Protein-protein interactions (PPIs) involving oncogenic drivers remain among the most intractable targets in cancer biology due to their dynamic conformations and limited accessibility to conventional small molecules. Although antibodies and indirect inhibitors have achieved clinical success against targets such as PD-1/PD-L1 and MYC, challenges persist related to tissue penetration, intracellular delivery, resistance, and incomplete blockade of key interface hotspots. Here, we present DesignForge, an integrated de novo protein design framework that combines deep-learning-based structure generation, sequence optimization, and energetic hotspot mapping to create compact miniprotein binders for PPIs. Using this approach, we engineered PD-1 mimetics predicted to disrupt the PD-1/PD-L1 immune checkpoint, designed scaffolds targeting the MYC/MAX dimerization interface, and generated KRAS binders in a manner predicted to occlude RAF interaction. The top designs showed high structural confidence by AlphaFold2, favorable stability metrics, and consistent hotspot engagement identified through MOE-based analyses. Collectively, these results establish DesignForge as a generalizable in silico platform for rational design of therapeutic protein binders that extend beyond antibody and small-molecule modalities to systematically target intractable oncogenic PPIs.

Quantitative image analysis is essential for advancing stem cell biology, developmental studies, and drug discovery, yet most workflows still rely on manual or semi-quantitative scoring that is slow, subjective, and poorly scalable. A major challenge is converting complex colony morphologies into reproducible, high-dimensional datasets. To address this gap, we developed ColonyQuant, an open-source platform that integrates automated colony segmentation, alkaline phosphatase (AP) intensity quantification, morphometric profiling, and statistical classification into a single workflow. ColonyQuant computes per-colony functional readouts alongside comprehensive shape descriptors, capturing both staining intensity and structural features in a unified framework. Applied to embryonic stem cells (ESCs) treated with a selective KDM4 histone-demethylase inhibitor, ColonyQuant revealed dose-dependent reductions in colony area and integrated AP signal, together with systematic remodeling of morphometric metrics. Multivariate analyses robustly stratified treatment groups and identified intensity and solidity as principal features capturing dose-dependent colony responses. By transforming subjective scoring into objective, scalable, and biologically interpretable phenotyping, ColonyQuant provides a reproducible platform for stem cell research and high-content screening.

Mouse embryonic stem cells (ESCs) occasionally transit into a rare two-cell-like (2C) state characterized by transient activation of endogenous retroviruses such as MERVL and expression of 2C-specific genes including the Zscan4 cluster. These 2C-like cells resemble early blastomeres and display expanded developmental potential, but their unstable and sporadic nature has hindered mechanistic studies. Here, we demonstrate the transiently stable maintenance of MERVLpositive ESCs that exhibit persistent MERVL expression and activation of 2C-associated genes. Live-cell imaging revealed uniform and sustained MERVL activity in these MERVLpositive ESCs, contrasting with the heterogeneous and transient expression observed in conventional ESCs. Transcriptome profiling demonstrated robust induction of 2C-specific regulatory networks, and embryoid body differentiation combined with machine learning uncovered increased lineage variability and altered developmental trajectories. Single-cell RNA sequencing revealed clear separation of control ESCs from MERVL-positive populations and redistribution across distinct transcriptional states, with Red and Mosaic lines showing graded shifts within a shared transcriptional manifold. Epigenomic profiling further revealed distinct chromatin states, specialized super-enhancer landscapes, and active enhancer marking at MERVL loci. Together, these findings demonstrate that stable maintenance of MERVLpositive ESCs is achievable in vitro, providing a powerful model to dissect endogenous retroviral element-driven transcriptional regulation, epigenomic remodeling, and 2C-like transcriptional and epigenetic programs.

Background/Objectives: Genetic generalized epilepsies (GGE) often remit in childhood, yet a subset of adults remain pharmacoresistant with substantial morbidity. The genetic basis of adult pharmacoresistant GGE is poorly defined. This descriptive study used whole-genome sequencing (WGS) to identify recurrent coding variants and pathways associated with pharmacoresistant adult GGE. Methods: WGS was performed in ten racially diverse adults (mean age 37.2 years; range 20-52) with electroencephalographically confirmed, pharmacoresistant GGE (mean onset 13.7 years). Analysis prioritized variants present in at least 80% of participants and which were either (i) missense variants predicted deleterious with ANNOVAR or (ii) loss-of-function variants predicted high-impact from snpEff. Pathway enrichment and overlap with a commercial clinical epilepsy gene panel were assessed. Results: Filtering identified 133 unique, deleterious coding variants across 69 genes shared by at least eight participants. Four genes (APOL4, KMT2C, SON, VDR) overlapped a clinical epilepsy panel, supporting the capacity of WGS to recover clinically relevant loci. Prioritized loci implicated gastrointestinal and metabolic regulators (e.g., MUC6, PNLIPRP2), chemosensory receptors (OR10D3, OR8U1, TAS2R19), neuroimmune mediators (LILRA2, SIGLEC12, OAS2), and ion transporters (KCNJ12, P2RX5, RHBG), consistent with multifactorial mechanisms of pharmacoresistance. Conclusions: This exploratory WGS study focused exclusively on adults with pharmacoresistant GGE, revealing shared high-impact variants and convergent pathways spanning absorption/metabolism, vitamin D signaling, immunity, and ion transport. Findings broaden the genetic landscape of pharmacoresistant GGE while motivating validation in larger, multiethnic cohorts.

Quantitative image analysis is essential for advancing stem cell biology, developmental studies, and drug discovery, yet most workflows still rely on manual or semi-quantitative scoring that is slow, subjective, and poorly scalable. A major challenge is converting complex colony morphologies into reproducible, high-dimensional datasets. To address this gap, we developed ColonyQuant, an open-source platform that integrates automated colony segmentation, alkaline phosphatase (AP) intensity quantification, morphometric profiling, and statistical classification into a single workflow. ColonyQuant computes per-colony functional readouts alongside comprehensive shape descriptors, capturing both staining intensity and structural features in a unified framework. Applied to embryonic stem cells (ESCs) treated with a selective KDM4 histone-demethylase inhibitor, ColonyQuant revealed dose-dependent reductions in colony area and integrated AP signal, together with systematic remodeling of morphometric metrics. Multivariate analyses robustly stratified treatment groups and identified intensity and solidity as principal features capturing dose-dependent colony responses. By transforming subjective scoring into objective, scalable, and biologically interpretable phenotyping, ColonyQuant provides a reproducible platform for stem cell research and high-content screening.

Mouse embryonic stem cells (ESCs) occasionally transit into a rare two-cell-like (2C) state characterized by transient activation of endogenous retroviruses such as MERVL and expression of 2C-specific genes including the Zscan4 cluster. These 2C-like cells (2CLCs) resemble early blastomeres and display expanded developmental potential, but their unstable and sporadic nature has hindered mechanistic studies. Here, we establish stable 2CLCs (s2CLCs) that maintain persistent MERVL expression and homogeneous 2C gene activation. Live-cell imaging revealed uniform and sustained MERVL activity in s2CLCs, contrasting with the heterogeneous and transient expression observed in conventional ESCs. Transcriptome profiling demonstrated robust induction of 2C-specific regulatory networks, and embryoid body differentiation combined with machine learning uncovered increased lineage variability and expanded developmental trajectories. Epigenomic profiling further revealed unique chromatin states, distinctive super enhancer landscapes, and active enhancer marking at MERVL loci. Together, these findings demonstrate that stable maintenance of the 2C-like state is achievable in vitro, providing a powerful model to dissect ERV-driven transcriptional regulation, epigenomic remodeling, and totipotent-like developmental potential.

Abstract Our goal is to identify, in situ, the regulatory genes that dictate the onset age and growth rate of spontaneously arising tumors to implement new intervention strategies. For genetic linkage analysis, we crossed HER2(neu) Tg mice with Diversity Outbred (DO) mice, an outbred population comprised of 8 founding strains (A/J, C57BL/6J, 129S1/SvlmJ, NOD/ShiLtJ, NZO/HILtJ, CAST/EiJ, PWK/PhJ, WSB/EiJ) maintained by non-sibling crossing to ensure each mouse is genetically unique. Using R/QTL package, we linked Quantitative Trait Loci (QTL) in Chr 1 and X with tumor onset age, and a Chr 10 QTL with tumor growth rate. The Chr1 (=human Chr2) QTL was linked to human breast cancer diagnosis in 11 Genome-Wide Association Studies (GWAS; the human GWAS catalog, https://www.ebi.ac.uk/gwas/). Because human GWAS data is deficient in Chr X mapping and also does not assess tumor growth rate as a trait, the QTL loci identified in mouse Chr X and 10 were discoveries beyond the capacity of human GWAS. In total, we identified 26 candidate genes across the 3 QTL. For clinical validation, the genes were analyzed with CodeAI (Caris LS) which associated cancer patient clinical data with candidate gene expression. We found 21/26 genes significantly associated with the survival of patients with primary (n=3,533) and/or metastatic (n=4,870) breast cancers and 17/26 were associated with lung cancer (n=11,334) survival, with 13 overlapping genes between lung and breast cancer, indicating broad applicability of these candidate genes. We interrogated the immune status in BALB NeuT and (BALBxPWK)F1 NeuT mice because the F1 mice developed tumors around 9 wk, much earlier than the 14 wk observed in BALB NeuT mice. Surprisingly, (BALBxPWK)F1 mice responded to cancer vaccines more vigorously. Single cell RNA (scRNA) libraries from (BALBxPWK)F1 NeuT tumors showed an expanded lymphocyte cluster (12%), compared with BALB NeuT (2%) and an enrichment of T cell activation genes. These findings may reflect a deficiency in tumor cell recognition by the activated anti-tumor effectors and a window of opportunity for correction. Of the 21 candidate genes, LILRB4, a myeloid cell checkpoint molecule, was of particular interest. Expression of LILRB4 is predictive of patient survival in both breast and lung cancer. Using scRNA sequencing, we found this transmembrane receptor expressed primarily by tumor infiltrating myeloid cells. In a reported scRNA sequencing dataset (GSE161529), LILRB4 was expressed by infiltrating macrophages in human breast cancers, and is also correlated with reduced survival. In a PAN Cancer ATACseq dataset (Corces et al., 2018), promoter accessibility of LILRB4 was associated with reduced survival. Taken together, there are multiple immune regulatory mechanisms of tumor progression and LILRB4 is an important new actionable target. CA76340 (WZW), KCI TBM program (JBJ) and CCSG P30 CA022453 (GB). Citation Format: Jennifer B. Jacob, Wei-Zen Wei, Benjamin L. Kidder, Tolulope Adeyelu, Andrew Elliott, Gerold Bepler, Joyce D. Reyes. Identification of the actionable target, LILRB4, through genetic linkage analysis of Diversity Outbred (DO) F1 mice expressing HER2(neu) [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 114.

Introduction: Cultured mouse trophoblast stem cells (mTSC) maintain proliferation/normal stemness (NS) under FGF4, which when removed, causes normal differentiation (ND). Hypoxic, or hyperosmotic stress forces trophoblast giant cells (TGC) differentiate. Hypoxic, hyperosmotic, and genotoxic benzo(a)pyrene (BaP), which is found in tobacco smoke, force down-regulation of inhibitor of differentiation (Id)2, enabling TGC differentiation. Hypoxic and hyperosmotic stress induce TGC by SAPK-dependent HAND1 increase. Here we test whether BaP forces mTSC-to-TGC while inducing SAPK and HAND1. Methods: Hand1 and SAPK activity were assayed by immunoblot, mTSC-to-TGC growth and differentiation were assayed at Tfinal after 72hr exposure of BaP, NS, ND, Retinoic acid (RA), or sorbitol. Nuclear-stained cells were micrographed automatically by a live imager, and assayed by ImageJ/FIJI, Biotek Gen 5, AIVIA proprietary artificial intelligence (AI) software or open source, CellPose artificial intelligence/AI software. Results: BaP (0.05-1 mu M) activated SAPK and HAND1 without diminishing growth. TSC-to-TGC differentiation was assayed with increasingly accuracy for 2-4 N cycling nuclei and >4 N differentiating TGC nuclei, using ImageJ/FIJI, Gen 5, AIVIA, or CellPose AI software. The AIVIA and Cellpose AI software matches human accuracy. The lowest BaP effects on SAPK activation/HAND1 increase are >10-fold more sensitive than similar effects for mESC. RA induces 44-47% 1st lineage TGC differentiation, but the same RA dose induces only 1% 1st lineage mESC differentiation. Discussion: First, these pilot data suggest that mTSC can be used in high throughput screens (HTS) to predict toxicant exposures that force TGC differentiation. Second, mTSC differentiated more cells than mESC for similar stress exposures, Third, open source AI can replace human micrograph quantitation and enable a miscarriage-predicting HTS.

Teratoma formation is key for evaluating differentiation of human pluripotent stem cells into embryonic germ layers and serves as a model for understanding stem cell differentiation and developmental processes. Its potential for insights into epigenome and transcriptome profiling is significant. This study integrates the analysis of the epigenome and transcriptome of hESC-generated teratomas, comparing transcriptomes between hESCs and teratomas. It employs cell type-specific expression patterns from single-cell data to deconvolve RNA-Seq data and identify cell types within teratomas. Our results provide a catalog of activating and repressive histone modifications, while also elucidating distinctive features of chromatin states. Construction of an epigenetic signature matrix enabled the quantification of diverse cell populations in teratomas and enhanced the ability to unravel the epigenetic landscape in heterogeneous tissue contexts. This study also includes a single cell multiome atlas of expression (scRNA-Seq) and chromatin accessibility (scATAC-Seq) of human teratomas, further revealing the complexity of these tissues. A histology-based digital staining tool further complemented the annotation of cell types in teratomas, enhancing our understanding of their cellular composition. This research is a valuable resource for examining teratoma epigenomic and transcriptomic landscapes and serves as a model for epigenetic data comparison.

ABSTRACT Deep neural networks have significantly advanced medical image analysis, yet their full potential is often limited by the relatively small dataset sizes. Generative modeling has stimulated attention for its potential applications in the synthesis of medical images. Recent advancements in diffusion models have exhibited a remarkable capacity for producing photorealistic images. Despite this promising development, the application of such models in the generation of medical images remains underexplored. In this study, we explored the potential of using diffusion models to generate medical images, with a particular emphasis on producing brain magnetic resonance imaging (MRI) scans, such as those depicting low-grade gliomas. Additionally, we examined the generation of contrast enhanced spectral mammography (CESM) images, as well as chest and lung X-ray images. Utilizing the Dreambooth platform, we trained stable diffusion models based on text prompts, class and instance images, subsequently prompting the trained models to produce medical images. The generation of medical imaging data presents a viable approach for preserving the anonymity of medical images, effectively reducing the likelihood of patient re-identification during the exchange of data for research. The findings of this study reveal that the application of diffusion models in generating images successfully captures attributes specific to oncology within imaging modalities. Consequently, this research establishes a framework that harnesses the power of artificial intelligence for the generation of cancer medical imagery.

HER2-targeted therapy has improved breast cancer survival, but treatment resistance and disease prevention remain major challenges. Genes that enable HER2/Neu oncogenesis are the next intervention targets. A bioinformatics discovery platform of HER2/Neu-expressing Diversity Outbred (DO) F1 Mice was established to identify cancer-enabling genes. Quantitative Trait Loci (QTL) associated with onset ages and growth rates of spontaneous mammary tumors were sought. Twenty-six genes in 3 QTL contain sequence variations unique to the genetic backgrounds that are linked to aggressive tumors and 21 genes are associated with human breast cancer survival. Concurrent identification of TSC22D3, a transcription factor, and its target gene LILRB4, a myeloid cell checkpoint receptor, suggests an immune axis for regulation, or intervention, of disease. We also investigated TIEG1 gene that impedes tumor immunity but suppresses tumor growth. Although not an actionable target, TIEG1 study revealed genetic regulation of tumor progression, forming the basis of the genetics-based discovery platform.

Identifying cancer type-specific genes that define cell states is important to develop effective therapies for patients and methods for detection, early diagnosis, and prevention. While molecular mechanisms that drive malignancy have been identified for various cancers, the identification of cell-type defining transcription factors (TFs) that distinguish normal cells from cancer cells has not been fully elucidated. Here, we utilized a network biology framework, which assesses the fidelity of cell fate conversions, to identify cancer type-specific gene regulatory networks (GRN) for 17 types of cancer. Through an integrative analysis of a compendium of expression data, we elucidated core TFs and GRNs for multiple cancer types. Moreover, by comparing normal tissues and cells to cancer type-specific GRNs, we found that the expression of key network-influencing TFs can be utilized as a survival prognostic indicator for a diverse cohort of cancer patients. These findings offer a valuable resource for exploring cancer type-specific networks across a broad range of cancer types.

Table S1. Primer sequences for Q-RT-PCR Table S2. SKY Analysis of shLuc ES cells, shSmyd5 ES cells, and shSmyd5 cancer cells.

Figure S1. Altered differentiation and formation of transformed cells during EB formation of SMYD5 depleted ES cells Figure S2. LTR and LINE repeats are enriched at CNA regions in shSmyd5 cancer cells Figure S3. Expression of G9a in SMYD5-depleted ES cells and embryoid bodies. Figure S4. Expression of ERVK and L1 repeat subfamilies in shSmyd5 cancer cells Figure S5. Depletion of SMYD5 in human ES cells leads to transformation during differentiation Figure S6. SMYD5-depleted MCF10A cells exhibit altered 3D growth characteristics Figure S7. Depletion of SMYD5 leads to altered gene expression of breast cancer cells

Aberrant expression of the Forkhead box transcription factor, FOXQ1, is a prevalent mechanism of epithelial-mesenchymal transition (EMT) and metastasis in multiple carcinoma types. However, it remains unknown how FOXQ1 regulates gene expression. Here, we report that FOXQ1 initiates EMT by recruiting the MLL/KMT2 histone methyltransferase complex as a transcriptional coactivator. We first establish that FOXQ1 promoter recognition precedes MLL complex assembly and histone-3 lysine-4 trimethylation within the promoter regions of critical genes in the EMT program. Mechanistically, we identify that the Forkhead box in FOXQ1 functions as a transactivation domain directly binding the MLL core complex subunit RbBP5 without interrupting FOXQ1 DNA binding activity. Moreover, genetic disruption of the FOXQ1-RbBP5 interaction or pharmacologic targeting of KMT2/MLL recruitment inhibits FOXQ1-dependent gene expression, EMT, and in vivo tumor progression. Our study suggests that targeting the FOXQ1-MLL epigenetic axis could be a promising strategy to combat triple-negative breast cancer metastatic progression.

The H3K4 demethylase KDM5B is overexpressed in multiple cancer types, and elevated expression levels of KDM5B is associated with decreased survival. However, the underlying mechanistic contribution of dysregulated expression of KDM5B and H3K4 demethylation in cancer is poorly understood. Our results show that loss of KDM5B in multiple types of cancer cells leads to increased proliferation and elevated expression of cancer stem cell markers. In addition, we observed enhanced tumor formation following KDM5B depletion in a subset of representative cancer cell lines. Our findings also support a role for KDM5B in regulating epigenetic plasticity, where loss of KDM5B in cancer cells with elevated KDM5B expression leads to alterations in activity of chromatin states, which facilitate activation or repression of alternative transcriptional programs. In addition, we define KDM5B-centric epigenetic and transcriptional patterns that support cancer cell plasticity, where KDM5B depleted cancer cells exhibit altered epigenetic and transcriptional profiles resembling a more primitive cellular state. This study also provides a resource for evaluating associations between alterations in epigenetic patterning upon depletion of KDM5B and gene expression in a diverse set of cancer cells.

Epigenetic mechanisms contribute to the initiation and development of cancer, and epigenetic variation promotes dynamic gene expression patterns that facilitate tumor evolution and adaptation. While the NCI-60 panel represents a diverse set of human cancer cell lines that has been used to screen chemical compounds, a comprehensive epigenomic atlas of these cells has been lacking. Here, we report an integrative analysis of 60 human cancer epigenomes, representing a catalog of activating and repressive histone modifications. We identify genome-wide maps of canonical sharp and broad H3K4me3 domains at promoter regions of tumor suppressors, H3K27ac-marked conventional enhancers and super enhancers, and widespread inter-cancer and intra-cancer specific variability in H3K9me3 and H4K20me3-marked heterochromatin domains. Furthermore, we identify features of chromatin states, including chromatin state switching along chromosomes, correlation of histone modification density with genetic mutations, DNA methylation, enrichment of DNA binding motifs in regulatory regions, and gene activity and inactivity. These findings underscore the importance of integrating epigenomic maps with gene expression and genetic variation data to understand the molecular basis of human cancer. Our findings provide a resource for mining epigenomic maps of human cancer cells and for identifying epigenetic therapeutic targets.

Heterochromatin, which is a densely packed chromatin state that is transcriptionally silent, is a critical regulator of gene expression. However, it is unclear how the repressive histone modification, H4K20me3, or the histone methyltransferase, SUV420H2, regulate embryonic stem (ES) cell fate by patterning the epigenetic landscape. Here, we report that depletion of SUV420H2 leads to a near complete loss of H4K20me3 genome-wide, dysregulated gene expression, and delayed ES cell differentiation. SUV420H2-bound regions are enriched with repetitive DNA elements, which are de-repressed in SUV420H2 knockout ES cells. Moreover, SUV420H2 regulation of H4K20me3-marked heterochromatin controls chromatin architecture, including fine-scale chromatin interactions in pluripotent ES cells. SUV420H2 plays a critical role in stabilizing the three-dimensional (3D) chromatin landscape of ES cells, where loss of SUV420H2 results in A/B compartment switching, perturbed chromatin insulation, and altered chromatin interactions of pericentric heterochromatin and surrounding regions, indicative of localized decondensation. In addition, depletion of SUV420H2 resulted in compromised interactions between H4K20me3 and gene regulatory regions. Together, these findings describe a novel role for SUV420H2 in regulating the chromatin landscape of ES cells.