Biology subjects

Mills, G. B.

Publications and source records attributed to Mills, G. B..

At least 19 recordsLinked to original sources

bioRxiv2026

SCORPy: Lowering the computational barrier to reproducible multiplexed imaging spatial single cell proteomics analysis

Spatially resolved single-cell proteomic imaging technologies, including cyclic immunofluorescence (CycIF), generate high-dimensional data, critical for tissue-scale biological analysis. However, single-cell analysis remains computationally demanding, lacks standardization across platforms and is often inaccessible to experimental biologists without programming expertise. Here we present SCORPy (Single-Cell proteOmics Research Platform), a standalone, cross-platform desktop application that provides an end-to-end, code-free workflow for the analysis of single-cell proteomic data extracted from imaging experiments. SCORPy introduces methodological advances for preprocessing multiplexed imaging data: an exposure-aware, cycle-matched background correction strategy, and a normalization framework that harmonizes signal distributions across markers while enabling batch correction across experiments. These approaches are integrated with quality control, interactive thresholding and cell phenotyping using a hierarchical cell reference library, and downstream compositional and spatial analyses within a unified interface. Sample-level metadata can be incorporated throughout the workflow to support integrative analyses and facilitate generation of publication-ready visualizations. By combining robust preprocessing methods with an accessible implementation, SCORPy reduces computational barriers and promotes broader adoption of spatial single-cell proteomics analysis.

cell biology↗

bioRxiv2026

ACCREDIT: A Quality-Aware Agentic Engine for Cell-resolved Cross-modal Image Registration with Dynamic Iterative Tuning

Spatial omics across complementary modalities is transforming our understanding of tissue architecture. Realizing this potential requires accurate and robust registration of cross-platform molecular images with hematoxylin-and-eosin (H&E) sections, the primary morphological reference for pathology. Existing methods, however, often fail silently when image orientation is unknown, image contrast is inverted, or tissue overlap is incomplete, producing erroneous registrations without alerting users or attempting recovery. Here, we present ACCREDIT, a quality-aware agentic framework that redefines cross-modal registration as an adaptive decision-making process rather than a one-shot computation. ACCREDIT combines deterministic registration pipelines with a reference-free composite quality score that automatically evaluates registration quality and rejects plausible but biologically incorrect registrations. When registration quality is insufficient, a large language model (LLM)-based rescue agent autonomously diagnoses failure modes and selects targeted recovery strategies, while an optional strategy-learning module captures expert-validated corrections for future reuse. Across Xenium, CODEX, cell-boundary, and IHC-to-H&E registration tasks, ACCREDIT outperformed competing methods by detecting registration failures and improving alignment quality through automated recovery and rescue. Ultimately, ACCREDIT enables robust integration of histology and spatial molecular profiling, providing a foundation for translating spatial omics into routine H&E-based pathology workflows.

bioinformatics↗

bioRxiv2026

High-Density Wild-Type IL-2 Nanoparticles Preferentially Enhance CD8⁺ T-Cell Expansion and Reprogram the Tumor Microenvironment

Low response rates to immune checkpoint inhibitors (ICIs) in solid tumors are often driven by insufficient tumor-infiltrating CD8 T cells and immunosuppressive tumor microenvironment (TME). Although interleukin-2 (IL-2) potently expands and activates CD8 T cells, its clinical use is limited by rapid clearance, dose-limiting toxicity, and regulatory T cell (Treg) stimulation. Engineered IL-2 variants have not yet achieved meaningful clinical efficacy. Here, polymer-modified mesoporous silica nanoparticles displaying dense, unmodified wild-type IL-2 on their surface (IL2-NP) are developed, conferring proteolytic stability and tumor retention. IL2-NP enables avidity-mediated CD8 T cell binding and enhances proliferation and effector function without increased Treg binding or proliferation. Intratumoral IL2-NP expands CD8 T cells, increases CD8/Treg ratios, and reprograms TME through dendritic cell activation and M1-like macrophage polarization. IL2-NP induces regression of both treated and untreated distant colorectal tumors in a CD8 T cell-dependent manner. IL2-NP synergizes with ICIs and leads to complete tumor regression and immunological memory that protect against rechallenge. Treatment is well tolerated, with strong efficacy also observed in triple-negative breast and metastatic ovarian cancer models. Overall, intratumoral IL2-NP elicits robust systemic antitumor immunity, offering a promising strategy to enhance ICIs, cancer vaccines, and adoptive T-cell therapies. Graphical abstractThis work introduces a nanoparticle platform that overcomes major shortcomings of IL-2 immunotherapy by presenting wild-type IL-2 at high density on the nanoparticle surface, thereby increasing binding avidity to effector T cells. The resulting IL-2 nanoparticles enhance cytotoxic T cell expansion, reprogram the tumor microenvironment, and augment responses to immune checkpoint blockade to achieve robust ant-tumor immune response in mouse tumor models. O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=182 SRC="FIGDIR/small/738558v1_ufig1.gif" ALT="Figure 1"> View larger version (82K): org.highwire.dtl.DTLVardef@a53915org.highwire.dtl.DTLVardef@146c137org.highwire.dtl.DTLVardef@3734ccorg.highwire.dtl.DTLVardef@6b5aef_HPS_FORMAT_FIGEXP M_FIG C_FIG

bioengineering↗

bioRxiv2026

Tumor-immune trajectory context connects static tissue architecture to clinical outcomes

Multiplexed tissue imaging (MTI) has revealed recurrent tumor microenvironment (TME) architectures with prognostic value, yet these measurements are inherently static, obscuring dynamic changes in the TME that govern therapeutic response. Here, we introduce a trajectory-centric framework that reconstructs continuous TME dynamics by integrating agent-based mathematical modeling and simulation with state space analysis. This approach yields a mechanistically constrained reference landscape built entirely from in silico simulation, and onto which static patient biospecimens can be projected and mapped onto simulated TME trajectories. Systematic simulation of tumor-immune interactions in triple-negative breast cancer identifies six metastable TME states connected by transition pathways spanning immune control to immune escape. Mapping MTI data from two independent patient cohorts, including longitudinal samples from a randomized immunotherapy trial, validates this landscape by positioning individual biospecimens along inferred TME trajectories rather than in static states. We show that treatment-phase TME states, but not pre-treatment configurations, robustly predict immunotherapy response, and identical terminal states can arise from distinct trajectory histories corresponding to immune failure or resolved inflammation. Thus, this framework enables mechanistic simulations to define a reference dynamical landscape that serves as a coordinate system for interpreting static clinical spatial data, providing a principled basis for evaluating consistency, predictiveness, and clinical relevance across independent patient cohorts. Altogether, this study advances spatial tumor profiling from static state classification of human tissues to dynamic trajectory inference, establishing a quantitative framework for trajectory-informed, state-guided, and temporally adaptive immunotherapy strategies.

cancer biology↗

bioRxiv2026

Global mRNA 3'UTR lengthening in small-cell neuroendocrine carcinoma

Small-cell neuroendocrine carcinoma (SCNC) is a rare but highly malignant tumor subtype that primarily arises in the lung, also rarely in other organs, and as a consequence of treatment induced lineage transdifferentiation of prostate adenocarcinomas. The molecular convergence of SCNC across diverse tissues enables its identification through conserved SCNC-specific molecular markers, facilitating tumor subtype classification. As a critical post-transcriptional regulatory mechanism, alternative polyadenylation (APA) modulates 3'UTR length and significantly impacts tumor progression. However, its role in SCNC remains largely unclear. Here, we report a global 3'UTR lengthening pattern driven by APA in SCNC. We identified a set of conserved 3'UTR lengthening events across SCNCs of different tissue origins, which are strongly associated with neural development and related signaling pathways. Furthermore, we developed a neural network-based prediction model to classify SCNC by leveraging these specific APA signatures. Our study provides new insights into the post-transcriptional landscape of SCNCs and highlights APA signatures as promising biomarkers for SCNC identification.

cancer biology↗

bioRxiv2026

miniMTI: minimal multiplex tissue imaging enhances biomarker expression prediction from histology

Virtual multiplexing from routine histology has advanced rapidly, yet morphology alone provides limited access to molecular state, imposing an intrinsic ceiling on H&E-only inference. Here, we introduce miniMTI, a molecularly anchored virtual staining framework that determines the minimal set of experimentally measured markers required, alongside H&E, to accurately reconstruct large multiplex tissue imaging (MTI) panels while preserving biologically and clinically relevant information. miniMTI learns from paired same-section H&E-MTI data using a unified multimodal generative model that can condition on arbitrary combinations of measured marker channels, coupled with an iterative panel selection strategy to rank informative molecular anchors. Across colorectal and prostate cancer cohorts spanning two MTI platforms and over 40 million cells, miniMTI reduces a 40-marker MTI assay to H&E plus as few as three measured molecular markers, while accurately recovering withheld markers, preserving cellular phenotypes and spatial tissue architecture, and disease-associated molecular programs, including Gleason grade-linked signatures. By integrating histology context with sparse molecular grounding, miniMTI overcomes the limitations of morphology-only virtual staining and provides a scalable, cost-effective approach for expanding MTI-level biomarker coverage with retained biological interpretability and clinical relevance.

bioinformatics↗

bioRxiv2025

A window trial in metastatic pancreatic ductal adenocarcinoma reveals resistance mechanisms to targeting the KRAS-MEK pathway

Copy number alterations of KRAS, mutated in over 90% of pancreatic ductal adenocarcinomas (PDAC), and MYC occur in 30-40% of PDAC. Here we demonstrate that KRAS and MYC are frequently co-gained and accompanied with worse prognosis in PDAC. In a Window-of-Opportunity clinical trial for metastatic PDAC, serial biopsies and deep multi-omics analyses were utilized to explore resistance mechanisms to MEK inhibition, as a surrogate for KRAS inhibition. Tumors from four of 14 patients showed Ki-67/CA19-9-based biomarker response (BR). Non-BR tumors were enriched for KRAS/MYC co-gain and KRASG12D variant. A transcriptomic signature of BR tumors was inversely correlated with KRASG12D/MYC co-gain in a large PDAC dataset and predictive for KRAS inhibitor response in multiple models. Finally, co-targeting KRAS and MYC was synergistic in KRASG12D/MYC co-gain PDAC. Together, this study provides insight into KRAS inhibitor resistance and supports MYC as an important target to improve patient outcomes in this deadly disease.

cancer biology↗

bioRxiv2025

Spatial organization of the tumor-immune microenvironment in ER-positive breast cancer: remodeling during treatment and associations with clinical response

BackgroundThe tumor microenvironment influences treatment response in ER-positive breast cancer, but what distinguishes responders from non-responders and how it changes during treatment is poorly understood. MethodsER-positive breast tumors treated with neoadjuvant chemotherapy with or without bevacizumab were profiled with bulk proteomics pre-(n = 95), on-(n = 84) and post-treatment (n = 100). A subset of tumors was profiled with spatial single-cell proteomics pre-(n = 13) and on-treatment (n = 11). Cell phenotypes, spatial location and activation states were determined, and cellular colocalization assessed with spatial metrics. Bulk and spatial features were evaluated against treatment response defined by residual cancer burden. ResultsTreatment with bevacizumab amplified chemotherapy effects on proteomic signaling. The immune contexture shifted from suppressive to supportive during treatment through decreased macrophage, regulatory and anergic T-cell density and increased colocalization between epithelial cells and CD8+, CD4+ T-cells and dendritic cells. At baseline, responders had high density of effector memory T-cells, while non-responders had more naive T-cells. In addition, responders had increased colocalization of epithelial cells with macrophages, and effector memory T-cells with M1-like macrophages compared to non-responders. ConclusionsSpatially distinct tumor-immune microenvironments influence response to neoadjuvant treatment, offering valuable insights for guiding treatment decisions.

cancer biology↗

bioRxiv2025

Language of Stains: Tokenization Enhances Multiplex Immunofluorescence and Histology Image Synthesis

Multiplex tissue imaging (MTI) is a powerful tool in cancer research, allowing spatially resolved, single-cell phenotype analysis. However, MTI platforms face challenges such as high costs, tissue loss, lengthy acquisition times, and complex analysis of large, multichannel images with batch effects. To address these challenges, we propose a novel computational method to model the interactions between dozens of panel markers and Hematoxylin & Eosin (H&E) staining, enabling in-silico generation of marker stains. This approach reduces the reliance on experimentally measured markers, bridging low-cost H&E data with MTIs high-content information. Our approach uses a two-stage frame-work for channel-wise bioimage synthesis: first, vector quantization learns a visual token vocabulary, then a bidirectional transformer infers missing markers through masked language modeling. Comprehensive bench-marking across different MTI platforms and tissue types demonstrates the effectiveness of our method in improving marker prediction while maintaining biological relevance. This advance makes high-dimensional multiplex tissue imaging more accessible and scalable, supporting deeper insights and potential clinical applications in cancer research.

bioinformatics↗

bioRxiv2025

Automated Machine Learning Profiling with MAP-HR for Quantifying Homologous Recombination Foci in Patient Samples

Accurate visualization and quantification of homologous recombination (HR)-associated foci in readily available patient samples are critical for identifying patients with HR deficiency (HRD) when they present for care to guide polyADP ribose polymerase (PARP) inhibitors (PARPi) or platinum-based therapies. Immunofluorescence (IF) assays have the potential to accurately visualize DNA repair processes as punctate foci within the nucleus. To ensure precise HRD assessment, we developed MAP-HR, (Machine-learning Assisted Profiling of Homologous Recombination), a scalable machine-learning (ML) analysis platform to enable effective patient triage and therapeutic decision-making. This workflow integrates high-resolution four-channel IF imaging and automated analysis of Geminin (cell cycle states), RAD51 (HR repair),{gamma} H2AX foci (double strand breaks) and DAPI (nuclear localization) in both cultured cell lines and in a single formalin-fixed, paraffin-embedded (FFPE) patient sample. Using a spinning disk confocal microscope, we optimized imaging parameters to improve resolution and signal-to-noise ratio. Our MAP-HR pipeline uses nested nuclei and segmentation of foci to analyze the HR status of each cell, unlike competing bulk or single-foci marker assays, allowing evaluation of HR functional heterogeneity across and within patient biopsies. This approach facilitates robust comparisons of HR and foci-based processes across diverse cell populations and patient tissues, enabling scalable, translational research. GRAPHICAL ABSTRACT O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=144 SRC="FIGDIR/small/639329v1_ufig1.gif" ALT="Figure 1"> View larger version (33K): org.highwire.dtl.DTLVardef@154d43eorg.highwire.dtl.DTLVardef@3a6fa6org.highwire.dtl.DTLVardef@782daborg.highwire.dtl.DTLVardef@1fe6c28_HPS_FORMAT_FIGEXP M_FIG C_FIG

cancer biology↗

bioRxiv2025

A spatially resolved single cell proteomic atlas of Small Bowel Adenocarcinoma

Small bowel adenocarcinoma (SBA) is a rare malignancy marked with a poor prognosis. The cellular and proteomic heterogeneity within the tumor immune microenvironment (TIME) of SBA is a likely driver of prognosis, disease progression and response to therapy. We have addressed a major gap in knowledge of the TIME in SBA using highly multiplexed, protein imaging of the SBA tumor-immune ecosystem generating a comprehensive, single-cell level, spatial, proteomic atlas of TIME in > 600,000 cells from 136 tumor and matched normal samples from clinically and genomically annotated SBA patients (N=37). The SBA TIME Atlas informs on spatial distribution and interactions of tumor-intrinsic processes, diverse immune cell types, immune checkpoints, and vascularization. Enrichment of proliferating epithelial tissue, stem cells, and likely pro-tumor immune signatures in the tumor niche is contrasted by the representation of naive and early-effector T-cells in the epithelial compartments of adjacent normal tissues. Epithelial-stem-immune cell spatial architectures within tumor and matched-normal niches were strongly associated with patient survival, suggesting malignancy is driven by spatial architecture beyond the tumor microenvironment. The blueprints of therapeutically actionable immune checkpoints at the interface between epithelial and microenvironmental T-cells as well as macrophages have established a guideline for precision immunotherapies tailored to the TIME composition in SBA. We expect that this SBA atlas will contribute to a deeper understanding of the immune contexture in this rare disease as well as other gastrointestinal cancers and help guide future precision immune-oncology strategies.

cancer biology↗

bioRxiv2025

Neoplastic immune mimicry is a generalizable phenomenon in breast cancer and epithelial CD69 enables early tumor progression

Dedifferentiation programs are commonly enacted during breast cancer progression to enhance tumor cell fitness. Increased cellular plasticity within the neoplastic compartment of tumors correlates with disease aggressiveness, often culminating in greater resistance to cytotoxic therapies or augmented metastatic potential. Here we report that subpopulations of dedifferentiated neoplastic breast epithelial cells express canonical leukocyte cell surface receptor proteins and have thus named this cellular program "immune mimicry." We document neoplastic cells engaging in immune mimicry within public human breast tumor single-cell RNA-seq datasets, histopathological breast tumor specimens, breast cancer cell lines, as well as in murine transgenic and cell line-derived mammary cancer models. Immune-mimicked neoplastic cells harbor hallmarks of dedifferentiation and are enriched in treatment-resistant and high-grade breast tumors. We corroborated these observations in aggressive breast cancer cell lines where anti-proliferative cytotoxic chemotherapies drove epithelial cells toward immune mimicry. Moreover, in subsequent proof-of-concept studies, we demonstrate that expression of the CD69 leukocyte activation protein by neoplastic cells confers a proliferative advantage that facilitates early tumor growth and therefore conclude that neoplastic breast epithelial cells upregulating leukocyte surface receptors potentiate malignancy. Moving forward, neoplastic immune mimicry should be evaluated for prognostic utility in additional breast cancer cohorts to determine its potential for patient stratification. Future research should evaluate correlates with distal metastases, progression-free survival, overall survival, and therapeutic response/resistance. Statement of SignificanceNeoplastic breast epithelial cells express surface receptors canonically attributed to leukocytes and are associated with therapy resistance and aggressive tumor behavior.

cancer biology↗

bioRxiv2024

Deciphering single-cell heterogeneity and cellular ecosystem dynamics during prostate cancer progression

Prostate cancer (PC) progresses from benign epithelium through pre-malignant lesions, localized tumors, metastatic dissemination, and castration-resistant stages, with some cases exhibiting phenotype plasticity under therapeutic pressure. However, high-resolution insights into how cell phenotypes evolve across successive stages of PC remain limited. Here, we present the Prostate Cancer Cell Atlas (PCCAT) by integrating [~]710,000 single cells from 197 human samples covering a spectrum of tumor stages. This comprehensive analysis dissects the cellular landscape and characterizes key cell types and molecular features that associate with PC progression and prognosis. In malignant cells, we highlight a distinctive profile denoted by high Major Histocompatibility Complex (MHC) expression, low Androgen Receptor (AR) activity, and enhanced stemness programs associated with enzalutamide resistance. Moreover, we reveal several cell states strongly correlated with PC progression and adverse prognosis, including lineage plasticity-like malignant cells (LPCs), neuroendocrine tumor cells, pericytes, and matrix cancer-associated fibroblasts (mCAFs). Furthermore, we uncover shared cell states that underpin the immune suppressive tumor microenvironment in advanced PC, including activated regulatory T cells, exhausted CD8+ T cells, and SPP1-expressing macrophages. Lastly, we pinpoint a spatial niche composed of mCAFs and SPP1-expressing macrophages localized near the tumor boundary in aggressive PC, which correlates with poor prognosis. Overall, our work provides a valuable resource and offers deeper insights into the diverse cell states, dynamics, and functional characteristics involved in PC progression at single-cell resolution.

cancer biology↗

bioRxiv2024

UniFORM: Towards Universal Immunofluorescence Normalization for Multiplex Tissue Imaging

Multiplexed tissue imaging (MTI) technologies enable high-dimensional spatial analysis of tumor microenvironments but face challenges with technical variability in staining intensities. Existing normalization methods, including Z-score, ComBat, and MxNorm, often fail to account for the heterogeneous, right-skewed expression patterns of MTI data, compromising signal alignment and downstream analyses. We present UniFORM, a non-parametric, Python-based pipeline that uses an automated rigid landmark functional data registration approach for normalizing both feature- and pixel-level MTI data. Designed specifically for the distributional characteristics of MTI datasets, UniFORM operates without prior distributional assumptions and performs robustly regardless of distribution modality, including both unimodal and bimodal patterns. It removes technical variation by aligning the biologically invariant component of the signal, typically the negative (non-expressing) population, while preserving biologically meaningful variation in the positive population, thereby maintaining tissue-specific expression patterns essential for downstream analysis. Benchmarking across three distinct MTI platform datasets demonstrates that UniFORM outperforms existing methods in mitigating batch effects while maintaining biological signal fidelity. This is evidenced by improved marker distribution alignment and positive population preservation, enhanced kBET and Silhouette scores, and improved downstream analyses such as UMAP visualizations and Leiden clustering. UniFORM also introduces a novel guided fine-tuning option for complex and heterogeneous datasets. Although optimized for fluorescence-based platforms, UniFORM provides a scalable and robust solution for MTI data normalization, enabling accurate and biologically meaningful interpretations.

bioinformatics↗

bioRxiv2024

A Comprehensive Proteogenomic and Spatial Analysis of Innate and Acquired Resistance of Metastatic Melanoma to Immune Checkpoint Blockade Therapies

While a subset of patients with metastatic melanoma achieves durable responses to immune checkpoint blockade (ICB) therapies, the majority ultimately exhibit either innate or acquired resistance to these treatments. However, the molecular mechanisms underlying resistance to ICB therapies remain elusive and are warranted to elucidate. Here, we comprehensively investigated the tumor and tumor immune microenvironment (TIME) of paired pre- and post-treatment tumor specimens from metastatic melanoma patients who were primary or secondary resistance to anti-CTLA-4 and/or anti-PD-1/PD-L1 therapies. Differentially expressed gene (DEG) analysis and single-sample gene set enrichment analysis (ssGSEA) with transcriptomic data identified cell cycle and c-MYC signaling as pathway-based resistance signatures. And weighted gene co-expression network analysis (WGCNA) revealed the activation of a cross-resistance meta-program involving key signaling pathways related to tumor progression in ICB resistant melanoma. Moreover, spatially-resolved, image-based immune monitoring analysis by using NanoStrings digital spatial profiling (DSP) and Cyclic Immunofluorescence (CyCIF) showed infiltration of suppressive immune cells in the tumor microenvironment of melanoma with resistance to ICB therapies. Our study reveals the molecular mechanisms underlying resistance to ICB therapies in patients with metastatic melanoma by conducting such integrated analyses of multi-dimensional data, and provides rationale for salvage therapies that will potentially overcome resistance to ICB therapies. Statement of translational relevanceThis study paves the way for the creation of innovative therapeutic strategies, aimed at subverting resistance to immune checkpoint blockade (ICB) therapies in metastatic melanoma patients. By unraveling the specific molecular mechanisms underlying resistance, scientists can design effective alternative treatments that target pathways such as pathways associated with cell cycle dysregulation and c-MYC signaling. Furthermore, through the application of advanced immune monitoring techniques such as NanoString Digital Spatial Profiling (DSP) and Cyclic Immunofluorescence (CyCIF), this study has significantly enriched our understanding of the tumor microenvironment. This enhanced characterization facilitates the discovery of potential biomarkers that may forecast a patients response to ICB treatment. Ultimately, these advancements could potentially refine patient outcomes and foster the development of more personalized cancer treatments in the future.

genetics↗

bioRxiv2024

COEXIST: Coordinated single-cell integration of serial multiplexed tissue images

Multiplexed tissue imaging (MTI) and other spatial profiling technologies commonly utilize serial tissue sectioning to comprehensively profile samples by imaging each section with unique biomarker panels or assays. The dependence on serial sections is attributed to technological limitations of MTI panel size or incompatible multi-assay protocols. Although image registration can align serially sectioned MTIs, integration at the single-cell level poses a challenge due to inherent biological heterogeneity. Existing computational methods overlook both cell population heterogeneity across modalities and spatial information, which are critical for effectively completing this task. To address this problem, we first use Monte-Carlo simulations to estimate the overlap between serial 5m-thick sections. We then introduce COEXIST, a novel algorithm that synergistically combines shared molecular profiles with spatial information to seamlessly integrate serial sections at the single-cell level. We demonstrate COEXIST necessity and performance across several applications. These include combining MTI panels for improved spatial single-cell profiling, rectification of miscalled cell phenotypes using a single MTI panel, and the comparison of MTI platforms at single-cell resolution. COEXIST not only elevates MTI platform validation but also overcomes the constraints of MTIs panel size and the limitation of full nuclei on a single slide, capturing more intact nuclei in consecutive sections and thus enabling deeper profiling of cell lineages and functional states.

bioinformatics↗

bioRxiv2024

Graph Structured Neural Networks for Perturbation Biology

1Computational modeling of perturbation biology identifies relationships between molecular elements and cellular response, and an accurate understanding of these systems will support the full realization of precision medicine. Traditional deep learning, while often accurate in predicting response, is unlikely to capture the true sequence of involved molecular interactions. Our work is motivated by two assumptions: 1) Methods that encourage mechanistic prediction logic are likely to be more trustworthy, and 2) problem-specific algorithms are likely to outperform generic algorithms. We present an alternative to Graph Neural Networks (GNNs) termed Graph Structured Neural Networks (GSNN), which uses cell signaling knowledge, encoded as a graph data structure, to add inductive biases to deep learning. We apply our method to perturbation biology using the LINCS L1000 dataset and literature-curated molecular interactions. We demonstrate that GSNNs outperform baseline algorithms in several prediction tasks, including 1) perturbed expression, 2) cell viability of drug combinations, and 3) disease-specific drug prioritization. We also present a method called GSNNExplainer to explain GSNN predictions in a biologically interpretable form. This work has broad application in basic biological research and pre-clincal drug repurposing. Further refinement of these methods may produce trustworthy models of drug response suitable for use as clinical decision aids. Availability and implementationOur implementation of the GSNN method is available at https://github.com/nathanieljevans/GSNN. All data used in this work is publicly available.

bioinformatics↗

bioRxiv2024

Transcriptional and phenotypic heterogeneity underpinning venetoclax resistance in AML

The venetoclax BCL2 inhibitor in combination with hypomethylating agents represents a cornerstone of induction therapy for older AML patients, unfit for intensive chemotherapy. Like other targeted therapies, venetoclax-based therapies suffer from innate and acquired resistance. While several mechanisms of resistance have been identified, the heterogeneity of resistance mechanism across patient populations is poorly understood. Here we utilized integrative analysis of transcriptomic and ex-vivo drug response data in AML patients to identify four transcriptionally distinct VEN resistant clusters (VR_C1-4), with distinct phenotypic, genetic and drug response patterns. VR_C1 was characterized by enrichment for differentiated monocytic- and cDC-like blasts, transcriptional activation of PI3K-AKT-mTOR signaling axis, and energy metabolism pathways. They showed sensitivity to mTOR and CDK inhibition. VR_C2 was enriched for NRAS mutations and associated with distinctive transcriptional suppression of HOX expression. VR_C3 was characterized by enrichment for TP53 mutations and higher infiltration by cytotoxic T cells. This cluster showed transcriptional expression of erythroid markers, suggesting tumor cells mimicking erythroid differentiation, activation of JAK-STAT signaling, and sensitivity to JAK inhibition, which in a subset of cases synergized with venetoclax. VR_C4 shared transcriptional similarities with venetoclax-sensitive patients, with modest over-expression of interferon signaling. They were also characterized by high rates of DNMT3A mutations. Finally, we projected venetoclax-resistance states onto single cells profiled from a patient who relapsed under venetoclax therapy capturing multiple resistance states in the tumor and shifts in their abundance under venetoclax selection, suggesting that single tumors may consist of cells mimicking multiple VR_Cs contributing to intra-tumor heterogeneity. Taken together, our results provide a strategy to evaluate inter- and intra-tumor heterogeneity of venetoclax resistance mechanisms and provide insights into approaches to navigate further management of patients who failed therapy with BCL2 inhibitors.

cancer biology↗