Session 1: Cell & Network Biology
Session 2: Cell Biology & Signaling
Effectively Advocating for Science to the Public
Collaborative Discussion Sessions
Session 3: Initiation & Progression
Pediatric Cancer Systems Biology
The Cancer Cell Map Initiative: 2025 Update
Trey Ideker, UC San Diego
In contrast to the substantial progress in decoding the genetic basis of cancer, systematic efforts to map cancer pathways are just beginning. The Cancer Cell Map Initiative (CCMI) has recently completed a multi-cancer protein–protein interaction map centered on the protein p53 measured in breast cancer, colon cancer, leukemia. We find that TP53 structural mutants interact robustly with chaperones and protein-folding proteins (e.g., DNAJA1), whereas TP53 DNA-binding mutants exhibit elevated metabolic signatures. These patterns illuminate the distinct interactions that differentiate different mutations of the same gene. To complement these efforts, we have completed a comprehensive map of synthetic lethal interactions between cancer genes and druggable targets across seven different tumor backgrounds, representing breast, lung and oropharyngeal cancers of heterogeneous oncogenic backgrounds. From combinatorial CRISPR knock outs, 1,805 synthetic lethal interactions were identified. Of these interactions, 226 translate to robust biomarkers of drug sensitivity, including frequent genetic alterations in the KDM5C/6A histone demethylases, which sensitize to inhibition of TIPARP (PARP7). Finally, we are using the above cancer network maps to guide interpretable precision oncology models called visible neural networks. Towards establishing these predictive models as physician-ready tools, we recently published a series of seven “Hallmarks of Predictive Oncology” (including Interpretability, Generalizability, and Fairness). This work includes a hallmarks-based scorecard and model checklist to guide both model creators and clinical evaluators, aiming to bridge the translational gap between machine learning innovations and clinical utility.
Other Authors: Trey Ideker, PhD, Professor, UC San Diego Marcus Kelley, PhD, Postdoc, UC San Diego Nevan Krogan, PhD, Professor, UC San Francisco Nadia Arang, PhD, Postdoc, UC San Francisco
scMINER2 - A causal feature learning framework to untangle cell-type-resolved hidden gene regulatory network (hGRN)
Yogesh Dhungana, St. Jude Children's Research Hospital
Single-cell technologies provide unprecedented resolution into immune cell biology, but they also intensify the curse of dimensionality, where sparsity and high dimensionality undermine traditional similarity and clustering approaches. Addressing this limitation requires not only dimensionality reduction but also mechanistic frameworks that preserve causal biological signals. To overcome these challenges, we developed scMINER2, a systems biology framework that simultaneously reconstructs hidden gene regulatory networks (hGRNs) and lineage trajectories from scRNA-seq and scATAC-seq data. By integrating transcriptional and chromatin accessibility profiles, scMINER2 identifies differentiation pathways and regulatory bottlenecks. Benchmarking against existing tools demonstrated that scMINER2 not only outperforms in accuracy but also enables deeper mechanistic interpretation. Applied to CAR-T cells, scMINER2 revealed a previously unrecognized subset of type-2 memory (T2M) cells. In contrast to historically suppressive type-2 programs, T2M cells emerged as long-lived, self-renewing, and indispensable drivers of CAR-T persistence, reframing type-2 immunity as a hidden dimension of therapeutic success in cancer immunotherapy. To validate this finding, we developed scPhenoMAP, a single-cell 2D embedding algorithm that projects T cells into a unified effector-state map across datasets. Using CAR-T scRNA-seq from global multi-center clinical trials spanning multiple continents, tumor types, and CAR constructs, scPhenoMAP consistently identified T2M enrichment in responders. This global reproducibility, further confirmed by preclinical experimental validation, underscores the universality and therapeutic relevance of T2M biology. Together, these findings highlight the ability of scMINER2 to overcome the curse of dimensionality, while preserving causal biological signals, reframing type-2 immunity as a hidden dimension of CAR-T persistence and durable cancer immunotherapy.
Other Authors: Jayadev Mavuluri, Song-Eun Lim, Darong Yang, Sheetal Bhatara, Xu Yang, Kevin Ye, Jai Mehta, Noemi Reyes, Mahesh Pujyan, Chun-Yang Lin, Qingfei Pan, Liang Ding, Koon-Kiu Yan, Terrence L. Geiger, Jiyang Yu
Clone-Specific Pathway and Program Inference in Low- to High-Grade IPMN Precursors of Pancreatic Cancer
Rachel Karchin, Johns Hopkins University
Intraductal papillary mucinous neoplasms (IPMNs) are common precursors to pancreatic cancer, yet the clonal and transcriptional transitions underlying progression from low- to high-grade lesions remain poorly understood. We collected 10 archival FFPE IPMN specimens containing both low- and high-grade regions and have begun in-depth analysis using Visium HD spatial transcriptomics and whole-exome sequencing. To interpret these data, we extended PictographPlus, originally developed to integrate bulk DNA and RNA sequencing for clonal evolution and clone-level expression inference. Because spatial transcriptomic data lack DNA information, we adapted PictographPlus to use copy-number clones inferred from spatial RNA profiles and identify altered pathways across clonal transitions using GSEA. CoGAPS identifies latent biological patterns from high-dimensional data, and ProjectR projects these patterns into independent datasets to enable robust cross-study and cross-platform comparisons. Here, we applied CoGAPS and ProjectR to transfer-learn patterns from a publicly-available bulk RNA-sequencing dataset containing IPMN, PDAC, and normal ductal epithelial samples. The patterns, representing cohort-level transcriptional programs, were projected onto our own IPMN Visium HD data to visualize their spatial localization. This dual framework of clone-aware pathway inference with PictographPlus and orthogonal projection of CoGAPS patterns with ProjectR will provide complementary insights into clonal architecture, transcriptional phenotypes, and microenvironmental interactions in early pancreatic tumorigenesis. Unlike existing approaches that rely solely on transcriptional profiling, our framework explicitly links spatial expression patterns to clonal evolution, establishing a novel paradigm for dissecting precancerous progression.
Other Authors: Jiaying Lai, Kathleen Noller, Prathima Nagendra, Adonis Dmello, Luciane Kagohara, Laura Wood, Elana Fertig
Perturb-DBiT for Spatially Resolved, Whole-Transcriptome In Vivo CRISPR Screens
Alev Baysoy, Yale University
Spatially resolved in vivo CRISPR screening combines gene editing with spatial transcriptomics to interrogate how genetic perturbations reshape gene expression within native tissue contexts. Current approaches are limited to small perturbation panels and capture only a subset of protein-coding RNAs. Here, we introduce Perturb-DBiT, a versatile method for simultaneous, base-level co-sequencing of spatial whole-transcriptome total RNA and single-guide RNAs (sgRNAs) on the same tissue section. This approach enables unbiased mapping of how genetic perturbations influence RNA regulation, cellular behavior, and tissue organization in situ. Applying Perturb-DBiT to a human cancer metastatic colonization model, we profiled large sgRNA panels across tumor colonies alongside their full RNA transcriptomes in consecutive tissue sections. This revealed previously unrecognized perturbation effects on long non-coding RNA (lncRNA) co-variation, microRNA–mRNA interactions, and tRNA alterations linked to amino acid metabolism, migration, and tumor growth. Integration with transcriptional pseudotime trajectories further uncovered perturbation-driven changes in clonal dynamics and cellular cooperation. In an immune-competent syngeneic mouse model, Perturb-DBiT enabled the spatial analysis of genetic perturbations within the tumor immune microenvironment, uncovering distinct and synergistic effects on immune infiltration and suppression. Overall, Perturb-DBiT provides a spatially resolved, comprehensive view of how gene knockouts impact diverse molecular and cellular processes—including small and large RNA regulation, tumor proliferation, metastasis, and immune interactions—offering a panoramic framework for understanding perturbation responses in complex tissues.
Other Authors: Paul Renauer, Xiaolong Tian, Feifei Zhang
Regulation of macropinocytosis in cancer cell settings with elevated Ras activity
Matthew Lazzara, University of Virginia
Oncogenes simultaneously confer proliferative advantages to and exert stresses on malignant cells. For example, mutant Ras drives tumor growth, but the energy demands of Ras-induced phenotypes lead malignant cells to engulf large quantities of nutrient-containing extracellular fluid via macropinocytosis. Overly rapid accumulation of macropinocytic vesicles stresses cells, leading to a non-apoptotic form of death termed methuosis. To identify the mechanisms and cell settings that promote or restrain macropinocytosis, we are integrating mechanistic modeling and high-content imaging approaches. It was previously shown that ectopic expression of Ras mutants in glioblastoma cells, which virtually never harbor RAS mutations, drives methuosis. We demonstrated that EGFR amplification, a common occurrence in glioblastoma, reproduces the methuosis phenotype but only in a subset of glioblastoma cells. To predict mechanisms that restrain methuosis, we developed a mechanistic model of EGFR-mediated signaling including a moving-boundary receptor endocytosis module. The model generated the nonintuitive but experimentally validated prediction that increased abundance of EGFR-containing endosomes impairs Ras activity through a reaction-diffusion mechanism involving the protein tyrosine phosphatase SHP2. Thus, natural variation among cells in EGFR abundance and endocytosis may create a path for EGFR-amplified disease progression without methuosis. In related studies, we tested growth factors and hypoxia as drivers of heterogeneous macropinocytosis in KRas-mutant pancreas cancer cells. Within cell populations, macropinocytosis correlated with epithelial-mesenchymal transition. We hypothesize this occurred because of macropinocytosis-permissive cytoskeletal rearrangements in more mesenchymal cells. Inferences from mutual information models based on multiplexed immunofluorescence imaging are being tested to identify signaling pathways that regulate these phenomena.
Other Authors: Sung Hyun Lee, Alex DeWalle, Seth Boehringer
Identification of immunotherapeutic targets in the GBM:immune microenvironment
Forest White, MIT
Over the past decade, checkpoint inhibitors have transformed patient care. for a variety of tumor types. Unfortunately, this success has not extended to glioblastoma, as multiple clinical trials with monotherapy checkpoint inhibitors or combination immunotherapies have failed. Targeted immunotherapies, including BiTES, CAR-T cells, and peptide/mRNA immunostimulating therapies have, in some diseases, shown improved therapeutic efficacy relative to immune checkpoint inhibitors. To identify antigens for these targeted immunotherapies we developed a quantitative immunopeptidomics platform and applied this platform to interrogate MHC I peptides displayed by GBM tumor cells and macrophages, alone or in co-culture. We showed that co-culture polarized macrophages to a TAM-like state, resulting in an altered immunopeptide repertoire. Similarly, tumor cells adapted to co-culture by altering their MHC I peptide presentation, with increased presentation of >20-fold for several GBM tumor cell peptides following co-culture. After validating the presentation of these co-cultured induced macrophage- or tumor cell-associated peptides in vivo, we developed an mRNA immunostimulatory therapy targeting multiple of these peptides. Treatment with this therapy resulted in tumor shrinkage for several days, generated a peptide-targeting immune response, and resulted in increased T-cell infiltration into the GBM tumors. However, overall survival was not improved, potentially due to T cell exhaustion in the tumor microenvironment. To improve therapeutic response, we have now identified a cohort of shared splicing-derived GBM tumor neoantigens, quantified tumor-specific vs. DC cross-presented peptides, and have begun to test the role of radiotherapy in potentiating immune response by stimulating a more inflammatory microenvironment.
Other Authors: Yufei Cui, Amy Wisdom, Heidi Temple, Kien Phuong, Franziska Michor, Stefani Spranger
A proteome-wide structurally resolved protein interactome for an atomistic perspective of cancer signaling
Diana Murray, Columbia University
Protein networks are ubiquitous, powerful tools in systems biology and the ways they can be applied to understand biological phenomena grow ever more diverse. There are many large protein-protein interaction (PPI) databases derived primarily from sequence information, literature curation, and low- and high-throughput experimental methods. However, most do not incorporate 3D-structural information. A goal of the Columbia Center for Cancer Systems Therapeutics (CaST) is to leverage proteome-wide structure-based PPI networks to model cellular signaling events. We previously developed a structure-based algorithm for proteome-wide prediction of PPIs mediated by structured domains. During the past year, we developed an additional pipeline to predict PPIs mediated by the binding of short linear motifs (SLiMs) in one protein to structured domains in another and identified high-confidence interactions involving 60 SLiM-recognition domains, ranging from those that are well-studied (e.g. Src-Homology domains) to those not accessible by other computational approaches (e.g. PDZ, BRCT, and Bcl-2 homology domains). A high-confidence human interactome that combines domain-domain and domain-SLiM interactions is comprised of 735,000 binary physical PPIs and serves as the reference interactome supporting much of the research at CaST. Algorithmic clustering of the CaST Interactome reveals organization at different scales, from subnetworks to protein complexes. Quite remarkably, individual clusters, generated entirely from structural evidence, contain proteins with a common biological theme allowing the identification of groups of interacting proteins that carry out distinct biological processes. Clusters that pertain directly to the Hallmarks of Cancer reveal expanded hallmark networks, crosstalk between networks and, in addition, suggest novel drug targets.
Other Authors: Aakash Saha, Barry Honig
Characterizing the metabolic landscape of high-grade glioma with respect to transcriptional subtype and imaging region
Kristin Swanson, Cedars-Sinai
High-grade glioma (HGG) continues to have a dismal prognosis owing in part to its heterogeneous and invasive nature both within and between patients. The most aggressive form of high-grade glioma, called glioblastoma, portends a median survival of around 15 months. Like many cancers, high-grade glioma rewires metabolism towards its own benefit, shifting away from the typical reliance of brain tissue on glucose towards other sources such as fatty acids and acetate. Using a cohort of image-localized high-grade glioma biopsies collected from diverse imageable regions with associated transcriptomics (58 patients, 202 biopsies), we are able to assess the spatial transcriptomic landscape of HGG. We have previously shown that the transcriptome of HGG can be characterized into a continuous transition between states, and we have shown that the cellular ecosystem follows a similar pattern, using a combination of trajectory inference and cellular deconvolution. These patterns are strongly associated with the non-enhancing and contrast-enhancing regions of the tumors, which are typically considered for treatment planning and progression assessment. Here, we applied differential expression of hallmark pathways to determine how metabolism factors into this paradigm. As expected, we find overall higher levels of metabolism-associated pathway expression in the contrast enhancing regions compared to the non-enhancing regions. Interestingly, we found metabolic expression differs between the ecologies associated with transcriptional subtypes of high-grade glioma. For example, fatty acid (p=0.003), xenobiotic (p
Other Authors: Lee Curtin, Kyle W. Singleton, Pamela R. Jackson, Peter Canoll, Kristin R. Swanson
Led by Todd Stukenberg, UVA, on behalf of the Education & Outreach Working Group
Scientists are increasingly recognizing the importance of communicating why science matters—not only for advancing knowledge, but also for improving lives and fueling the American economy. The Outreach Core directors of the CSBC centers have been actively developing programs to promote scientific literacy, from middle school classrooms to medical training. Yet, in today’s climate, we all face new challenges in effectively advocating for science and engaging the public. This breakout session offers an open discussion and brainstorming opportunity to share strategies, identify tools, and explore how we can better champion science as one of the great challenges of our time.
Dr. Charles W.M. Roberts, MD, PhD
Executive Vice President
Director, Comprehensive Cancer Center
St. Jude Children's Research Hospital
Led by Amy Brock and David Basanta
During this session attendees will break out into small groups for a discussion on the future of Cancer Systems Biology. Podium presenters in Sesion 1-4 are asked to include one slide in their presentation that addresses critical unmet needs in their work and the cancer systems biology community that could be addressed in the next five years.
Co-evolution of tumor progenitors and their microenvironment in progression of lung premalignancy to adenocarcinoma
Humam Kadara, The University of Texas MD Anderson Cancer Center
The co-evolution and interactions of different cell subsets in progression of precursor lesions to lung adenocarcinoma (LUAD) are incompletely understood. We generated spatial transcriptomic maps of 56 human precursor lesions and LUADs from 25 patients and of an independent cohort of 36 lesions from 19 patients, analyzing a total of 486,519 spots and 5.4 million cells. We identify region-specific expression programs that distinguish precursors from LUADs. Spatially-resolved clonal architectures reveal patient-specific heterogeneity in evolution of precursors to LUADs. We find progenitors expressing tumor-associated meta-programs and residing in niches enriched with proinflammatory IL1B high macrophages. Epithelial-proinflammatory niches are prevalent in precursor lesions but become less frequent in LUADs. Epithelial-inflammatory niches are stage-specific, shape early LUAD development and represent promising targets for interception.
Other Authors: Fuduan Peng, Ansam Sinjab, Yibo Dai, Sujuan Yang, Minyue Chen, Linghua Wang
Oncogenic and tumor-suppressive forces converge on a progenitor-orchestrated niche to shape early tumorigenesis
José Reyes, Memorial Sloan Kettering Cancer Center
The transition from benign to malignant growth is a pivotal yet poorly understood step in cancer progression that marks the shift from a pathologically inert condition to a clinically lethal disease. Here, we integrate lineage tracing, single-cell and spatial transcriptomics to visualize the molecular, cellular and tissue-level events that promote or restrain malignancy during the tumor initiation in mouse models of pancreatic ductal adenocarcinoma (PDAC). We identify a discrete progenitor-like population of KRAS-mutant cells that co-activates oncogenic and tumor-suppressive programs-including p53, CDKN2A, and SMAD4-engaging senescence-like responses and remodeling their microenvironment, ultimately assembling a niche that mirrors invasive PDAC. KRAS inhibition depletes progenitor-like cells and dismantles their niche. Conversely, p53 suppression enables progenitor cell expansion, epithelial-mesenchymal reprogramming, and immune-privileged niche formation. These findings position the progenitor-like state as the convergence point of cancer-driving mutations, plasticity, and tissue remodeling-revealing a critical window for intercepting malignancy at its origin.
Other Authors: Isabella Del Priore, Andrea C Chaikovsky, Nikhita Pasnuri, Ahmed M Elhossiny, Tobias Krause, Andrew Moorman, Catherine Snopkowski, Meril Takizawa, Cassandra Burdziak, Nalin Ratnayeke, Ignas Masillioni, Yu-Jui Ho, Ronan Chaligné, Paul B Romesser, Aveline Filliol, Tal Nawy, John P Morris 4th, Zhen Zhao, Marina Pasca Di Magliano, Direna Alonso-Curbelo, Dana Pe'er, Scott W Lowe
Mouse lymph node colonization models predict human tumor metastasis in melanoma
Brendan Ball, Stanford University
Aggressive tumor progression and distant metastases of cancers such as skin cutaneous melanoma is thought to be initiated by lymph node (LN) involvement. Previous mouse studies have selectively enriched for tumor cells with enhanced LN migratory capacity but have not been directly translated to human outcomes. As a result, the involvement of LN colonization in distant metastasis remains unclear. To translate findings from mouse models to human, we used a computational framework termed Translatable Components Regression (TransComp-R) to synthesize mouse tumor samples from the LN with human primary and metastatic tumors. Using TransComp-R, we identified a principal component encoding transcriptomic variation of early- and late-stage mouse tumor generations that stratified primary and metastatic tumors from human melanoma patients. We identified enriched biological pathways associated with the cell cycle and G2/M checkpoint. To reveal potential cell types involved in tumor LN colonization, we applied CIBERSORTx, a deconvolution approach to infer cell type proportions. Notably, we identified differentially abundant CD4 memory activated and follicular helper T cells across human metastatic and primary tumor samples. In both cells, we identified a panel of genes driving LN colonization that were also differentially expressed across metastatic and primary tumors in human. These cells and genes may contribute to the tumor-immune microenvironment that facilitates LN colonization in cases of melanoma. Nevertheless, our synthesis of pre-clinical models with human data reveals potential biological pathways and transcriptomic signatures for future investigation.
Other Authors: Brendan K. Ball, Andrew J. Gentles
ErbB Pathway Activation and SMARCD3 Regulons Drive Resistance in ER+ Breast Cancer
Aritro Nath, City of Hope
Estrogen receptor-positive breast cancer (ER+BC) constitutes approximately 70% of all breast cancer cases, with a significant subset developing resistance to endocrine therapy and progressing to metastatic disease. While the addition of CDK4/6 inhibitors (CDK4/6i) has improved outcomes, resistance remains a major clinical challenge. Using single-cell RNA sequencing of longitudinal samples from the FELINE trial, we identified a consistent activation of the ErbB signaling pathway in resistant tumors, marked by upregulation of ERBB2/ERBB4 and downstream transcription factors. This shift supports an estrogen-independent proliferative state. To counteract this, we combined ribociclib with afatinib in 3D ER+BC models, revealing synergistic effects that restored G1/S and G2/M checkpoint control and induced apoptosis via dual inhibition of CDK2 and CDK1 complexes. Complementing this, transcriptomic profiling of everolimus-resistant ER+BC cells uncovered persistent activation of IGF1R, ESR1, RAS, and MAPK pathways. Network modeling implicated SMARCD3 regulons as key resistance drivers, with elevated activity in patient tumors. Co-treatment with everolimus and trametinib effectively suppressed resistant cell growth, highlighting MEK1/2 inhibition as a viable strategy. Finally, phylogenetic reconstruction of end-stage metastatic tumors using scRNA-seq revealed genetic and transcriptomic heterogeneity, immune suppression, and clonal expansion. These findings underscore the complexity of resistance mechanisms and the need for multi-targeted therapeutic approaches. Together, these studies illuminate the dynamic plasticity of ER+BC and propose rational combination therapies to overcome resistance and improve patient outcomes.
Other Authors: Jason Griffiths, Kimya Karimi, Eric Medina, Isaac Bishara, Feng Chi, Eleni Farmaki, Patrick Cosgrove, Andrea Bild
Phase i trials in cancer: From board to bench to bedside and back again
Alexander Anderson, Moffitt Cancer Center
Modern cancer treatments are largely target-driven, when a targetable mutation is present, with patients receiving the maximum tolerated dose (MTD) therapies. While these approaches often yield strong initial responses, cancers—complex evolving systems—adapt through resistance mechanisms, leading to relapse. Although the role of evolution in tumor progression, metastasis, and treatment response is increasingly recognized, MTD strategies remain dominant in precision oncology. Evolutionary therapy offers an alternative paradigm by exploiting how cancers adapt under treatment through optimized dosing and sequencing, often guided by mathematical modeling. Adaptive therapy, a form of evolutionary therapy, seeks to delay resistance by maintaining competition between drug-sensitive and resistant populations to control tumor burden. Developed through mathematical insights, adaptive therapy has shown promise in preclinical models (prostate, ovarian, melanoma, breast) and pilot clinical trials (NCT02415621; NCT05189457; NCT03543969). This talk will discuss how mathematical models, integrating differential equations and deep reinforcement learning, can optimize treatment strategies such as adaptive therapy and drive Phase i (imaginary) trials. We will highlight how model-driven virtual patients can capture variability in tumor dynamics and treatment response, bridge bench to bedside research, and be calibrated from historical clinical data to generate virtual cohorts. Such cohorts can power Phase I trials, inform treatment stratification and switching thresholds, predict trial outcomes, and guide scheduling decisions. We will also examine how appointment frequency influences outcomes and how robust adaptive therapy remains when patients miss visits. By integrating model-driven insights into trial design, we aim to accelerate translation of evolutionary therapies and improve durable cancer control.
Other Authors: Maximilian Strobl, Kit Gallagher, Jill Gallaher, Mark Robertson-Tessi, Philip Maini, Robert Gatenby
Leveraging AI for Brain Tumor Resection: Predicting 5-ALA Fluorescence on Preoperative MRI
Pamela Jackson, Cedars Sinai
High-grade gliomas (HGGs) are aggressive, infiltrative brain tumors with poor survival outcomes and limited therapeutic options. To enhance surgical strategies, fluorescence-guided surgery (FGS) using 5-aminolevulinic acid (5-ALA) has emerged as a transformative approach in neurosurgical oncology. We have developed a radiomics-based machine learning model to predict 5-ALA fluorescence using preoperative MRI features. Using the Medtronic StealthStation neuronavigation system, we have collected over 1,000 spatially localized intraoperative biopsy samples from more than 250 patients undergoing resection for glioma. Each biopsy site was recorded with 3D coordinates, labeled in real time, flash-frozen, and annotated with metadata including collection time, anatomical location, and surgical plan. A subset of this data including 268 biopsies from 52 patients (25 male, 27 female) with reported intraoperative 5-ALA fluorescence was used to develop a radiomics model predicting locoregional patterns of 5-ALA positivity based on preoperative MRI. Image features were extracted from the biopsy sample coordinates across multiple MRI sequences and paired with corresponding 5-ALA fluorescence status to train the model on regional fluorescence prediction. The final model achieved 83% accuracy on the training set and 86% on the validation set using a 70-15-15 data split; this included 208 samples for training and 42 for validation, stratified by patient to ensure no overlap of samples from the same individual across sets. These findings demonstrate the feasibility of predicting 5-ALA fluorescence using radiomic features from standard preoperative MRI, offering a promising noninvasive strategy for improving intraoperative decision-making and enabling more precise, personalized resections in patients with high-grade gliomas.
Other Authors: Destiny Green, BS, Pamela R. Jackson, PhD, Lee Curtin, PhD, Kyle W. Singleton, PhD, Michael Vogelbaum, MD, PhD, William Curry, MD, Bob Carter, MD, PhD5, Alfredo Quinones-Hinojosa, MD, Bernard Bendok, MD, Kristin R. Swanson, PhD
Mathematical Modeling of JNK Pathway Dynamics to Optimize Therapeutic Strategies for Triple-Negative Breast Cancer Lung Metastasis
Tatiana Miti, H. Lee Moffitt Cancer Center
Triple-negative breast cancer (TNBC) remains a clinical challenge due to early metastatic dissemination and limited efficacy of current therapies. While chemotherapy is the standard of care, it paradoxically activates the JNK pathway, a stress-activated cascade that promotes metastatic initiation and survival in the lung microenvironment. Attempts to block individual components of JNK-mediated cross-talk between tumor cells, macrophages, endothelial cells, and fibroblasts have not eliminated metastases. Direct JNK inhibition reduced tumor–fibroblast signaling, extracellular matrix remodeling, and macrophage activation, while enhancing chemotherapy response, but was insufficient to eradicate metastases. Critically, therapeutic outcomes varied with the timing of inhibition (pre-seeding, pre-chemotherapy, or concurrent), underscoring the importance of temporal and ecological context. To systematically investigate these dynamics, we developed an agent-based model (ABM) of TNBC–lung ecosystem interactions. The model captures spatiotemporal changes in tumor cell proliferation, death, and cross-talk with stromal components, recapitulating experimental patterns observed in short-term xenografts and aligning with longer-term datasets. Using this framework, we evaluated eco-evolutionary dynamics of JNK+ cell interactions and identified treatment schedules that more effectively disrupt metastatic niche formation. Model predictions reveal that therapeutic success depends strongly on the sequence of interventions and the spatial context of tumor–stromal interactions during early metastatic colonization. This integrative modeling approach provides mechanistic insight into how ecological dynamics govern treatment outcomes and offers a strategy to rationally design therapies capable of reducing or eradicating TNBC metastatic burden. Beyond TNBC, this framework represents a broadly applicable paradigm for modeling metastatic niche ecology and optimizing therapeutic interventions across cancer types.
Other Authors: Bina Desai, Daria Myroshnychenko, Andriy Marusyk, David Basanta
Quantum Cancer Biology: Systematic Elucidation and pharmacologic targeting of Master Regulator dependencies presiding over coexisting pancreatic cancer subpopulations, at the single cell level
Andrea Califano, Chan Zuckerberg Biohub of New York
Abstract TBD
Other Authors: n/a
Face Behond the Findings
Angelique C. Graham Harlan, Patient Advocate, ALSAC
Defying the odds as a 30+ year survivor of acute myeloid leukemia treated at St. Jude Children’s Research Hospital, Angelique C. Graham Harlan channels resilience into transformative healthcare leadership. As Sr. Advisor – Development Operations at ALSAC, Angelique is renowned for driving enterprise-wide innovation in workforce development, onboarding, and inclusive culture-building. Her strategic vision has elevated employee experience through initiatives like the Development Attaché Program, standardized interview guides, and onboarding roadmaps.
Angelique’s journey is further distinguished by her clinical experience as a Hematology- Oncology- Bone Marrow Transplant Nurse Specialist treating our nation's leaders as well as her academic excellence and advocacy. She holds a Doctorate in Nursing Practice (DNP) in Nursing Administration/Leadership, serves as a Commission on Collegiate Nursing Education (CCNE) onsite evaluator, and is an ALSAC Global Advisor. Her expertise spans strategic planning, nursing leadership, and advocacy as an HPV champion and SJCRH PFCC advisor. Through operational excellence, cross-functional collaboration, and a deep personal connection to the mission of St. Jude, Angelique continues to shape the future of healthcare and inspire others with her unwavering commitment
Systems Biology Approaches to Pediatric Oncology
Franziska Michor, Dana-Farber Cancer Institute
The U54-funded KIDSROBIN center studies two pediatric cancers of neuro-ectodermal origin: diffuse midline glioma (DMG) and high-risk neuroblastoma (NBL). Both are genetically simple yet display striking intra-tumoral heterogeneity, making them well suited to a systems biology approach. Insights from such rare pediatric cancers have often proved broadly relevant to adult cancers. KIDSROBIN integrates single-cell genomics, epigenetics, and proteomics with computational modeling to chart the evolutionary dynamics of these tumors. A Data Science Core synthesizes these datasets into predictive models of tumor behavior and therapeutic response, while an AI and Imaging Core provides independent validation beyond genomics. Together, these efforts exemplify a systems-level strategy: linking molecular profiles, cellular states, and clinical outcomes to guide intervention design. Finally, the ROBIN-NEST training program extends the center’s impact by fostering cross-disciplinary education in systems biology and oncology, preparing the next generation of researchers and clinicians.
Other Authors: Daphne Haas-Kogan, David Kozono, Franziska Michor
Characterizing the spatial architecture of SMARCB1-deficient epithelioid sarcomas with Modulator
Wesley Tansey, Memorial Sloan Kettering Cancer Center
Epithelioid sarcoma (ES) is a rare mesenchymal neoplasm driven by biallelic inactivation of SMARCB1, mostly affecting adolescents and young adults. The clinical course of ES is generally unfavorable, with a fatal outcome in ~50% of cases despite aggressive multimodal therapies combining surgery, chemotherapy, and irradiation. Tazemetostat, an EZH2 inhibitor (EZH2i) recently FDA approved for ES, but the vast majority of tumors do not respond or develop resistance. The role of the tumor microenvironment (TME) in ES therapeutic response is generally unknown. We used the 10X Xenium platform to spatially profile 5K genes in 64 ES samples from a cohort of 19 patients. To analyze these data, we developed Modulator, a biologically-informed, spatially-aware method for automated cell typing in spatial transcriptomics data. Modulator is specifically crafted for rare cancers like ES, where no reference atlas or known marker gene list is available for typing diseased cells. Analyzing over 3B transcripts and 1.7M typed cells reveals a complex signaling environment that is strongly conserved across all samples, with ES cells co-opting every major cell type. Rather than shifting the TME in response to treatment, differential expression between pre- and post-Tazemetostat samples shows a cancer cell-intrinsic transition to a neural crest state. Spatial niche analysis implicates a tumor-induced M2-like macrophage population thwarting effector T cell development in budding tertiary lymphoid sites. Overall, our results suggest potential for both direct cytotoxic cancer cell targets and immune-stimulating TME targets for epithelioid sarcomas.
Other Authors: Jiayi Fan, Jeff Quinn, Alex Kentsis, Mrinal Gounder, Cristina Antonescu
Conception-calibrated pediatric tumor mitotic clocks
Darryl Shibata, USC School of Medicine
Molecular clocks can reconstruct tumorigenesis, but their calibration is limited by uncertainty in cancer ages. Pediatric cancers simplify this problem because age ranges are narrowly bounded by conception and a minimum of ~30 divisions. We developed mitotic clocks from rapidly fluctuating CpG (fCpG) DNA methylation on the X chromosome in male cancers, applying a binary Markov model to estimate mitotic age and epimutation rates. Across acute lymphoblastic leukemia, acute myeloid leukemia, neuroblastoma, and embryonal brain tumors, modeled and observed methylation data were highly concordant, yielding epimutation rates of ~10-3 per division. The clocks also resolved remission dynamics, inferring relapses seeded by small numbers of variably dormant residual cancer cells. By exploiting the unique age constraints of pediatric tumors, calibrated fCpG clocks provide a quantitative framework to reveal otherwise hidden features of human tumor evolution.
Other Authors: Jeremiah John
A hidden-driver approach identifies synergistic brain-penetrant drug combinations for high-risk medulloblastoma and beyond
Jiyang Yu, St. Jude Children's Research Hospital
Targeted strategies are urgently needed but remain elusive for high-risk Group 3 medulloblastoma, a metastatic, recurrence-prone subtype that responds poorly to current non-specific therapies. To accelerate discovery of targeted therapies for this malignancy, we developed SINBA (Synergy Inference by Data-driven Network-Based Bayesian Analysis), a systems biology framework that identifies covert disease drivers and then nominates synergistic drug pairs that target them. The inclusion of a medulloblastoma-specific gene network and a curated drug-gene interaction database allows SINBA to prioritize drug combinations that are both synergistic and blood-brain-barrier-permeable. Screening of 320 drug combinations derived from the in silico analysis of over 10,000 driver pairs produced 19 synergistic pairs; of these, a combination of MEK inhibitors and the p38 inhibitor regorafenib was particularly efficacious. This combination suppressed G3 tumor progression in mice, with therapeutic benefits further enhanced by low-dose radiation. Single-cell RNA sequencing found early UBC progenitor-like tumor cells to be especially sensitive to the combination treatment, which remodels the tumor microenvironment to promote antitumor immune responses. These results demonstrate SINBA’s great potential as a drug discovery platform, clinically actionable G3 MB strategy, and precision oncology framework.
Other Authors: Jingjing Liu, Xu Yang, Mingrui Zhu, Xinran Dong, Honglei Zhou, Brandon Bianski, Barbara M. Jonchere, Wenwei Lin, Xiang Fu, Abigail Wang,Ruilin Jiang, Lei Yang, Burgess B. Freeman III, Taosheng Chen, Giles W. Robinson, Martine F. Roussel, Thomas E. Merchant, Amar Gajjar