Biology subjects

Chen, Y.-A.

Publications and source records attributed to Chen, Y.-A..

At least 19 recordsLinked to original sources

bioRxiv2026

Telomere Dysfunction in Human Astrocytes Drives Acrocentric Chromosome Instability and Nucleolar Reorganization

Loss of p53 and Rb enables continued cell division despite progressive telomere erosion, ultimately triggering telomere crisis. While recent work has clarified mechanisms of chromatin bridge resolution, the longitudinal dynamics of structural damage through crisis remain incompletely understood. We transduced normal human astrocyte (NHA) cells with HPV18 E6/E7 and tracked them across extended population doublings. NHA E6/E7 cells showed progressive telomere shortening, anaphase bridges, and a growth plateau consistent with crisis. Multicolor FISH revealed subclonal chromosomal abnormalities largely invisible to short-read sequencing, with acrocentric chromosomes disproportionately affected: chromosome 13 was abnormal in over 92% of metaphases, and chromosomes 21 and 22 were similarly enriched. Translocations involving chromosome 13 were transient, replaced at later passages, while numerical aberrations persisted. Immunofluorescence revealed compact spherical nucleoli replaced by dispersed necklace-like structures, indicating a large-scale reorganization of nucleolar structure in response to telomere dysfunction. To determine whether these changes reflected altered chromosomal organization in the nucleus, we performed Hi-C and deployed KaryoScope, our alignment-free k-mer-based approach that recovers trans-chromosomal signal from repetitive acrocentric short arms. Inter-chromosomal contacts among nucleolar organizing region-bearing acrocentric chromosomes were markedly and persistently depleted in E6/E7 cells. Together, cytogenetic, imaging, and chromatin-contact data identify the nucleolus as a structural nexus linking telomere dysfunction to large-scale genomic rearrangement.

cancer biology↗

bioRxiv2026

KaryoScope: rapid, alignment-free sequence annotation for the pangenome era

The pangenome era is producing long-read sequencing data and complete genome assemblies (1-3) at a pace that current annotation methods cannot match. Existing tools were each built for a single feature class (repeats, centromeric satellites, or genes) and falter precisely where the genome is most variable and harbours clinically important variation: the centromeres, subtelomeres, and acrocentric short arms. Here we present KaryoScope, an alignment-free method to annotate an assembly at base-pair resolution across any desired feature classes in a single pass, completing in minutes on a standard workstation. Applied to the Human Pangenome Reference Consortium Release 2 assemblies (3), KaryoScope identifies the SST1 macrosatellite as the recurrent sequence at Robertsonian translocation fusion points (4, 5), delivers the first pangenome-wide census of D4Z4 macrosatellite structural diversity at the 4q and 10q subtelomeres relevant to facioscapulohumeral muscular dystrophy (6), and reveals previously uncharacterised centromere structural polymorphism, including chromosome-specific satellite loss and megabase-scale rearrangement validated by fluorescence in situ hybridization. A pre-built KaryoScope database for the human genome is distributed alongside the tool, and additional databases can be built for any reference genome or annotation source. Together, these capabilities bring the most variable regions of the genome within reach for comparative, clinical, and pangenome-scale analysis. KaryoScope is available at https://github.com/barthel-lab/KaryoScope.

bioinformatics↗

bioRxiv2026

BiLSTM-Powered Bilinear Attention for Protein-Ligand Prediction

Rapid and accurate prediction of protein-ligand bindings is essential for drug discovery. While generative AI has driven rapid advancements in structure-based approaches, sequence-based methods remain significantly faster and more cost-effective. Here, we present a weakly supervised deep learning framework integrating graph convolutional networks (GCN) for molecular encoding and bidirectional long short-term memory (BiLSTM) for protein modeling. The latter represents long-range dependencies better than the widely used convolutional neural network (CNN). Leveraging a bilinear attention network (BAN), this model learns protein-ligand pairwise interactions without requiring three-dimensional structural supervision. By using the publicly available BindingDB dataset, the model was trained, solely on affinity labels, and successfully classified binder and non-binders with AUROC of 0.96 and an AUPRC of 0.95. The model generates interpretable attention maps that serve as a "GPS" to locate binding sites. Remarkably, despite the lack of structural training data, it can pinpoint key contact residues confirmed by crystal structures. Our method could function as a scalable filter for giga-scale libraries, allowing rapid screening of drug candidates with direct structural insights into the protein-ligand interface.

bioinformatics↗

bioRxiv2026

A single-cell and spatial atlas of prostate cancer reveals the combinatorial nature of gene modules underlying lineage plasticity and metastasis

Prostate cancer encompasses a spectrum of disease states driven by complex cellular heterogeneity. To delineate the transcriptional programs underlying lineage plasticity and metastasis, we constructed a comprehensive single-cell atlas of 128 patients, spanning localized, castration-resistant, and metastatic disease. Lineage plasticity was prevalent in localized disease, with subsets of tumor cells adopting distinct basal-like and club-like states. Luminal-like cancer cells also displayed extensive lineage infidelity, defined not by a binary loss of identity but by the combinatorial erosion of luminal gene modules associated with higher grade and stage. In the metastatic setting, gene program association analysis (GPAS) identified a broad induction of cell-cycle gene modules across organ sites as well as an induction of organ-specific gene modules, including osteomimetic signaling in bone, neuro-migratory genes in brain, and erythroid-like transitions in liver. Neuroendocrine prostate cancers (NEPCs) were not monolithic but defined by combinations of NE-associated gene modules including a novel HES6 program. Notably, these modules were detected at intermediate levels in localized samples, suggesting molecular plasticity precedes histological transformation. We also developed a refined NE signature that could distinguish NEPC tumors more accurately than previously published signatures. Within the tumor microenvironment (TME), we observed an elevation of pro-inflammatory Th17 T-cells in African American patients and identified a rare Schwann cell population. Finally, we present PCformer, a transformer-based foundation model trained on >500,000 cells to automate cell-state classification. Together, this comprehensive atlas demonstrates the complex nature of gene modules underlying lineage infidelity and plasticity in cancer cells and highlights distinct immune and stromal populations within the tumor ecosystem.

genomics↗

bioRxiv2025

Integrative single-cell and spatial mapping of oxidative stress response uncovers GCLC+ mesenchymal tumor cell state linked with favorable outcomes in triple negative breast cancer

Inhibiting oxidative stress response (OSR) proteins has been suggested as a therapeutic strategy in triple negative breast cancer (TNBC). However, the cell type specificity and spatial distribution of OSR genes and proteins, such as GCLC and NQO1, is unknown. Using single cell and spatial transcriptomics datasets we found that OSR genes were highly expressed in TNBC tumor cells, which localized in spatial clusters. Multiplex immunofluorescence imaging of 345 TNBC samples from 186 patients demonstrated that OSR proteins GCLC and NQO1 exhibit distinct expression profiles across tumor, immune, and stromal cell populations and are elevated in inflamed histological regions. Tumor cell OSR protein expression was associated with the composition of the adjacent cellular neighborhood. Furthermore, we identified GCLC and vimentin positive (GCLC+VIM+) mesenchymal-like tumor cells, residing near immune cells and exhibiting increased proliferation and decreased anastasis signatures, suggesting sensitivity to chemotherapy. Across a panel of thirteen TNBC cell lines, GCLC expression was positively correlated with sensitivity to cisplatin. Cox regression analysis revealed that patients with higher proportions of GCLC+VIM+ tumor cells had a longer overall survival. Collectively, our results demonstrate that individual OSR proteins are expressed in distinct microenvironments and tumor cell states, potentially contributing to patient outcomes.

cancer biology↗

bioRxiv2025

Mapping the Telomeric 3D Interactome with Telomere-C Reveals Repetitive Element Hubs Associated with Telomere Maintenance

Telomeres are essential for genome integrity, but the accurate, high-resolution mapping of their three-dimensional (3D) chromatin interactions, a process thought to mediate gene regulation and telomere maintenance, has been limited by the repetitive nature of subtelomeric DNA and shortcomings of traditional sequencing-based chromatin conformation capture methodologies. To overcome these challenges, we developed Telomere Conformation Capture by sequencing (Telomere-C), a novel high-throughput approach for the genome-wide identification of telomeric chromatin interactions. We applied Telomere-C to three normal fibroblast cell lines (BJ, IMR90, WI38), one telomerase-positive cancer cell line (HeLa S3) and two cancer cell lines (U2-OS and WI38-VA13) that maintain telomeres by alternative lengthening of telomeres (ALT). We generated millions of telomere-associated reads per sample and identified up to 41,026 interaction peaks. We successfully validated previously-described telomeric interaction sites (e.g., TERT and DUX4), and critically, revealed a novel class of ultra-long-range interactions extending over 5 Mb away from the telomere, which surprisingly constitute over 80% of all identified contacts. Contrary to prior studies focusing on protein coding gene interactions, we discovered that the telomeric 3D interactome is overwhelmingly anchored at repetitive element hubs, particularly interstitial telomeric sequence (ITS), telomere-associated repeat 1 (TAR1) and D20S16 elements. We found that the clustering of these interactions correlates strongly with cell-type-specific telomere maintenance. Most notably, D20S16 telomeric interactions were uniquely and highly enriched in ALT cancer cells, suggesting a mechanistic link. Taken together, our study effectively constructs the first high resolution maps of the telomeric 3D interactome, redefining its scope to be dominated by ultra-long-range contacts with repetitive elements. This work provides fundamental insights into telomeric nuclear organization and establishes the telomeric D20S16 interaction as a molecular signature for the ALT pathway.

genomics↗

bioRxiv2025

Morphology-Aware Profiling of Highly Multiplexed Tissue Images using Variational Autoencoders

Spatial proteomics (highly multiplexed tissue imaging) provides unprecedented insight into the types, states, and spatial organization of cells within preserved tissue environments. To enable single-cell analysis, high-plex images are typically segmented using algorithms that assign marker signals to individual cells. However, conventional segmentation is often imprecise and susceptible to signal spillover between adjacent cells, interfering with accurate cell type identification. Segmentation-based methods also fail to capture the morphological detail that histopathologists rely on for disease diagnosis and staging. Here, we present a method that combines unsupervised, pixel-level machine learning using autoencoders with traditional segmentation to generate single-cell data that captures information on protein abundance, morphology, and local neighborhood in a manner analogous to human experts while overcoming signal spillover. We demonstrate the generality of this approach by applying it to CyCIF, Lunaphore COMET, and CODEX data, and show that it can learn histological features across multiple spatial scales.

pathology↗

bioRxiv2025

Spatial determinants of tumor cell dedifferentiation and plasticity in primary cutaneous melanoma

Localized cutaneous melanoma can be cured by excision but success critically depends on early detection and risk assessment of primary lesions. However, their initiation, progression, and immunology remain poorly understood, partly due to high intra- and inter-tumor heterogeneity. We studied this heterogeneity using spatial profiling in over 300 histological domains, each representing a single progression stage, and found that 200-600 cell neighborhoods from a single melanoma can be as different in RNA and protein expression as neighborhoods from different tumors. These differences are not stochastic, however, and disease progression can be mapped at the neighborhood level onto a cell state landscape defined by the activity of the melanocyte master regulator MITF and genes associated with a dedifferentiated neural crest phenotype. Position in this landscape is influenced by proximity to immune cells, perivascular environments, and other tissue features, but no single association is absolute, giving rise to complex spatial patterns.

cancer biology↗

bioRxiv2024

Multimodal Spatial Profiling Reveals Immune Suppression and Microenvironment Remodeling in Fallopian Tube Precursors to High-Grade Serous Ovarian Carcinoma

High-Grade Serous Ovarian Cancer (HGSOC) originates from fallopian tube (FT) precursors. However, the molecular changes that occur as precancerous lesions progress to HGSOC are not well understood. To address this, we integrated high-plex imaging and spatial transcriptomics to analyze human tissue samples at different stages of HGSOC development, including p53 signatures, serous tubal intraepithelial carcinomas (STIC), and invasive HGSOC. Our findings reveal immune modulating mechanisms within precursor epithelium, characterized by chromosomal instability, persistent interferon (IFN) signaling, and dysregulated innate and adaptive immunity. FT precursors display elevated expression of MHC-class I, including HLA-E, and IFN-stimulated genes, typically linked to later-stage tumorigenesis. These molecular alterations coincide with progressive shifts in the tumor microenvironment, transitioning from immune surveillance in early STICs to immune suppression in advanced STICs and cancer. These insights identify potential biomarkers and therapeutic targets for HGSOC interception and clarify the molecular transitions from precancer to cancer. STATEMENT OF SIGNIFICANCEThis study maps the immune response in fallopian tube precursors of high-grade serous ovarian cancer, highlighting localized interferon signaling, CIN, and competing immune surveillance and suppression along the progression axis. It provides an explorable public spatial profiling atlas for investigating precancer mechanisms, biomarkers, and early detection and interception strategies.

cancer biology↗

bioRxiv2024

Integrating spatial profiles and cancer genomics to identify immune-infiltrated mismatch repair proficient colorectal cancers

Predicting the progression of solid cancers based solely on genetics is challenging due to the influence of the tumor microenvironment (TME). For colorectal cancer (CRC), tumors deficient in mismatch repair (dMMR) are more immune infiltrated than mismatch repair proficient (pMMR) tumors and have better prognosis following resection. Here we quantify features of the CRC TME by combining spatial profiling with genetic analysis and release our findings via a spatially enhanced version of cBioPortal that facilitates multi-modal data exploration and analysis. We find that [~]20% of pMMR tumors exhibit similar levels of T cell infiltration as dMMR tumors and that this is associated with better survival but not any specific somatic mutation. These T cell-infiltrated pMMR (tipMMR) tumors contain abundant cells expressing PD1 and PDL1 as well as T regulatory cells, consistent with a suppressed immune response. Thus, like dMMR CRC, tipMMR CRC may benefit from immune checkpoint inhibitor therapy. SIGNIFICANCEpMMR tumors with high T cell infiltration and active immunosuppression are identifiable with a mid-plex imaging assay whose clinical deployment might double the number of treatment-naive CRCs eligible for ICIs. Moreover, the low tumor mutational burden in tipMMR CRC shows that MMR status is not the only factor promoting immune infiltration.

cancer biology↗

bioRxiv2024

Cross-institutional HER2 assessment via a computer-aided system using federated learning and stain composition augmentation

The rapid advancement of precision medicine and personalized healthcare has heightened the demand for accurate diagnostic tests. These tests are crucial for administering novel treatments like targeted therapy. To ensure the widespread availability of accurate diagnostics with consistent standards, the integration of computer-aided systems has become essential. Specifically, computer-aided systems that assess biomarker expression have thrusted through the widespread application of deep learning for medical imaging. However, the generalizability of deep learning models has usually diminished significantly when being confronted with data collected from different sources, especially for histological imaging in digital pathology. It has therefore been challenging to effectively develop and employ a computer-aided system across multiple medical institutions. In this study, a biomarker computer-aided framework was proposed to overcome such challenges. This framework incorporated a new approach to augment the composition of histological staining, which enhanced the performance of federated learning models. A HER2 assessment system was developed following the proposed framework, and it was evaluated on a clinical dataset from National Taiwan University Hospital and a public dataset coordinated by the University of Warwick. This assessment system showed an accuracy exceeding 90% for both institutions, whose generalizability outperformed a baseline system developed solely through the clinical dataset by 30%. Compared to previous works where data across different institutions were mixed during model training, the HER2 assessment system achieved a similar performance while it was developed with guaranteed patient privacy via federated learning.

pathology↗

bioRxiv2023

RPL22 is a tumor suppressor in MSI-high cancers and a key splicing regulator of MDM4

Microsatellite instability high (MSI-H) tumors are malignant tumors that, despite harboring a high mutational burden, often have intact TP53. One of the most frequent mutations in MSI-H tumors is a frameshift mutation in RPL22, a ribosomal protein. Here, we identified RPL22 as a modulator of MDM4 splicing through an alternative splicing switch in exon 6. RPL22 loss increases MDM4 exon 6 inclusion, cell proliferation, and augments resistance to the MDM inhibitor Nutlin-3a. RPL22 represses expression of its paralog, RPL22L1, by mediating the splicing of a cryptic exon corresponding to a truncated transcript. Therefore, damaging mutations in RPL22 drive oncogenic MDM4 induction and reveal a common splicing circuit in MSI-H tumors that may inform therapeutic targeting of the MDM4-p53 axis and oncogenic RPL22L1 induction.

cancer biology↗

bioRxiv2023

uniLIVER: a Human Liver Cell Atlas for Data-Driven Cellular State Mapping

The liver performs several vital functions such as metabolism, toxin removal and glucose storage through the coordination of various cell types. The cell type compositions and cellular states undergo significant changes in abnormal conditions such as fatty liver, cirrhosis and liver cancer. As the recent breakthrough of the single-cell/single-nucleus RNA-seq (sc/snRNA-seq) techniques, there is a great opportunity to establish a reference cell map of liver at single cell resolution with transcriptome-wise features. In this study, we build a unified liver cell atlas uniLIVER by integrative analyzing a large-scale sc/snRNA-seq data collection of normal human liver with 331,125 cells and 79 samples from 6 datasets. Besides the hierarchical cell type annotations, uniLIVER also proposed a novel data-driven strategy to map any query dataset to the normal reference map by developing a machine learning based framework named LiverCT. Applying LiverCT on the datasets from multiple abnormal conditions (1,867,641 cells and 439 samples from 12 datasets), the alterations of cell type compositions and cellular states were systematically investigated in liver cancer.

bioinformatics↗

bioRxiv2023

Single cell transcriptional profiling of benign prostatic hyperplasia reveals a progenitor-like luminal epithelial cell state within an inflammatory microenvironment

Benign prostatic hyperplasia (BPH) is characterized by excessive cell proliferation and inflammation and affects most aging men. The development of new therapies for BPH requires a deeper understanding of the underlying pathophysiology and cellular components of BPH. Here, we characterize at single cell resolution the cellular states of BPH and identify cell populations enriched in BPH that contribute to cell proliferation and inflammation. Single-cell RNA-sequencing was performed on prostate tissue from 15 patients undergoing holmium laser enucleation of the prostate for treatment of BPH. Clustering and differential expression analysis on aligned single cell RNA-seq data was performed to annotate all cell types. Pseudotime, gene set enrichment, gene ontology, and ligand-receptor analyses were performed. 16,234 cells were analyzed and specific stromal, epithelial, and immune subgroups were found to be strongly associated with inflammation. A rare luminal subgroup was identified and pseudotime analysis indicated this luminal subgroup was more closely related to club and basal cells. Using a gene set derived from epithelial stem cells, we found that this luminal subgroup had a significantly higher stem cell signature score than all other epithelial subgroups, suggesting this subgroup is a luminal precursor state. Ligand-receptor interactions between stromal, epithelial, and immune cells were explored with CellPhoneDB. Unique interactions highlighting MIF, a pro-inflammatory cytokine that promotes epithelial cell growth and inflammation in the prostate, were found between fibroblasts and the progenitor luminal subgroup. This luminal subgroup also interacted with neutrophils and macrophages through MIF. Our single-cell profiling of BPH provides a roadmap for inflammation-linked cell subgroups and highlights a novel luminal progenitor subgroup interacting with other cell groups via MIF that may contribute to the inflammation and cell proliferation phenotype associated with BPH.

genomics↗

bioRxiv2023

Longitudinal and multimodal auditing of tumor adaptation to CDK4/6 inhibitors in HR+ metastatic breast cancers

CDK4/6 inhibitors (CDK4/6i) have transformed the treatment of hormone receptor-positive (HR+), HER2-negative (HR+) breast cancers as they are effective across all clinicopathological, age, and ethnicity subgroups for metastatic HR+ breast cancer. In metastatic ER+ breast cancer, CDK4/6i lead to strong and consistent improvement in survival across different lines of therapy. To improve understanding of how metastatic HR+ breast cancers become refractory to CDK4/6i, we have created a multimodal and longitudinal tumor atlas to investigate therapeutic adaptations in malignant cells and in the tumor immune microenvironment. This atlas is part of the NCI Cancer Moonshot Human Tumor Atlas Network and includes seven pairs of pre- and on-progression biopsies from five metastatic HR+ breast cancer patients treated with CDK4/6i. Biopsies were profiled with bulk genomics, transcriptomics, and proteomics as well as single-cell ATAC-seq and multiplex tissue imaging for spatial, single-cell resolution. These molecular datasets were then linked with detailed clinical metadata to create an atlas for understanding tumor adaptations during therapy. Analysis of our atlas datasets suggests a diverse set of tumor adaptations to CDK4/6i therapy. Malignant cells may adapt to therapy via mTORC1 activation, cell cycle bypass, and increased replication stress. The tumor immune microenvironment displayed evidence of both immune activation and immune suppression during therapy. Together, our metastatic ER+ breast cancer atlas represents a rich multimodal resource to better understand HR+ breast cancer tumor therapeutic adaptations to CDK4/6i therapy.

cancer biology↗

bioRxiv2023

Anaplerotic nutrient stress drives synergy of angiogenesis inhibitors with therapeutics targeting tumor metabolism

Tumor angiogenesis is a cancer hallmark, and its therapeutic inhibition has provided meaningful, albeit limited, clinical benefit. While anti-angiogenesis inhibitors deprive the tumor of oxygen and essential nutrients, cancer cells activate metabolic adaptations to diminish therapeutic response. Despite these adaptations, angiogenesis inhibition incurs extensive metabolic stress, prompting us to consider such metabolic stress as an induced vulnerability to therapies targeting cancer metabolism. Metabolomic profiling of angiogenesis-inhibited intracranial xenografts showed universal decrease in tricarboxylic acid cycle intermediates, corroborating a state of anaplerotic nutrient deficit or stress. Accordingly, we show strong synergy between angiogenesis inhibitors (Avastin, Tivozanib) and inhibitors of glycolysis or oxidative phosphorylation through exacerbation of anaplerotic nutrient stress in intracranial orthotopic xenografted gliomas. Our findings were recapitulated in GBM xenografts that do not have genetically predisposed metabolic vulnerabilities at baseline. Thus, our findings cement the central importance of the tricarboxylic acid cycle as the nexus of metabolic vulnerabilities and suggest clinical path hypothesis combining angiogenesis inhibitors with pharmacological cancer interventions targeting tumor metabolism for GBM tumors.

cancer biology↗

bioRxiv2022

3D multiplexed tissue imaging reconstruction and optimized region-of-interest (ROI) selection through deep learning model of channels embedding

Tissue-based sampling and diagnosis are defined as the extraction of information from certain limited spaces and its diagnostic significance of a certain object. Pathologists deal with issues related to tumor heterogeneity since analyzing a single sample does not necessarily capture a representative depiction of cancer, and a tissue biopsy usually only presents a small fraction of the tumor. Many multiplex tissue imaging platforms (MTIs) make the assumption that tissue microarrays (TMAs) containing small core samples of 2-dimensional (2D) tissue sections are a good approximation of bulk tumors although tumors are not 2D. However, emerging whole slide imaging (WSI) or 3D tumor atlases that employ MTIs like cyclic immunofluorescence (CyCIF) strongly challenge this assumption. In spite of the additional insight gathered by measuring the tumor microenvironment in WSI or 3D, it can be prohibitively expensive and time-consuming to process tens or hundreds of tissue sections with CyCIF. Even when resources are not limited, the criteria for region-of-interest (ROI) selection in tissues for downstream analysis remain largely qualitative and subjective as stratified sampling requires the knowledge of objects and evaluates their features. Despite the fact TMAs fail to adequately approximate whole tissue features, a theoretical subsampling of tissue exists that can best represent the tumor in the whole slide image. To address these challenges, we propose deep learning approaches to learn multi-modal image translation tasks from two aspects: 1) generative modeling approach to reconstruct 3D CyCIF representation and 2) co-embedding CyCIF image and Hematoxylin and Eosin (H&E) section to learn multi-modal mappings by a cross-domain translation for minimum representative ROI selection. We demonstrate that generative modeling enables a 3D virtual CyCIF reconstruction of a colorectal cancer specimen given a small subset of the imaging data at training time. By co-embedding histology and MTI features, we propose a simple convex optimization for objective ROI selection. We demonstrate the potential application of ROI selection and the efficiency of its performance with respect to cellular heterogeneity.

bioinformatics↗

bioRxiv2022

Multi-modal digital pathology for colorectal cancer diagnosis by high-plex immunofluorescence imaging and traditional histology of the same tissue section

Precision medicine is critically dependent on better methods for diagnosing and staging disease and predicting drug response. Histopathology using Hematoxylin and Eosin (H&E) stained tissue - not genomics - remains the primary diagnostic method in cancer. Recently developed highly-multiplexed tissue imaging methods promise to enhance research studies and clinical practice with precise, spatially-resolved, single-cell data. Here we describe the "Orion" platform for collecting and analyzing H&E and high-plex immunofluorescence (IF) images from the same cells in a whole-slide format suitable for diagnosis. Using a retrospective cohort of 74 colorectal cancer resections, we show that IF and H&E images provide human experts and machine learning algorithms with complementary information that can be used to generate interpretable, multiplexed image-based models predictive of progression-free survival. Combining models of immune infiltration and tumor-intrinsic features achieves a hazard ratio of [~]0.05, demonstrating the ability of multi-modal Orion imaging to generate high-performance biomarkers.

cancer biology↗