicians, other health care professionals, survivors, patients, and advocates gather to share the latest advances in cancer science and medicine. From population science and prevention; to cancer biology, translational, and clinical studies; to survivorship and advocacy; the AACR Annual Meeting highlights the work of the best minds in cancer research from institutions all over the world.

• Language: English
• Publisher: CTI Meeting Technology
• Release date: May 13, 2022
• ISBN: 9781005543747

Book preview

Proceedings of the AACR

Volume 61 | April 2022

Part B: Monday, April 11 through Wednesday, April 13

TABLE OF CONTENTS

Presentations: Monday, April 11

BIOINFORMATICS / CONVERGENCE SCIENCE / SYSTEMS BIOLOGY

SINGLE-CELL AND SPATIAL PROFILING OF TUMOR CELLS AND THE IMMUNE MICROENVIRONMENT

DATABASE RESOURCES AND WEB APPLICATIONS

NEW ALGORITHMS AND TOOLS FOR DATA ANALYSIS

COMPUTATIONAL FRAMEWORKS AND RESOURCES

MACHINE LEARNING ACROSS CANCER RESEARCH

CHEMISTRY

DRUG DESIGN AND LEAD OPTIMIZATION STRATEGIES TOWARD NOVEL OR DIFFICULT-TO-DRUG CANCER TARGETS

HIGH-THROUGHPUT SCREENING, DRUG DESIGN, AND NATURAL PRODUCTS IN CANCER

APPLIED NANOTECHNOLOGY AND DRUG DELIVERY APPROACHES TOWARD POTENTIAL CANCER THERAPEUTICS

CLINICAL RESEARCH EXCLUDING TRIALS

BIOMARKERS 2

BIOMARKERS PREDICTIVE OF THERAPEUTIC BENEFIT 1

BIOMARKERS PREDICTIVE OF THERAPEUTIC BENEFIT 2

IMMUNE MECHANISMS INVOKED BY OTHER THERAPIES

CIRCULATING TUMOR CELLS AND PROGNOSTIC BIOMARKERS

IMMUNE RESPONSE TO THERAPIES / IMMUNE MONITORING AND CLINICAL CORRELATES

INFLAMMATION / MODIFIERS OF THE TUMOR MICROENVIRONMENT

PEDIATRIC ONCOLOGY: CLINICAL INVESTIGATION

SPATIAL PROTEOMICS AND TRANSCRIPTOMICS

CLINICAL TRIALS

PATIENT SELECTION STRATEGIES FOR MOLECULARLY TARGETED AGENTS IN CLINICAL TRIALS

NEOADJUVANT AND PERIOPERATIVE IMMUNOTHERAPY CLINICAL TRIALS

PHASE II CLINICAL TRIALS 1

PHASE II TRIALS IN PROGRESS

PHASE I CLINICAL TRIALS 1

PHASE II CLINICAL TRIALS 2

EXPERIMENTAL AND MOLECULAR THERAPEUTICS

LATE-BREAKING RESEARCH: EXPERIMENTAL AND MOLECULAR THERAPEUTICS 1 / CHEMISTRY

BIOMARKERS, MODELS, AND MECHANISMS

EMERGING NEW ANTICANCER AGENTS

BIOLOGICAL THERAPEUTIC AGENTS

COMBINATIONS

DRUG CONJUGATES / BISPECIFIC ANTIBODIES

LUNG CANCER TREATMENT RESISTANCE

MECHANISMS OF DRUG ACTION 3

PRECLINICAL AND CLINICAL PHARMACOLOGY

ANTIBODY-DRUG CONJUGATES

BREAST CANCER DRUG RESISTANCE AND NOVEL TARGETS

DRUG TARGETS

MECHANISMS OF DRUG ACTION 1

MECHANISMS OF DRUG ACTION 2

NEW TECHNOLOGIES FOR DRUG DISCOVERY

IMMUNOLOGY

LATE-BREAKING RESEARCH: IMMUNOLOGY 1

CANCER BIOLOGY AND TUMOR IMMUNITY

IMMUNE MECHANISMS INVOKED BY OTHER THERAPIES

INFLAMMATION, TUMOR INITIATION AND PROGRESSION

MOLECULAR SIGNALING AND METABOLIC AND EPIGENETIC REGULATION IN ADAPTIVE TUMOR IMMUNITY

TUMOR ANTIGENS, ANTIGEN PRESENTATION, AND TUMOR IMMUNITY

IMMUNE RESPONSE TO THERAPIES 1

IMMUNOMODULATORY AGENTS AND INTERVENTIONS 1

INFLAMMATION, IMMUNITY, AND CANCER

INNATE IMMUNITY TO CANCER

MOLECULAR/CELLULAR BIOLOGY AND GENETICS

LATE-BREAKING RESEARCH: MOLECULAR/CELLULAR BIOLOGY AND GENETICS 1

CANCER GENOMICS AND BIOLOGY

CANCER METABOLISM

CANCER GENOMICS 3

GENE EXPRESSION AND CELL DIFFERENTIATION

GENOMIC INSTABILITY 1

ONCOGENES AND TUMOR SUPPRESSOR GENES 1

ONCOGENES AND TUMOR SUPPRESSOR GENES 2

POST-TRANSCRIPTIONAL AND TRANSLATIONAL CONTROL

CHROMATIN, ENHANCERS, PROMOTERS, AND REGULATION OF TRANSCRIPTION FACTOR FUNCTION

DIAGNOSTIC AND THERAPEUTIC APPLICATIONS OF MICRORNAS

GENOMIC INSTABILITY 2

GROWTH FACTOR AND RECEPTORS

SHORT AND LONG NONCODING RNA REGULATING CANCER BIOLOGY

TUMOR-MICROENVIRONMENT INTERACTIONS

MULTIDISCIPLINARY

ADVANCING CANCER RESEARCH THROUGH AN INTERNATIONAL CANCER REGISTRY: AACR PROJECT GENIE USE CASES

AACR PROJECT GENIE USE CASES 1

AACR PROJECT GENIE USE CASES 2

POPULATION SCIENCES

DIET, ALCOHOL, TOBACCO USE, AND OTHER LIFESTYLE RISK FACTORS

GENETIC AND ENVIRONMENTAL CANCER RISK FACTORS

PREVENTION / EARLY DETECTION / INTERCEPTION

PRECLINICAL PREVENTION, EARLY DETECTION, AND INTERCEPTION 2

PRECLINICAL PREVENTION, EARLY DETECTION, AND INTERCEPTION 3

REGULATORY SCIENCE AND POLICY

SCIENCE AND HEALTH POLICY / REGULATORY SCIENCE AND POLICY

SURVIVORSHIP

ADVOCATES POSTER SESSION 1

TUMOR BIOLOGY

LATE-BREAKING RESEARCH: TUMOR BIOLOGY 2

MOLECULAR PRINCIPLES OF METASTASIS

NEXT-GEN GENETICALLY ENGINEERED MOUSE MODELS: DEDICATED TO THE MEMORY OF BEATRICE MINTZ

CANCER STEM CELLS

GENETICALLY ENGINEERED MOUSE MODELS 1

GENETICALLY ENGINEERED MOUSE MODELS 2

METASTATIC CELL PLASTICITY: EMT AND STEM CELL

TRANSCRIPTIONAL AND METABOLIC REGULATION OF METASTASIS

TUMOR ADHESION

CAUSES AND CONSEQUENCES OF TUMOR HETEROGENEITY

CELL LINE AND ANIMAL MODELS

HUMANIZED MOUSE MODELS, IMAGING, AND IMMUNO-ONCOLOGY

LEVERAGING SINGLE CELL SEQUENCING TECHNOLOGIES AND NEW MODELS TO STUDY PEDIATRIC MALIGNANCIES

METHODS TO MEASURE TUMOR EVOLUTION AND HETEROGENEITY

TUMOR IMMUNE PROFILING AND SPATIAL ANALYSIS

Presentations: Tuesday, April 12

BIOINFORMATICS / CONVERGENCE SCIENCE / SYSTEMS BIOLOGY

LATE-BREAKING RESEARCH: BIOINFORMATICS, CONVERGENCE SCIENCE, AND SYSTEMS BIOLOGY

APPLICATION OF BIOINFORMATICS TO CANCER BIOLOGY 1

MATHEMATICAL MODELS

APPLICATION OF BIOINFORMATICS TO CANCER BIOLOGY 2

CHEMISTRY

STRUCTURAL AND CHEMICAL BIOLOGY

CLINICAL RESEARCH EXCLUDING TRIALS

LATE-BREAKING RESEARCH: CLINICAL RESEARCH 2

RESISTANCE MECHANISMS AND NEW ADVANCES IN IMMUNOTHERAPEUTICS

ADOPTIVE CELL THERAPY

BIOMARKERS PREDICTIVE OF THERAPEUTIC BENEFIT 3

DIAGNOSTIC BIOMARKERS

CELL-FREE DNA 1

CELL-FREE DNA 2

COMBINATION IMMUNOTHERAPIES / THERAPEUTIC ANTIBODIES

TRANSLATIONAL AND CLINICAL SCIENCE

TRANSLATIONAL RESEARCH: MOLECULAR AND CLINICAL

CLINICAL TRIALS

IMMUNOTHERAPY COMBINATION STRATEGIES IN CLINICAL TRIALS

COMBINATION IMMUNOTHERAPY CLINICAL TRIALS

PHASE I CLINICAL TRIALS 2

PHASE I TRIALS IN PROGRESS 1

PHASE III CLINICAL TRIALS

PHASE III TRIALS IN PROGRESS

COVID-19 AND CANCER

COVID-19 AND CANCER

ENDOCRINOLOGY

MOLECULAR, PRECLINICAL, AND CLINICAL ENDOCRINOLOGY

EXPERIMENTAL AND MOLECULAR THERAPEUTICS

PROMISING NEW THERAPIES

TARGETING THE RAS ONCOGENE

CELL CYCLE, REPLICATION INHIBITORS, AND IMMUNOTHERAPY AGENTS

DNA DAMAGE RESPONSE AND REPAIR

GASTROINTESTINAL CANCER TREATMENT RESISTANCE

NOVEL TARGETS AND PATHWAYS

SIGNALING PATHWAY INHIBITORS

DRUG RESISTANCE AND REVERSAL OF RESISTANCE

EPIGENETIC TARGETS

GENE AND VECTOR-BASED THERAPY

PRECLINICAL RADIOTHERAPEUTICS

TYROSINE KINASE AND PHOSPHATASE INHIBITORS

IMMUNOLOGY

ADOPTIVE CELL THERAPY

NOVEL CELLULAR MECHANISMS FOR IMMUNE EVASION IN CANCER

ADOPTIVE CELL THERAPY 3

ADOPTIVE CELL THERAPY 4

THERAPEUTIC ANTIBODIES 1

THERAPEUTIC ANTIBODIES 2

IMMUNOMODULATORY AGENTS AND INTERVENTIONS 2

MICROBIOME AND CANCER

MODIFIERS OF THE TUMOR MICROENVIRONMENT

VACCINES: ONCOLYTIC AND PROPHYLACTIC

MOLECULAR/CELLULAR BIOLOGY AND GENETICS

LATE-BREAKING RESEARCH: MOLECULAR/CELLULAR BIOLOGY AND GENETICS 2

CANCER EPIGENETICS: FROM NUCLEOTIDES TO 3D GENOME STRUCTURE

CANCER GENOMICS AND THERAPEUTIC VULNERABILITIES

CANCER GENOMICS 4

CELL CYCLE CONTROL AND CELL CYCLE REGULATORS AS THERAPEUTIC TARGETS

METABOLIC THERAPIES, DETECTION, AND TECHNOLOGIES

ONCOGENIC TRANSCRIPTION FACTORS

TUMOR-ASSOCIATED METABOLIC CHANGES

CANCER GENOMICS 5

CHROMATIN MODIFIERS: FROM MECHANISM TO THERAPEUTIC OPPORTUNITY

GTPASE AND UBIQUITIN SIGNALING

NON-APOPTOTIC CELL DEATH / AUTOPHAGY

NUTRIENTS, DIET, AND METABOLIC REQUIREMENTS OF CANCER

MULTIDISCIPLINARY

MINISYMPOSIUM: LATE-BREAKING RESEARCH

POPULATION SCIENCES

LATE-BREAKING RESEARCH: POPULATION SCIENCES / PREVENTION, EARLY DETECTION, AND INTERCEPTION

CANCER HEALTH DISPARITIES

BIOMARKERS OF ENDOGENOUS OR EXOGENOUS EXPOSURES, EARLY DETECTION, AND BIOLOGIC EFFECTS

SURVIVORSHIP AND PROGNOSIS BIOMARKERS

PREVENTION / EARLY DETECTION / INTERCEPTION

PREMALIGNANT LESIONS, INTERVENTION, AND HEALTH DISPARITIES

RISK ASSESSMENT, BIOMARKERS, EARLY DETECTION, AND SCREENING

SURVIVORSHIP

ADVOCATES POSTER SESSION 2

TUMOR BIOLOGY

AGE AND THE TUMOR MICROENVIRONMENT

STROMAL-EPITHELIAL INTERACTIONS AND PDAC PROGRESSION

CELLULAR AND MOLECULAR PROCESSES OF METASTASIS 1

CELLULAR AND MOLECULAR PROCESSES OF METASTASIS 2 / TUMOR DORMANCY

IN VIVO IMAGING 1

IN VIVO IMAGING 2

OMICS APPROACHES: DISEASE CLASSIFICATION, BIOMARKERS OF RESPONSE, AND OUTCOME PREDICTION IN PEDIATRIC ONCOLOGY

ORGAN-SPECIFIC TUMOR MICROENVIRONMENT 1

ORGAN-SPECIFIC TUMOR MICROENVIRONMENT 2

MICROBIOME

ORGANOID-BASED MODELS

PATIENT-DERIVED XENOGRAFTS

SECRETED SOLUBLE FACTORS AND EXOSOMES IN THE MICROENVIRONMENT

STEM CELLS AND REGULATORY PATHWAYS IN CANCER

STROMAL TUMOR INTERACTIONS

TUMOR ANGIOGENESIS

Presentations: Wednesday, April 13

BIOINFORMATICS / CONVERGENCE SCIENCE / SYSTEMS BIOLOGY

INTEGRATIVE SINGLE-CELL ANALYSIS

CHEMISTRY

PROTEOMICS, SIGNALING NETWORKS, AND BIOMARKER DISCOVERY

CLINICAL RESEARCH EXCLUDING TRIALS

PRECISION ONCOLOGY / SURGICAL ONCOLOGY

REAL-WORLD DATA (RWD) AND REAL-WORLD EVIDENCE (RWE) / OUTCOMES RESEARCH

TRANSLATIONAL RESEARCH: IMAGING

VACCINES / IMMUNOMODULATORY AGENTS AND INTERVENTIONS

CLINICAL TRIALS

PHASE I TRIALS IN PROGRESS 2

EXPERIMENTAL AND MOLECULAR THERAPEUTICS

LATE-BREAKING RESEARCH: EXPERIMENTAL AND MOLECULAR THERAPEUTICS 2

EMERGING NEW ANTICANCER AGENTS

HEMATOLOGICAL AND PEDIATRIC MALIGNANCY AND SARCOMA TREATMENT RESISTANCE

IDENTIFICATION OF MOLECULAR TARGETS

MOLECULAR CLASSIFICATION OF TUMORS

MOLECULAR PHARMACOLOGY

NEW CHEMOTHERAPY AGENTS

IMMUNOLOGY

LATE-BREAKING RESEARCH: IMMUNOLOGY 2

COMBINATION IMMUNOTHERAPIES 1

COMBINATION IMMUNOTHERAPIES 2

IMMUNOMODULATORY AGENTS AND INTERVENTIONS 3

THERAPEUTIC ANTIBODIES 3

MOLECULAR/CELLULAR BIOLOGY AND GENETICS

APOPTOSIS

CHROMATIN MODIFIERS: MUTATIONS AND NOVEL THERAPEUTICS

DNA MODIFICATION: MECHANISM TO BIOMARKERS TO TREATMENT

EPIGENOMICS TO MOLECULAR MARKERS

MITOCHONDRIAL METABOLISM IN CANCER

POPULATION SCIENCES

CANCER HEALTH DISPARITIES

PREVENTION / EARLY DETECTION / INTERCEPTION

OBESITY, BEHAVIORAL SCIENCE, AND QUALITY OF LIFE

TUMOR BIOLOGY

CLONAL EVOLUTION

GENE EXPRESSION AND THE MICROENVIRONMENT

INNERVATION, ECM, AND CELL SURFACE RECEPTORS

MODELS AND TECHNICAL APPROACHES TO ANALYZE AND EXAMINE THE TUMOR MICROENVIRONMENT

THERAPEUTIC TARGETS AND DRUG RESISTANCE IN PEDIATRIC CANCERS

Monday, April 11, 2022

BIOINFORMATICS / CONVERGENCE SCIENCE / SYSTEMS BIOLOGY

Single-cell and Spatial Profiling of Tumor Cells and the Immune Microenvironment

#2125

Towards a spatial view of immune cell function in cancer.

Maryam Pourmaleki,¹ Caitlin J. Jones,² Brian D. Greenstein,² Sabrina D. Mellinghoff,² Daniel A. Navarrete,² Smrutiben A. Mehta,² Nicholas D. Socci,² Ingo K. Mellinghoff,² Travis J. Hollmann². ¹Memorial Sloan Kettering Cancer Center/Weill Cornell Medicine, New York, NY; ²Memorial Sloan Kettering Cancer Center, New York, NY.

Immunotherapy can result in lasting tumor regressions, but despite its success, only a subset of patients and cancer types benefit from immunotherapy. Thus, uncovering features of the tumor microenvironment (TME) that contribute to this differential response can inform candidates for novel therapies and biomarkers for patient stratification. Recent advances in single-cell technologies allow for profiling of cell states and their spatial interactions within a tumor. Here, we performed multiplexed immunofluorescence (mpIF), a method for in situ single-cell measurement of 30+ proteins, as well as bulk transcriptomics and genomics on 180 tumor samples across 4 immunotherapy-responsive and -resistant solid and liquid cancers: in-transit melanoma (ITM), non-small cell lung cancer (NSCLC), glioblastoma multiforme (GBM), and classical Hodgkin lymphoma (cHL). In aggregate, we measured over 100 million single cells and identified over 1,000 unique tumor and immune cell states. We uncovered immune-suppressive cell states, such as macrophages negative for PD-L1 and B7-H3 in immunotherapy-resistant ITM tumors and all GBM tumors, which generally fail to respond to immunotherapy. In GBM, macrophages exhibited two distinct spatial topologies, infiltrated or excluded. In ITM, we identified pre-treatment cell states and gene expression signatures that associate with immunotherapy response such as MHC class I expression on the tumor cell surface, B cell aggregates, exhausted PD-1/LAG-3/TIM-3 triple-positive CD8+ T cells, and expression of interferon-gamma genes. We observed a similar immune checkpoint-rich (PD-1/LAG-3/TIM-3 triple-positive) TME in all cHL tumors, which have remarkably high immunotherapy response rates. We spatially defined the microniche (30-micron radius neighborhood) of exhausted CD8+ T cells, and within the microniches, found antigen presentation competent tumor cells, proliferating T cells, and B7-H3 positive macrophages, which together contribute to an activated TME. Together, we present a statistical workflow for the integrated analysis of spatially resolved multidimensional data for cancer target discovery that is tailored towards application in routinely collected formalin-fixed paraffin-embedded cancer biospecimens.

#2126

Single-cell sequencing of early-stage lung adenocarcinomas reveals prominent intratumoral heterogeneity and epithelial plasticity programs.

Guangchun Han,¹ Ansam Sinjab,¹ Warapen Treekitkarnmongkol,¹ Dapeng Hao,¹ Enyu Dai,¹ Luisa M. Solis,¹ Edwin R. Parra,¹ Stephen Swisher,¹ Tina Cascone,¹ Boris Sepesi,¹ Jichao Chen,¹ Steven Dubinett,² Junya Fujimoto,¹ Ignacio I. Wistuba,¹ Christopher S. Stevenson,³ Avrum E. Spira,³ Linghua Wang,¹ Humam Kadara¹. ¹UT MD Anderson Cancer Center, Houston, TX; ²University of California Los Angeles, Los Angeles, CA; ³Lung Cancer Initiative at Johnson and Johnson, Boston, MA.

Decoding the complex molecular and cellular processes during lung adenocarcinoma (LUAD) development is needed to devise early intervention strategies. To comprehensively capture LUAD neoplastic heterogeneity and cellular plasticity, we performed single-cell RNA-sequencing (scRNA-seq) of 257,481 enriched epithelial cells (EPCAM+ sorting) from 16 early-stage LUADs, each with 3 matched normal lung (NL) samples at defined spatial proximities to the tumor (n=47). 29,076 LUAD-derived cells clustered by patient and harbored distinct gene expression features (e.g., oxidative stress response), signifying interpatient LUAD heterogeneity. We also identified, using whole exome sequencing (WES) of matching lung and germline control samples, recurrent oncogenic driver alterations (e.g., EGFR, TP53, KRAS). Transcriptomic features of malignant cells were shared between LUADs (e.g., loss of lineage-specific gene expression) or private such as those associated with driver mutation status (e.g., KRAS). Indeed, clusters of malignant cells were overall segregated based on driver mutations (e.g., KRAS, EGFR). Malignant cells from KRAS-mutant LUADs (KM-LUADs) had increased activation of NF-kB, estrogen and hypoxia signaling, comprising a unique gene module (GM) that correlated with a less differentiated state. We also found hallmark pathways (cholesterol metabolism, DNA replication, cell fate decision) specific to EGFR-mutant LUADs (EM-LUADs). Notably, cells from one EM-LUAD and its 3 multiregion NL tissues clustered closely and had activated pro-tumor lymphoid signatures (CD4 naïve, Treg). Mutation burden increased with tumor proximity and intriguingly, EGFR exon20 mutation was evident in the tumor (VAF = 0.29) and its most proximal NL (VAF = 0.05), signifying a mutational field effect. Copy number variations (CNVs) derived from WES of all samples were overall consistent with those inferred from scRNA-seq data. Relative to EM-LUADs, malignant cells from KM-LUADs displayed lower CNV burdens. Interpatient CNV heterogeneity was prominent even among LUADs harboring the same oncogenic drivers. Notably, intratumor heterogeneity (ITH) was high among epithelial cells within single regions from the same LUAD. Among LUADs, malignant cell clades with KRAS mutations and lower CNV scores displayed less differentiated states. To investigate biological pathways driving ITH, we derived 6 GMs with tumor-relevant functional features, including a transcription/translation regulation GM that consistently correlated with reduced differentiation. Our analysis of a large number of lung epithelial cells from LUAD patients reveals in-depth insights into LUAD taxonomy which can help identify epithelial heterotypes, unravel the continuum of early differentiation events and expand our understanding of early LUAD pathogenesis.

#2127

Master regulator analysis of the tumor microenvironment and the distinctive tumor sub-populations in pancreatic ductal adenocarcinoma.

Jinjie Ling,¹ Lorenzo Tomassoni,² Alvaro Curiel,³ Kenneth Olive,³ Andrea Califano². ¹Vagelos College of Physicians and Surgeons, Columbia University, New York, NY; ²Department of Systems Biology, Columbia University, New York, NY; ³Department of Medicine, Columbia University, New York, NY.

By 2030, pancreatic ductal adenocarcinoma (PDAC) is projected to become the second leading cause of cancer-associated mortality in the United States. However, current strategies in targeting key driver mutations such as KRAS, CDKN2A, TP53, and SMAD4, have found limited translational success. As a result, the standard of care for PDAC patients continues to rely on cytotoxic chemotherapies. Master regulator (MR) proteins integrate upstream genomic alterations to generate a common downstream transcriptional signature characteristic of a tumor cell population. Importantly, aberrant MR protein activity is requisite to implementing and maintaining a pro-tumor state. Therefore, MR proteins serve as critical tumor dependencies and represent a promising class of therapeutic targets for cancer treatment. Here, we propose a novel, network-based approach to identify top MR proteins by utilizing a combination of ARACNe (Margolin et al, 2006) and VIPER (Alvarez et al, 2016) algorithms for the analysis of single-nuclei transcriptomic data. Specifically, we used VIPER to investigate the heterogeneity of the tumor microenvironment (TME) and to characterize the tumor sub-populations that co-exist in a patient PDAC specimen. Clustering analysis based on protein activity revealed six populations in the TME. The SingleR classifier subsequently annotated these clusters as myeloid and lymphoid immune, fibroblast, neuroendocrine, epithelial, and putative tumor cells. Next, we validated the putative tumor cells by performing InferCNV which detected widespread chromosomal copy number variations. We identified two distinct tumor subpopulations which were found to be enriched by the MR signatures of the Oncogenic Precursor (OP) and Lineage subtypes, previously demonstrated to represent more and less differentiated tumor cell populations, respectively (Laise et al, 2020). Interestingly, top MR proteins identified for the putative tumor cells using the single-nuclei transcriptomics closely corresponded with those identified by single-cell and previous bulk transcriptomics, providing orthogonal validation for the inferred protein activities. We characterized the diverse cell populations in the PDAC TME, delineated two distinct tumor cell subtypes, and demonstrated high concordance of inferred MR protein activity at the single-cell, single-nuclei, and bulk levels. These results demonstrate the feasibility of applying systems biology-based methods to identify MR proteins in an N = 1 context and represent a novel precision medicine approach in the treatment of PDAC patients. Future lines of inquiry center on identifying treatments against each tumor subpopulation via OncoTreat, an algorithm that leverages high-throughput drug perturbation data to predict effective therapies based on inferred ability to invert MR protein activity (Alvarez, 2018).

#2128

Single-cell replication dynamics in genomically unstable cancers.

Adam C. Weiner, Andrew McPherson, Sohrab P. Shah. Memorial Sloan Kettering Cancer Center, New York, NY.

Background: Single-cell whole genome sequencing (scWGS) methods such as direct library preparation (DLP) provide amplification-free capture of cells in all cycle phases and have enabled rich interrogation into the cell to cell genomic diversity of cancer genomes. Previous DLP-driven clonal evolution studies removed S-phase cells as replicated loci confounded phylogenetic inference from somatic copy number aberrations (CNAs), leaving single-cell replication timing (scRT) as an unstudied property. We thus developed a method that assigned S-phase cells to phylogenetic clones and used such assignments to infer scRT profiles in aneuploid cell lines and tumors. We applied this method to determine the relative proliferation rate between clones and obtain a single-cell picture of DNA replication fork progression for genomically unstable cancers.

Methods: S-phase cells were assigned to phylogenetic clones based on copy number profile similarity and subsequently scRT profiles were inferred by normalizing out the effects of clonal, subclonal, and cell-specific CNAs before binarizing each genomic bin as replicated or unreplicated. scRT heterogeneity was quantified via a summary statistic representing the aggregate difference between expected and observed replication times (T-width). The scRT profiles were grouped into clone- and sample-level pseudobulks (aggregations of single cells) to determine genomic regions with differential replication timing. We then assessed the ability of S-phase enrichment to predict clonal expansions on time-series patient-derived xenografts of HER2+ and triple negative breast cancers that were either treatment-naive or exposed to cisplatin. In addition, we studied genetically engineered cell lines with inactivated DNA damage response genes (TP53, BRCA1, BRCA2) and primary high-grade serous ovarian cancer samples to quantify scRT heterogeneity at different states of genomic instability.

Results: We analyzed 15 datasets containing over 25,000 cells with ~1/3 of cells in S-phase. Time-series datasets revealed that S-phase enriched clones were highly proliferative in the treatment-naive setting and were most sensitive to cisplatin. T-width values calculated from inferred scRT profiles increased as a function of genomic instability, with the degree of heterogeneity varied between genomic regions. Clone-level pseudobulk replication timing profiles revealed that whole-chromosome aneuploidies were more likely to have unperturbed replication timing compared to complex structural variants such as translocations, inversions, and micronuclei.

Conclusions: Our time-series results show that S-phase cells enable the prediction of clone-specific proliferation and chemosensitivity using scWGS data from a single time point. The quantification of scRT heterogeneity across our collection of data implicates dysfunctional DNA replication as both driver and consequence of genomic instability.

#2129

Spatial charting of single cell transcriptomes in tissues.

Runmin Wei,¹ Siyuan He,¹ Shanshan Bai,¹ Emi Sei,¹ Min Hu,¹ Alastair Thompson,² Ken Chen,¹ Savitri Krishnamurthy,¹ Nicholas Navin¹. ¹The University of Texas MD Anderson Cancer Center, Houston, TX; ²Baylor College of Medicine, Houston, TX.

Single cell RNA sequencing (scRNA-seq) methods can profile the transcriptomes of single cells but cannot preserve spatial information. Conversely, spatial transcriptomics (ST) assays can profile spatial regions in tissue sections but do not have single cell genomic resolution. To address this issue, computational approaches (e.g., cell2location, RCTD) have been designed to deconvolute ST spots into proportions of different cell types. However, the spatial deconvolution method has the following limitations, 1) it can only infer cell type proportions of each spot and cannot reach higher granular cell states mapping; 2) it only predict categorical labels and cannot infer continuous cell information (e.g., lineage trajectories, gene signatures, continuous phenotypes) at a spatial resolution.

Here, we developed a computational approach called CellTrek that combines these two datasets to achieve single cell spatial mapping. Using a machine learning-based metric learning approach, CellTrek learned a cell-spot graph and then transfer spatial coordinates to cells. The CellTrek toolkit also provides two downstream analysis modules, including SColoc for spatial colocalization analysis and SCoexp for spatial co-expression analysis.

We benchmarked CellTrek using a simulation and two in situ datasets. We then applied CellTrek to reconstruct cellular spatial structures in existing datasets from normal mouse brain and kidney tissues. We also performed scRNA-seq and ST experiments on two ductal carcinoma in situ (DCIS) tissues and applied CellTrek to identify subclones that were restricted to different ducts, and specific T cell states adjacent to the tumor areas.

Our data shows that CellTrek can accurately map single cells in diverse tissue types to resolve their spatial organization into cellular neighborhoods and tissue structures. This method provides a new paradigm that is distinct from ST deconvolution, enabling a more flexible and direct investigation of single cell data with spatial topography.

#2130

Uncovering the spatial landscape of tumor-immune interactions using latent spaces from spatial transcriptomics.

Atul Deshpande,¹ Melanie Loth,² Dimitrios Sidiropoulos,¹ QingFeng Zhu,¹ Genevieve Stein-O'Brien,¹ NIkhil Rao,³ Cedric Uytingco,³ Stephen Williams,³ Cesar Santa-Maria,¹ Daniele M. Gilkes,¹ Lei Zhang,¹ Elizabeth Jaffee,¹ Robert Anders,¹ Ludmila Danilova,¹ Luciane T. Kagohara,¹ Elana J. Fertig¹. ¹Johns Hopkins University, Baltimore, MD; ²University of Pennsylvania, Philadelphia, PA; ³10x Genomics, Pleasanton, CA.

Recent advances in spatial transcriptomics (ST) enable us to measure gene expression from cancer tissues while retaining their spatial context. We present a novel bioinformatics pipeline to infer molecular changes from tumor and immune cell interactions in the tumor microenvironment (TME) from ST data. Latent space methods enable inference of biological patterns from ST without the need for spot deconvolution into cell-based spatial features. While linear latent space methods yield interpretable biological patterns, interactions between tumor and immune cells can be nonlinear. To enable comprehensive inference of the pathways in the TME, we developed novel algorithms to characterize biological patterns from ST data using linear latent space methods and further nonlinear effects from their interactions. For any given set of genes, the patternSpotter tool visualizes the spatial variation in the relative contribution of individual patterns to the aggregate expression at each location in the tumor sample. Application of this tool to latent features identified using CoGAPS non-negative matrix factorization on a Visium ST (10x Genomics) data from a lymph node with pancreatic cancer metastasis confirms its known immune cell architecture. Furthermore, we develop a patternMarker algorithm to identify sets of coexpressed genes associated with the patterns, which help us to pinpoint the underlying biological processes and cell types. Further analyzing a breast cancer sample with invasive carcinoma and multiple precursor lesions demonstrates that this approach can uncover tumor and immune regions without prior reliance on pathology annotations from H&E imaging. In this case, an ensemble-based factorization of multiple dimensions enhances our resolution of intra-tumor heterogeneity and identifies distinct hormone receptor pathways in different precursor lesions with the patternMarker algorithm. Additional latent features are associated with immune cells, revealing further heterogeneity in immune infiltration between the invasive carcinoma and distinct precursor lesions. Still, the molecular interactions resulting from this infiltration induce a further non-linear alteration to transcription not captured through the inferred latent spaces. To resolve this, we develop a further interactionMarker statistic to identify regions of inter-pattern interaction and the associated genes. We apply this approach to detect additional intra-tumor heterogeneity in immune signaling from infiltration suggestive of differences in immune attack of invasive lesions. Altogether, our pipeline for latent space analysis of ST can identify the location and context-specific molecular interactions within the TME, broadly applicable to a better understanding of the key drivers of tumorigenesis and resistance to immune attack in cancer.

#2131

Curating cartography: Enabling the harmonisation, visualisation, and reuse of diverse multiplexed imaging data through the Human Tumor Atlas Network Data Coordinating Center.

Adam J. Taylor,¹ Milen Nikolov,¹ Ino de Brujin,² Jeremy Muhlich,³ Mialy De Felice,¹ Artem Sokolov,³ Denis Schapiro,⁴ Peter Sorger,³ Julie Bletz,¹ Nikolaus Schultz,² Vésteinn Thorsson,⁵ James Eddy,¹ Ethan Cerami⁶. ¹Sage Bionetworks, Seattle, WA; ²Memorial Sloan Kettering Cancer Center, New York, NY; ³Harvard Medical School, Boston, MA; ⁴University of Heidelberg, Heidelberg, Germany; ⁵Institute for Systems Biology, Seattle, WA; ⁶Dana Farber Cancer Institute, Boston, MA.

The Human Tumor Atlas Network (HTAN) is a National Cancer Institute (NCI) Cancer Moonshot Initiative to generate three-dimensional molecular and spatial atlases of diverse human tumors and characterize crucial transitions in cancer progression and treatment. A series of manuscripts describing an extensive array of genomics, transcriptomics, proteomics, and imaging datasets are emerging but careful curation is required to maximize the data’s utility. HTAN centers have generated data from over 30 assay types including more than a dozen imaging methods (e.g., fluorescence microscopy, metal-tagged imaging, digital pathology, spatial transcriptomics, and electron microscopy). The HTAN Data Coordinating Center (DCC) develops infrastructure and tools to ingest, curate, explore, and share these data in a findable, accessible, interoperable, and reusable (FAIR) manner. As of November 2021 the DCC had ingested over 150 TB of data, including nearly 150,000 imaging data files. HTAN metadata schemas were developed through a community-driven Request for Comments process, including schemas for raw, processed and QC-checked imaging data, segmentation masks and feature arrays. This work led to a set of Minimum Information for Highly Multiplexed Tissue Imaging (MITI) guidelines, which are complemented by detailed metadata on participants and biospecimens. The Schematic python package is used to provide a user-friendly interface for defining data-model schemas, generate (meta)data submission spreadsheets, and asset-store interfaces on various cloud platforms. Across assay types, the HTAN schema currently encompasses over 700 attributes from 35 components. The DCC manages the HTAN Data Portal, which provides community access to the atlases and enables filtering of released atlas data based on metadata fields, and pointers to data availability. To enable visualization and exploration of imaging data directly from the portal, centers can submit Minerva stories, which provide guided storytelling of multiplexed tissue images. We provide methods to automatically generate default stories with visually appropriate overlays of sequential four-channel groups. We developed Miniature to generate unsupervised and interpretable multiplexed image thumbnails to provide rapid contextual information when browsing image collections. Dimensionality reduction is used to reduce a multiplexed image pyramid to three dimensions. Pixels are recolored according to their coordinates in low-dimensional space, where Pixels with similar marker expression are assigned similar colours. Public data sharing requires careful consideration of data egress costs and de-identification of images and their metadata. The DCC is working closely with NCI Cancer Research Data Commons to ensure long term-legacy and reuse of HTAN data.

Database Resources and Web Applications

#1189

OncoKB, MSK’s precision oncology knowledge base.

Sarah P. Suehnholz, Moriah Nissan, Hongxin Zhang, Ritika Kundra, Calvin Lu, Benjamin Xu, Maria E. Arcila, Marc Ladanyi, Michael F. Berger, Ahmet Zehir, Aijaz Syed, Julia E. Rudolph, Ross L. Levine, Ahmet Dogan, Jianjiong Gao, David B. Solit, Nikolaus Schultz, Debyani Chakravarty. Memorial Sloan Kettering, New York, NY.

OncoKB, Memorial Sloan Kettering Cancer Center’s (MSK) precision oncology knowledge base (www.oncokb.org), is a comprehensive database that annotates the oncogenic effects and clinical actionability of somatic alterations in cancer. OncoKB supports variant interpretation by the cBioPortal for Cancer Genomics and is used to annotate >12,000 MSK patient sequencing reports annually. Since its introduction in 2016, OncoKB has expanded to include 5685 alterations in 682 genes, and in October 2021, it became the first somatic knowledge base to be partially recognized by the FDA. The scope of the OncoKB FDA recognition includes clinically actionable variants that map to an FDA level of evidence, the processes of variant curation, and policies regarding database oversight, personnel training and transparency of data sources and operations. This recognition credentials OncoKB as providing accurate, reliable and clinically meaningful information to the medical and scientific communities.

The OncoKB Therapeutic (Tx) Levels of Evidence categorize variants based on their tumor type-specific predictive value of sensitivity or resistance to matched standard care or investigational targeted therapies. To date, OncoKB includes 43 Level 1 genes (included in the FDA drug label), 23 Level 2 genes (included in professional guidelines), 25 Level 3A genes (predictive of drug response in well-powered clinical studies), 23 Level 4 genes (predictive of drug response based on compelling biological evidence), and 11 R1 or R2 resistance genes. Initially focused on solid tumors, OncoKB was expanded to include hematologic disease annotation in 2019 and introduced Diagnostic (Dx) and Prognostic (Px) levels of evidence. All three level of evidence systems (Tx, Dx and Px) are consistent with the guidelines for evidence-based categorization of somatic variants published as a joint consensus recommendation by AMP/ASCO/CAP.

OncoKB is governed by a Clinical Genomics Annotation Committee, composed of MSK physicians and scientists who ensure that the information captured is accurate and current, and an external advisory board composed of leaders in the clinical oncology and genomics communities who oversee OncoKB updates and progress. OncoKB curation rules and processes are transparent and documented in the OncoKB Curation Standard Operating Procedure, which is publicly available via the website. User feedback to OncoKB content is encouraged via the website and through cBioPortal. Queries or suggestions by OncoKB users are addressed by the OncoKB team within 72 hours.

OncoKB offers licenses for academic, commercial and hospital use, with which users can programmatically access the web API. Future work includes coverage of additional cancer-associated genes, annotation of germline alterations that are predictive of drug response and/or associated with increased heritable cancer risk and the development of a clinical trial matching system.

#1190

MMHCdb: A knowledgebase for the evolving landscape of mouse models of human cancer.

Dale A. Begley, Debbie M. Krupke, Steven Neuhauser, John Sundberg, Carol J. Bult. The Jackson Laboratory, Bar Harbor, ME.

The laboratory mouse is the premier mammalian model organism for interrogating the genetic and molecular basis of human cancer and for preclinical investigations into targets for the prevention and treatment of cancer. The distributed and heterogenous nature of information about these model systems makes it difficult for researchers to integrate and interpret the information to determine the state of the field and to identify the most relevant models for basic and preclinical research. The Mouse Models of Human Cancer database (http://tumor.informatics.jax.org) is an expertly curated knowledgebase about genetically defined mouse strains and Patient Derived Xenograft (PDX) models of human cancer. Data in MMHCdb are obtained from peer-reviewed scientific publications and direct data submissions from individual investigators and large-scale programs. MMHCdb is built on FAIR data management principles (Findable, Accessible, Interoperable, Reusable). The enforcement of metadata standards and official gene, allele and strain nomenclature ensure accurate and comprehensive search results for cancer models. MMHCdb has long represented data from spontaneous or endogenously induced tumors from genetically defined mice and for PDXs which have been the foundation of basic cancer research and preclinical studies for decades. MMHCdb has expanded to include cancer models such as Diversity Outbred and Collaborative Cross mice which are ideally suited for research into the relationship of genetic variation with cancer susceptibility and for modeling the genetics of variability in treatment responses. The MMHCdb contains over 109,266 curated tumor frequency records for over 8,275 mouse strains. Tumor types in the database have been indexed to over 21,000 literature citations. PDX models and data available in MMHCdb are also accessible from the Patient Derived Cancer Models resource at EMBL-EBI which currently provides information for over 4,000 PDXs (https://cancermodels.org).MMHCdb is supported by NCI R01 CA089713  

#1191

The Cancer Complexity Knowledge Portal: Enabling the exploration, discovery and reuse of resources for interdisciplinary cancer research.

James Eddy, milen Nikolov, Brynn Zalmanek, Verena Chung, Julie Bletz. Sage Bionetworks, Seattle, WA.

The National Cancer Institute Division of Cancer Biology supports multiple research programs composed of interdisciplinary scientific communities that integrate approaches, data, and tools to address fundamental challenges in basic and translational cancer research. As the coordinating center for the CSBC and PS-ON, Sage Bionetworks is dedicated to fostering an open and collaborative scientific culture in which researchers can rapidly collaborate across institutional boundaries.

Towards this end, we developed the Cancer Complexity Knowledge Portal (cancercomplexity.synapse.org) as a community research resource that synthesizes and exposes the activities and outputs of the CSBC, PS-ON, and affiliated consortia. The portal links related resources (e.g., a grant to its publications) and provides search and faceting to accelerate discovery and collaboration in the cancer research community. We aim to provide rich context about along with access to the activities and contributors that have produced the resources hosted within this and other repositories. The portal currently hosts >75 grants, >2100 publications, >450 datasets, >150 tools, and ~250,000 files. These have been annotated to facilitate search and discoverability in collaboration with and through the generous efforts of the CSBC/PS-ON community and the NCI.

The Cancer Complexity Knowledge Portal supports flexible, responsive exploration of and access to curated resources and distilled knowledge through modern web components. Those resources are hosted in community databases (e.g., GEO and SRA) or within the Synapse data-sharing platform (synapse.org/csbcpson). As a back end for the portal, Synapse continues to provide a community workspace to upload, update, manage, browse, and download data via a web UI and APIs. We are excited to provide this portal as a resource to drive additional insight, discovery, and collaboration in the field. In the coming year, we will continue to make updates to the portal, ensuring the latest publications, tools and data are available and identifying new features to add value to the portal for the community. We plan to provide a richer catalog of computational tools, leveraging in part the efforts, interests, and methodological output of CSBC/PS-ON working groups. We will continue to work with Division of Cancer Biology programs to ensure that the portal is enabling the broader community to integrate the approaches, data, and tools needed to address important questions in basic and translational cancer research.

#1192

The Clinical Genome Resource (ClinGen) somatic cancer clinical domain working group.

Jason Saliba,¹ Gordana Raca,² Angshumoy Roy,³ Ian King,⁴ Shamini Selvarajah,⁴ Xinjie Xu,⁵ Rashmi Kanagal-Shamanna,⁶ Laveniya Satgunaseelan,⁷ David Meredith,⁸ Mark Evans,⁶ Alanna Church,⁹ Panieh Terraf,¹⁰ Yassmine Akkari,¹¹ Heather E. Williams,¹² Wan-Hsin Lin,¹³ Chimene Kesserwan,¹⁴ Deborah I. Ritter,³ Kilannin Krysiak,¹ Arpad Danos,¹ Alex Wagner,¹⁵ Marilyn M. Li,¹⁶ Dmitriy Sonkin,¹⁷ Jonathan S. Berg,¹⁸ Sharon E. Plon,³ Heidi L. Rehm,¹⁹ Shashikant Kulkarni,³ Ramaswamy Govindan,¹ Obi L. Griffith,¹ Malachi Griffith,¹ on behalf of the ClinGen Somatic CDWG. ¹Washington University School of Medicine, St. Louis, MO; ²Children's Hospital Los Angeles, Los Angeles, CA; ³Baylor College of Medicine, Houston, TX; ⁴University Health Network and University of Toronto, Toronto, Ontario, Canada; ⁵Mayo Clinic, Rochester, MN; ⁶The University of Texas MD Anderson Cancer Center, Houston, TX; ⁷Royal Prince Alfred Hospital, Sydney, Australia; ⁸Dana-Farber Cancer Institute and Harvard Medical School, Boston, MA; ⁹Boston Children’s Hospital and Harvard Medical School, Boston, MA; ¹⁰Memorial Sloan Kettering Cancer Center, New York City, NY; ¹¹Legacy Health, Portland, OR; ¹²Tempus Labs, Inc, Chicago, IL; ¹³Mayo Clinic, Jacksonville, FL; ¹⁴National Cancer Institute, Bethesda, MD; ¹⁵Nationwide Children's Hospital, Columbus, OH; ¹⁶Children’s Hospital of Philadelphia, Philadelphia, PA; ¹⁷National Cancer Institute, Rockville, MD; ¹⁸University of North Carolina School of Medicine, Chapel Hill, NC; ¹⁹Massachusetts General Hospital and Broad Institute of MIT and Harvard, Cambridge, MA.

Interpretation of the clinical significance of somatic gene variants in cancer remains a major challenge in cancer diagnosis, prognosis and treatment response prediction. We will report on progress and plans of the Clinical Genome Resource (ClinGen) Somatic Cancer Clinical Domain Working Group (CDWG). The CDWG membership consists of over 150 multi-disciplinary experts in cancer biology, oncology, pathology, genetics, genomics and informatics. The mission of the ClinGen Somatic Cancer CDWG is to facilitate the development of data curation guidelines and standards to determine the clinical significance of somatic alterations in cancer, thereby enhancing the usability, dissemination and implementation of cancer somatic changes in the ClinGen resource and other knowledgebases including CIViC, ClinVar, and the Variant Interpretation for Cancer Consortium (VICC) MetaKB. Our goal is to create high-quality assertions of the clinical significance of specific somatic variants in cancer by leveraging the CIViC curation interface, adapting the germline procedures of ClinGen to somatic variant interpretation, and implementing the interoperability standards of the Global Alliance for Genomics and Health (GA4GH). The ClinGen Somatic effort is overseen by the Somatic CDWG and reports progress to the overall ClinGen consortium. There are Somatic Cancer subdomains focused on particular clinically important domains of cancer variant interpretation including three Task Forces (covering Pediatric Cancer, Hematologic Cancer, and Solid Tumors) and a growing number of Somatic Cancer Variant Curation Expert Panels (SC-VCEPs). To improve quality and consistency of clinical interpretations, each candidate somatic cancer VCEP must complete a four step approval process adapted from ClinGen’s work in Germline disease domains. The Somatic CDWG works to ensure that each group is aware of available training materials and detailed standard operating procedures. Each SC-VCEP also coordinates with the ClinGen Cancer Variant Interpretation Committee (CVI) whose goal is to support development of granular specifications for the AMP/ASCO/CAP guidelines for somatic variant interpretation. New SC-VCEPs are anticipated to focus on specific clinically relevant genes, pathways, disease entities, variant classes or treatment modalities. Currently, three SC-VCEPs have begun to work through the four step process (focused on FGFR mutations, NTRK fusions, and FLT3 mutations respectively), and two more SC-VCEPs are in the planning stage (Histone H3 and Ph-like ALL). To date, ClinGen Somatic groups have contributed 619 evidence lines into CIViC from 353 published papers and 21 assertions of clinical significance. Input from the AACR community is critical for the establishment of new SC-VCEPs that address areas of variant interpretation with the greatest need.

#1193

Enhancing pediatric cancer variant curation and representation through standardized classification and automation.

Jason Saliba,¹ Jake Lever,² Kilannin Krysiak,¹ Arpad Danos,¹ Alex Wagner,³ Heather E. Williams,⁴ Laveniya Satgunaseelan,⁵ David Meredith,⁶ Cameron J. Grisdale,⁷ Chimene Kesserwan,⁸ Jianling Ji,⁹ Shruti Rao,¹⁰ Catherine Cottrell,³ Alanna Church,¹¹ Mark Evans,¹² Yasmina Jaufeerally-Fakim,¹³ Lynn M. Schriml,¹⁴ Angshumoy Roy,¹⁵ Gordana Raca,⁹ Malachi Griffith,¹ Obi L. Griffith¹. ¹Washington University School of Medicine, St. Louis, MO; ²University of Glasglow, Glasglow, United Kingdom; ³Nationwide Children's Hospital, Columbus, OH; ⁴Tempus Labs Inc. and Yale University School of Management, Chicago, IL; ⁵Royal Prince Alfred Hospital, Sydney, Australia; ⁶Dana-Farber Cancer Institute and Harvard Medical School, Boston, MA; ⁷Canada’s Michael Smith Genome Sciences Centre, Vancouver, British Columbia, Canada; ⁸National Cancer Institute, Bethesda, MD; ⁹Children's Hospital Los Angeles, Los Angeles, CA; ¹⁰Georgetown University Medical Center, Washington DC, DC; ¹¹Boston Children’s Hospital and Harvard Medical School, Boston, MA; ¹²The University of Texas MD Anderson Cancer Center, Houston, TX; ¹³University of Mauritius, Moka, Mauritius; ¹⁴University of Maryland School of Medicine, Baltimore, MD; ¹⁵Baylor College of Medicine, Houston, TX.

Childhood cancers present unique challenges for variant interpretation in a clinical context due to their rarity, low mutation burden, diversity of molecular alterations, and heterogeneity among patients. Consequently, genes and variants associated with childhood tumors are under-represented in public cancer databases and knowledgebases. A focused effort is needed for the structured curation of genetic variant-level data to document diagnostic, prognostic, and therapeutic biomarkers for childhood cancers. The Pediatric Cancer Curation Advancement Subcommittee (PCCAS), a collaboration between the Clinical Interpretations of Variants in Cancer knowledgebase (CIViC; civicdb.org), the ClinGen Somatic Pediatric Cancer Taskforce, Disease Ontology (DO; disease-ontology.org) and CIViCmine (bionlp.bcgsc.ca/civicmine/), is addressing this challenge through enhanced curation, tagging, and automation.

PCCAS created a pediatric specific curation standard operating procedure (SOP) to harmonize pediatric evidence entered in CIViC. Our SOP provides general guidance and considerations to define and classify childhood cancers and to represent childhood cancer evidence on a spectrum of age-related incidence and presentation. For instance, pediatric evidence in CIViC is now tagged using Human Phenotype Ontology (HPO) age of onset terms, allowing pediatric evidence to be easily searched, tracked, and sorted. We also initiated the addition of new age of onset terms to enhance the granularity of these tags.

WHO ICD-O nomenclature has been chosen for pediatric disease classification in CIViC. ICD-O provides updated terminology including specific genetic subtypes, which are important in pediatric cancers where their underlying molecular profiles often define the disease. To aid curator selection of disease, we verified pediatric relevant ICD-O terms inclusion in DO and restructured DO disease hierarchies to ensure proper mapping.

CIViC highlights our pediatric cancer initiative in multiple areas including a homepage feature linking directly to a dedicated pediatric advanced search that returns all evidence tagged with pediatric or young adult age of onset. Most importantly, our childhood specific SOP and initiatives are included in all ClinGen Somatic Cancer and CIViC training sessions for consistent implementation.

CIViCmine supports CIViC by using natural language processing to identify important cancer biomarkers in the literature. To better identify pediatric biomarkers, we are adapting and refining CIViCmine to use MeSH terms and other approaches to enhance accuracy in the identification of childhood evidence in both the literature and CIViC. In conclusion, implementation of these procedures, features, and automation are pushing to make childhood cancer variant evidence more accessible and interpretable.

#1194

Redesigning CIViC: Enhancing the structured curation of complex cancer variant data.

Kilannin C. Krysiak,¹ Adam C. Coffman,¹ Susanna Kiwala,¹ Joshua F. McMichael,¹ Arpad M. Danos,¹ Jason Saliba,¹ Cameron J. Grisdale,² Jake Lever,³ Lana Sheta,¹ Shruti Rao,¹ Alex H. Wagner,⁴ Malachi Griffith,¹ Obi L. Griffith¹. ¹Washington University School of Medicine, St Louis, MO; ²Canada’s Michael Smith Genome Sciences Centre, Vancouver, British Columbia, Canada; ³University of Glasgow, Glasgow, United Kingdom; ⁴Nationwide Children's Hospital and The Ohio State University College of Medicine, Columbus, OH.

The Clinical Interpretation of Variants in Cancer (CIViC; www.civicdb.org) knowledgebase is a curation platform designed to capture evidence from the published literature which support or refute the significance of genomic variants in various cancer types. Since the launch of the beta user interface in 2014, this knowledgebase has undergone substantial evolution and redesign to support the needs of the clinical and research communities. Significant work has been done by the community to release guidelines and resources to support cancer variant interpretation. CIViC is a crowd sourced, expert moderated resource with a community of over 300 active curators. In response to community feedback, evolving guidelines and improved understanding of the genetics of cancer, we have developed CIViC 2.0, a major redesign that expands the capabilities of this widely used resource.

CIViC 2.0 has been designed to support complex variant relationships (termed molecular profiles in CIViC) such as the poor prognostic impact of the combination of variants in FLT3, DNMT3A and NPM1 in acute myeloid leukemia. These highly structured molecular profiles link details about each constituent variant, including genomic coordinates, HGVS, aliases, and additional metadata. Molecular profiles will also support the absence of a variant, which has become critical in targeted therapy decision making such as colorectal cancer where EGFR expression without a KRAS variant is an indication for targeted inhibitor therapy. These molecular profiles also support structural variants. While fusions have always been supported, CIViC 2.0 will support complex genomic coordinate or cytoband specific queries. 

Significant technical changes have improved the speed and performance of the UI and the underlying API. A complete redesign of the API, utilizing GraphQL, empowers users to ask more complex questions and more deeply explore this highly curated data. More robust data schemas and comprehensive validation of evidence structure will allow for the programmatic submission of new evidence. By utilizing site-wide full text faceted search, identifying relevant content in summaries, evidence items, or even curator comments is straightforward.

The CIViC 2.0 redesign supports the ever-increasing complexity of cancer variant information and provides a powerful tool to explore and utilize the carefully curated data within the knowledgebase for a multitude of questions related to basic, translational, and clinical research.

#1195

BC-BET v2.0: Updates to an online Bladder Cancer Biomarker Evaluation tool.

Garrett M. Dancik. Eastern Connecticut State University, Willimantic, CT.

The Bladder Cancer Biomarker Evaluation Tool (BC-BET) is a web application for quickly evaluating candidate diagnostic and prognostic biomarkers in bladder cancer, based on publicly available gene expression datasets. Specifically, BC-BET evaluates whether a selected gene is differentially expressed between normal cells and tumor samples; between low- and high- grade tumors; and between non-muscle invasive and muscle-invasive tumors. BC-BET also evaluates whether a selected gene is significantly associated with survival, and whether the gene is associated with survival in two clinically important patient subsets: patients with low-grade, non-muscle invasive tumors; and patients with high-grade, muscle invasive tumors. Here we describe updates to BC-BET that include the inclusion of additional patient cohorts, and new features including multi-gene search, data visualization, additional data export options, and faster search times. BC-BET v2.0 contains gene expression data from 17 cohorts (N = 1846), and users can now query up to 500 genes at a time in order to quickly identify genes associated with bladder cancer. For single gene queries, dot plots can be generated for visualization of gene expression across groups; while Kaplan-Meier curves can be generated for survival analyses. BC-BET v2.0 also allows users to export the gene expression and clinical data, which would allow researchers to generate their own visualizations or to carry out additional analyses. Finally, the BC-BET backend was re-designed, resulting in more than a 2x fold increase in performance efficiency. We hope that this tool will continue to serve the bladder cancer research community and will ultimately lead to a better understanding of and treatment of this disease. The BC-BET web application as well as its source code is freely available from https://gdancik.github.io/BC-BET/.

#1196

Elucidating compound mechanism of action and polypharmacology with a large-scale perturbational profile compendium.

Lucas Zhongming Hu,¹ Eugene Douglass,¹ Ron Realubit,¹ Charles Karan,¹ Mariano Alvarez,² Andrea Califano¹. ¹Columbia University Irving Medical Center, New York, NY; ²DarwinHealth, Inc., New York, NY.

Traditional approaches to elucidate small molecule mechanism of action (MoA) are usually based on affinity binding assays. These often fail to detect lower affinity binding targets and secondary effectors that are usually highly tissue context-dependent. As such, these can vary dramatically across different cancer types. To address this challenge and facilitate more quantitative approaches to targeted cancer therapy, a more comprehensive picture of drug MoA, including poly-pharmacology and toxicity effects mediated by the full range of its high-affinity (primary), as well as its context-specific lower-affinity (secondary) and downstream effector (tertiary) targets. For this purpose, we generated transcriptional genome-wide RNA-Seq profiles of cancer cell lines, following compounds perturbation with FDA approved and late-stage experimental oncology drugs using a fully automated and highly efficient PLATE-Seq technology¹. Compared to prior approaches, this provides (a) genome-wide readouts, rather than readouts limited to a small set of landmark genes, (b) in cell lines that were specifically selected as the highest-fidelity models for published human cancer cohorts, using the OncoMatch algorithm², and (c) for a drug repertoire that is clinically relevant. In addition, to avoid confounding effects from drug stress/death response pathway activation—a common issue in previous studies—drugs were titrated at their maximum sub-lethal concentration (48-hr IC20) using 10-point dose response curves in each cell line. The ~20,000 molecular profiles generated by these assays were analyzed using the VIPER algorithm³ to reproducibly assess the effect of each drug on the activity of ~6,500 regulatory and signaling proteins compared to vehicle control. The resulting PanACEA (Pancancer Activity-based Compound Efficacy Analysis), a database comprising drug perturbation profiles for 23 different cancer cell lines and > 700 oncology drugs, representing the largest resources of functionally annotated, genome-wide perturbational profiles for clinically relevant drugs. Systematic analysis identified critical drug mechanism of action and drugs capable of reproducibly targeting undruggable proteins, such as MYC or KRAS. We also leveraged graph theory-based network approach to generate a conserved drug functional network and drug functional networks in different cancer contexts, which pinpoint critical novel poly-pharmacology effects.

#1197

Refining the drug-gene interaction database for precision medicine pipelines.

Matthew Cannon,¹ James Stevenson,¹ Kori Kuzma,¹ Colin O'Sullivan,¹ Katherine Miller,¹ Olivia Grischow,¹ Adam Coffman,² Susanna Kiwala,² Joshua F. McMichael,² Dorian Morrissey,² Kelsy Cotto,² Obi Griffith,² Malachi Griffith,² Alex Wagner¹. ¹Nationwide Children's Hospital, Columbus, OH; ²Washington University, St. Louis, MO.

The Drug-Gene Interaction Database (DGIdb, www.dgidb.org) is a publicly accessible resource that aggregates 102,426 gene records and 57,498 drug records from 40 drug-gene interaction data sources to aid both researchers and clinicians in identifying associations between genes of interest and available drugs and therapeutics. By using peer-reviewed data sources and publications, DGIdb represents a stand-alone resource with over 100,000 drug-gene interaction claims across 30 interaction types to drive hypothesis generation in precision medicine and interpretation pipelines. The background process that normalizes drugs to a harmonized ontological concept has been upgraded. These improvements have increased concept normalization for drugs by 20% and are now available as a stand-alone service for use (https://normalize.cancervariants.org/therapy/). Leveraging our platform’s ability to find relationships between disease-critical genes and available therapeutics, DGIdb has been used in clinical interpretation pipelines to find drugs for specific diseases with an emphasis on regulatory approval status. DGIdb now uses annotations from Drugs@FDA as an additional source to provide more accurate descriptors for market and maturity status of drugs, when available. Lastly, to enhance the annotation potential for DGIdb in precision medicine pipelines, we have updated our druggable gene category sources with an additional curated list of 2,217 genes. Used alone or in combination with existing categories-such as the heavily-utilized ‘clinically actionable’ category-this additional source will give precision medicine and interpretation pipelines the power to find concise, actionable annotations for specific diseases including pediatric cancers and epilepsy. These lists are managed and maintained as a publicly-available resource to provide up-to-date annotations on disease-associated genes as they become available.

#1198

Discovering novel bispecific antibody targets through the mining of large-scale bulk and single cell RNA-seq databases.

Eugene Chekalin, Shreya Paithankar, Bin Chen. Michigan State University, Grand Rapids, MI.

Bispecific antibodies (bsAbs) that bind two distinct surface proteins in cancer cells are emerging as an appealing therapeutic strategy in cancer immunotherapy. Among thousands of surface proteins, experimentally identifying the best target pairs that are expressed only in cancer cells, but not in normal cells, is costly and time-consuming. The open bulk RNASeq and single-cell (sc)RNASeq offers a great resource to identify novel bispecific targets. Bulk RNASeq has been widely explored to identify therapeutic targets and biomarkers, resulting in voluminous data for various cancers and normal tissues, but the mixture of cell types in bulk RNASeq could not characterize the precise expression of targets in cancer cells, resulting in considerable false positives. scRNASeq provides a high resolution of target expression in individual cells, however, the challenges in solving the dropout issue and cell type classification hinder the direct use of scRNASeq in bsAbs target identification. Utilizing the OCTAD database consisting of over 20,000 bulk RNASeq samples, we proposed an approach that identifies target pairs that separate tumors from healthy the most, taking into account cluster heterogeneity, the distance between tumors and healthy as well as the angle between potential markers. Among the top pairs in Hepatocellular Carcinoma (HCC), CD33~PLVAP, for example, was a false positive because CD33 is myeloid and lymphoid cell lineage-specific. We thus assembled a scRNAseq database of healthy vital organs containing 39361 cells to aid selection. By comparing their expression with the expression of 18000 malignant cells predicted from 72000 cells of eight HCC samples, we identified target pairs that mostly express on the surface of the malignant HCC cells and had low or zero expression in vital organs. The most promising marker pair was GPC3~MUC13, presenting on the surface of over 30% of malignant HCC cells, with very low expression in vital organs. We further developed an R package to navigate the bsAbs target selection from open bulk RNASeq and scRNASeq.

#1200

COSMIC cancer mutation census: Classifying somatic coding variants by their potential to drive cancer.

Zbyslaw Sondka, Bhavana Harsha, Helder Pedro, Nidhi Bindal Dhir, Charlie Hathaway, Sumodh Nair, Doron Sondheimer, Simon A. Forbes. Wellcome Sanger Institute, Cambridge, United Kingdom.

Somatic mutations accumulate in cells throughout their life. Most of them do not bring any negative effect. However, certain mutations change protein behaviour, structure, or level of expression. More importantly, some mutations are known to initiate and drive oncogenic transformation. These mutations often make good therapeutic targets but recognising this small subset in a cancer sample is a major challenge. The average cancer cell carries a life-long baggage of somatic mutations, and the mutational process is sped up in these cells through genomic instability (one of the hallmarks of cancer). As a result, there are hundreds of thousands of variants of unknown significance identified through sequencing of cancer DNA. COSMIC Cancer Mutation Census (CMC) answers this challenge by identifying coding mutations with a potential to drive cancer. This is achieved by combining manually curated information regarding cancer genes and genetic variants with data on variant frequencies in cancer and non-cancer populations, and algorithmic evaluation of variant significance. It applies a simple and transparent set of rules to the whole set of coding mutations in COSMIC to identify variants with the highest potential of clinical relevance. In current version (v95, November 2021) the CMC describes 4.7 million somatic variants and segregates them into four tiers. Tier 1 is the highest confidence set. This set includes 1558 mutations that are found in Cancer Gene Census genes and are also described as pathogenic in cancer by ClinVar. Tiers 2 and 3 contain variants with less extensive evidence of involvement in carcinogenesis. The dN/dS algorithm is used to include variants that are under positive selection in cancer cells. Finally, mutations without evidence for driving cancer are classified as Tier 4. In addition to this classification, CMC integrates and presents the information used to prioritise variants, including their frequencies in various cancer types (COSMIC), germline frequencies (gnomAD), ClinVar annotations, dN/dS analysis results, and nucleotide and amino acid conservation. Data can be accessed and scrutinised through a dedicated website at https://cancer.sanger.ac.uk/cmc.

#1201

GenomeCruzer: A tool for interactive 3D visualization and analysis of multidimensional cancer omics.

Claudio Isella,¹ Riccardo Corsi,² Jamal Elhasnaoui,¹ Andrea Bertotti,¹ Luca Vezzadini,² Enzo Medico¹. ¹University of Torino, Candiolo, Italy; ²Kairos3D s.r.l., Torino, Italy.

In cancer genomics, integrative analysis of data obtained from biological samples of patient cohorts requires handling large groups of patients, each with molecular data of different type for thousands of genes, such as expression, mutation, copy number, and others.

Currently available tools allow visualization and analysis of such complex data, however forcing the user to work with partial views due to the amount of information. Also, these systems usually involve somewhat laborious procedures, so that for a complete analysis of the information it is often necessary to use many tools in sequence.

Here we present GenomeCruzer, a software for 3D interactive visualization and analysis of multidimensional omic data sets (e.g. gene expression, methylation, copy number alteration). GenomeCruzer is particularly useful to quickly achieve clinical and biological insight, providing easy integration and interactive visualization of genomic data of any kind. GenomeCruzer has the dual objectives of representing large amounts of genomic data in an easily readable way, allowing interactive analysis and providing visual results in real-time. Powerful graphics also allow simultaneous display of different types of values for the same objects, through multiple dimensions and customizable visual metaphors. The user can navigate this 3D environment and interact with the different elements, which constitute a representation of the starting datasets, building analysis paths and using statistical tools e.g. for the comparison of groups to highlight correlations.

In conclusion, GenomeCruzer enables users without bioinformatics background to browse, analyse and interpret complex omics datasets for multiple groups of patients and genes simultaneously, extracting clinical correlates and biological knowledge.

#1202

A repository of PDX histology images for exploring spatial heterogeneity and cancer dynamics.

Brian S. White*,¹ Xingyi Woo*,¹ Soner Koc*,² Todd Sheridan,¹ Steven B. Neuhauser,³ Akshat M. Savaliya,¹ Lacey E. Dobrolecki,⁴ John D. Landua,⁴ Matthew H. Bailey,⁵ Maihi Fujita,⁶ Kurt W. Evans,⁷ Bingliang Fang,⁷ Junya Fujimoto,⁷ Maria Gabriela Raso,⁷ Shidan Wang,⁸ Guanghua Xiao,⁸ Yang Xie,⁸ Sherri R. Davies,⁹ Ryan C. Fields,⁹ R Jay Mashl,⁹ Jacqueline L. Mudd,⁹ Yeqing Chen,¹⁰ Min Xiao,¹⁰ Xiaowei Xu,¹⁰ Melinda G. Hollingshead,¹¹ Shahanawaz Jiwani,¹² PDXNet Consortium, Yvonne A. Evrard,¹² Tiffany A. Wallace,¹¹ Jeffrey A. Moscow,¹¹ James H. Doroshow,¹¹ Nicholas Mitsiades,⁴ Salma Kaochar,⁴ Chong-xian Pan,¹³ Moon S. Chen,¹³ Luis G. Carvajal-Carmona,¹³ Alana L. Welm,⁶ Bryan E. Welm,⁶ Michael T. Lewis,⁴ Ramaswamy Govindan,⁹ Li Ding,⁹ Shunqiang Li,⁹ Meenhard Herlyn,¹⁰ Michael A. Davies,⁷ Jack A. Roth,⁷ Funda Meric-Bernstam,⁷ Carol J. Bult,³ Brandi Davis-Dusenbery,² Dennis A. Dean,² Jeffrey H. Chuang¹. ¹The Jackson Laboratory for Genomic Medicine, Farmington, CT; ²Seven Bridges Genomics, Inc, Charlestown, MA; ³The Jackson Laboratory, Bar Harbor, ME; ⁴Baylor College of Medicine, Houston, TX; ⁵Simmons Center for Cancer Research, Brigham Young University, Provo, UT; ⁶Huntsman Cancer Institute, University of Utah, Salt Lake City, UT; ⁷The University of Texas MD Anderson Cancer Center, Houston, TX; ⁸University of Texas Southwestern Medical Center, Dallas, TX; ⁹Washington University School of Medicine, St. Louis, MO; ¹⁰The Wistar Institute, Philadelphia, PA; ¹¹National Cancer Institute, Bethesda, MD; ¹²Leidos Biomedical Research Inc., Frederick National Laboratory for Cancer Research, Frederick, MD; ¹³University of California - Davis, Davis, CA.

Patient-derived xenografts (PDXs) recapitulate intratumoral spatial heterogeneity and simulate a tumor microenvironment in which human immune and stromal cells in the PDX are replaced over passages by murine cells partially lacking immune function. Histological imaging enables exploring the spatial heterogeneity and dynamics of cancer, stromal, and immune cell interactions as correlates of tumor stage and therapeutic response over passages. We created a repository of curated, haematoxylin and eosin (H&E) images as a community resource for addressing these questions.

Images were generated at five sites within the NCI’s PDX Development and Trial Centers Research Network (PDXNet) and the NCI Patient-Derived Models Repository. Over 900 images, including 739 from PDXs and 190 from paired patients, are hosted on the Seven Bridges Genomics Cancer Genomics Cloud. They represent 42 cancer subtypes, including breast cancer (n=134), colon adenocarcinoma (COAD; n=94), pancreatic cancer (n=87), lung adenocarcinoma (LUAD; n=80), melanoma (n=71), and squamous cell lung cancer (LUSC; n=65). Paired human / PDX images are available for each of these cancers. Human and / or PDX images generated following patient treatment are available for 37 of the subtypes. Most images are from early passages (P0: 158; P1: 292; P2: 152; P3: 69; >P3: 55). Annotations include sex, age, race, ethnicity, and, for most images, pathological assessment of tissue-level percent cancer, stromal, and necrotic cell content (n=639) and tumor stage (n=650). RNA and exome sequencing data are available for 99 and 228 images, respectively, matched at the patient or sample level.

Quality control was performed using HistoQC. Cells were segmented and labeled as neoplastic, necrotic, immune, stromal, or other using Hover-Net and predictions of total neoplastic cell area correlated with whole-slide pathological assessment of cancer cell percentage (COAD: r=0.51; LUSC: r=0.59). HD-Staining, another classification approach, was applied to a subset of images and our clinical annotations will facilitate validation of this and related methods. Features of 512 x 512 pixel tiles were computed using the Inception V3 convolutional neural network pre-trained on ImageNet. Unsupervised clustering of these features demonstrate inter-patient heterogeneity within pathologist-annotated tumor regions. A classifier developed using pathologist-annotated cancer, stromal, and necrotic regions and trained on the features in LUSC images (n=10 images) achieved a cross-validation accuracy of 96% for cancer tiles across (n=5) LUAD images. Accuracy was lower for stromal classification (90%), likely reflecting current limitations of our small, but growing, labeled training set.

Our repository of clinically-annotated PDX H&E images should aid the community in studying spatial heterogeneity and in training deep learning-based image analysis methods.

New Algorithms and Tools for Data Analysis

#1205

Modeling spatially resolved cell-type-specific gene expression by weighted regression with SPACER.

Amelia R. Schroeder, Kyle Coleman, Jian Hu, Mingyao Li. University of Pennsylvania, Philadelphia, PA.

Transcriptomic analysis has substantially advanced our understanding of human diseases, but the complex nature of tissues is often ignored. Recent development of single-cell RNA-seq (scRNA-seq) technologies has made it possible to characterize cellular heterogeneity of solid tissues. Along with analyzing gene expression patterns on a single-cell level, there is a critical need to explore the spatial patterns exhibited by the genes across the tissue sample. Spatially resolved transcriptomics (SRT) is key to understanding cellular functions in their morphological state. However, current barcoding-based SRT technologies, such as 10x Genomics Visium, lack single-cell resolution, greatly hindering the investigation of the spatial differences in gene expression profiles as failure to account for cell-type variations can lead to an obscured understanding of the spatial patterns detected across the tissue. The overall goal of this research is to integrate spatial transcriptomics and scRNA-seq data in order to infer the gene expression profiles for each of the bulk-level spots in the spatially resolved data. This will allow us to study spatial patterns of genes with cell-type level resolution. SPACER expands the spatial transcriptomics data into the individual cell types by utilizing the spatial similarities and high-resolution histology information as a weight in a non-negative least squares regression. Having this integrated understanding allows us to gain an additional level of information about the gene activity for each cell type present in the tissue. We have performed benchmark evaluations for our method based on data generated from the 10x Visium platform and have seen promising results. The evaluations have shown high correlations for the gene expression patterns predicted by SPACER for varying cell types on the benchmark evaluations. We next analyzed spatial transcriptomics data for pancreatic cancer, breast cancer, and melanoma tissue samples to better study the complex nature of cancerous tissues. A tumor is consistently interacting with its microenvironment, which can impact tumor growth and cell proliferation. Thus, it is important to study the spatial gene expression patterns for different cell types in cancerous tissues.