Image-based Screening Research Articles

TMOD-36. NANOGLIOS: A MINIATURIZED PATIENT AVATAR ENABLING RAPID AND HIGH-THROUGHPUT PHENOTYPIC SCREENING FOR GLIOMAS

Abstract BACKGROUND Establishing pre-clinical patient derived orthtopical xenograft models for IDH-mutant and H3K27M-mutant gliomas has presented significant histological challenges. There is increasing interest in developing faithful patient-derived ex vivo 3D models to replicate these type of gliomas, aiming to accelerate drug screening efforts. METHOD We developed a rapid approach for establishing nanoliter-sized miniature glioma organoids, termed NanoGlios, by leveraging emulsion microfluidics. We tested the robustness of NanoGlios for neurosphere culture, glioma xenografts, and primary patient samples, focusing on H3K27M-mutant diffuse midline gliomas (DMGs) and IDH-mutant gliomas. We compared the morphological features of NanoGlios to their counterparts using high-content imaging, H&E staining, and histological analysis. Additionally, we evaluated the mutation and gene expression profiles of NanoGlios relative to their counterparts. Furthermore, we conducted proof-of-concept drug screening of NanoGlios with various inhibitors and established an AI-enhanced high-throughput phenotypic imaging-based screening platform. RESULTS We successfully established NanoGlios from various sample sources, including neurospheres, xenografts, and primary tissues harboring H3K27M or IDH1/2 mutations. NanoGlios derived from xenografts or primary tissues captured more heterogeneous patterns as compared to the neuropsheres. We were able to generate thousands of NanoGlios and perform drug assays for hundreds of conditions at passage 0 within two weeks from both resected glioma tissues and xenografts. The AI phenotypic screening enabled us to identify differential responses of NanoGlios to ONC-201 treatment and drug responses to mutant IDH inhibitors Vorasidenib and Ivosidenib, which was not able to be detected by conventional cell viability bulk assays.

Arabidopsis KNS3 and its two homologs mediate endoplasmic reticulum-to-plasma membrane traffic of boric acid channels.

Membrane proteins targeted to the plasma membrane are first transported from the endoplasmic reticulum (ER) to the Golgi apparatus. This study explored the mechanisms controlling plasma membrane trafficking of the boric acid channel AtNIP5;1 from the ER. Imaging-based screening using transgenic Arabidopsis identified six mutants in which GFP-NIP5;1 was localized in the ER in addition to the plasma membrane. Genetic mapping and whole-genome resequencing identified the responsible gene in four among the six mutants as KAONASHI3 (KNS3)/SPOTTY1/IMPERFECTIVE EXINE FORMATION. Among the plasma membrane-localized proteins tested, NIP5;1 and its homolog NIP6;1 were retained in the ER of the kns3 mutants. Our genetic analysis further discovered that two homologs of KNS3, KNSTH1 and KNSTH2, were also involved in the ER exit of NIP5;1. In Arabidopsis protoplasts and tobacco leaves, mCherry-fused KNS3 localized to the ER and Golgi, whereas KNSTH2 localized to the ER. The cytosolic C-terminal tail of KNS3 contains amino acids important for Golgi-to-ER trafficking. Furthermore, the ER-to-Golgi trafficking of KNS3 depended on KNSTH1 and KNSTH2, and the accumulation of these three proteins in Arabidopsis roots depended on each other. We propose that KNS3, KNSTH1, and KNSTH2 function as a cargo-receptor complex mediating the ER exit of NIP5;1.

Multiplexed, image-based pooled screens in primary cells and tissues with PerturbView.

Optical pooled screening (OPS) is a scalable method for linking image-based phenotypes with cellular perturbations. However, it has thus far been restricted to relatively low-plex phenotypic readouts in cancer cell lines in culture due to limitations associated with in situ sequencing of perturbation barcodes. Here, we develop PerturbView, an OPS technology that leverages in vitro transcription to amplify barcodes before in situ sequencing, enabling screens with highly multiplexed phenotypic readouts across diverse systems, including primary cells and tissues. We demonstrate PerturbView in induced pluripotent stem cell-derived neurons, primary immune cells and tumor tissue sections from animal models. In a screen of immune signaling pathways in primary bone marrow-derived macrophages, PerturbView uncovered both known and novel regulators of NF-κB signaling. Furthermore, we combine PerturbView with spatial transcriptomics in tissue sections from a mouse xenograft model, paving the way to in situ screens with rich optical and transcriptomic phenotypes. PerturbView broadens the scope of OPS to a wide range of models and applications.

Clathrin-associated carriers enable recycling through a kiss-and-run mechanism.

Endocytosis and recycling control the uptake and retrieval of various materials, including membrane proteins and lipids, in all eukaryotic cells. These processes are crucial for cell growth, organization, function and environmental communication. However, the mechanisms underlying efficient, fast endocytic recycling remain poorly understood. Here, by utilizing a biosensor and imaging-based screening, we uncover a recycling mechanism that couples endocytosis and fast recycling, which we name the clathrin-associated fast endosomal recycling pathway (CARP). Clathrin-associated tubulovesicular carriers containing clathrin, AP1, Arf1, Rab1 and Rab11, while lacking the multimeric retrieval complexes, are generated at subdomains of early endosomes and then transported along actin to cell surfaces. Unexpectedly, the clathrin-associated recycling carriers undergo partial fusion with the plasma membrane. Subsequently, they are released from the membrane by dynamin and re-enter cells. Multiple receptors utilize and modulate CARP for fast recycling following endocytosis. Thus, CARP represents a previously unrecognized endocytic recycling mechanism with kiss-and-run membrane fusion.

Hybrid convolutional neural network optimized with an artificial algae algorithm for glaucoma screening using fundus images.

We developed an optimized decision support system for retinal fundus image-based glaucoma screening. We combined computer vision algorithms with a convolutional network for fundus images and applied a faster region-based convolutional neural network (FRCNN) and artificial algae algorithm with support vector machine (AAASVM) classifiers. Optic boundary detection, optic cup, and optic disc segmentations were conducted using TernausNet. Glaucoma screening was performed using the optimized FRCNN. The Softmax layer was replaced with an SVM classifier layer and optimized with an AAA to attain enhanced accuracy. Using three retinal fundus image datasets (G1020, digital retinal images vessel extraction, and high-resolution fundus), we obtained accuracy of 95.11%, 92.87%, and 93.7%, respectively. Framework accuracy was amplified with an adaptive gradient algorithm optimizer FRCNN (AFRCNN), which achieved average accuracy 94.06%, sensitivity 93.353%, and specificity 94.706%. AAASVM obtained average accuracy of 96.52%, which was 3% ahead of the FRCNN classifier. These classifiers had areas under the curve of 0.9, 0.85, and 0.87, respectively. Based on statistical Friedman evaluation, AAASVM was the best glaucoma screening model. Segmented and classified images can be directed to the health care system to assess patients' progress. This computer-aided decision support system will be useful for optometrists.

P22-11 An Image-based toxicity screening method: Calcein AM as an indicator of esterase activity in Daphnia magna to detect sublethal effects in vivo

P22-11 An Image-based toxicity screening method: Calcein AM as an indicator of esterase activity in Daphnia magna to detect sublethal effects in vivo

Cardiovascular risk and subclinical atherosclerosis in first-degree relatives of patients with premature cardiovascular disease

BackgroundScreening first-degree relatives (FDRs) of patients with premature coronary artery disease (CAD) is recommended but not routinely performed. ObjectivesTo assess the diagnostic yield and impact on clinical management of a clinical and imaging-based screening program of FDRs delivered in the setting of routine clinical care. MethodsWe recruited FDRs of patients with premature CAD with no personal history of CAD and prospectively assessed for: 1) cardiovascular risk and presence of significant subclinical atherosclerosis (SA) defined as plaque on carotid ultrasound, stenosis >50% or extensive atherosclerosis on coronary computed tomography angiography, or coronary artery calcium scores >100 Agatston units or >75% percentile for age and sex; 2) utilization of preventive medications and lipid levels prior enrolment and after completion of the assessment. ResultsWe assessed 132 FDRs (60.6% females), mean (SD) age 47(17) years old. Cardiovascular risk was high in 38.2%, moderate in 12.2%, and low in 49.6% of FDRs. SA was present in 34.1% of FDRs, including 12.5% in low, 51.9% in moderate, and 55.0% in high calculated risk groups. After assessment, LLT was initiated in 32.6% of FDRs and intensified in 16.0% leading to mean (SD) LDL-C decrease of 1.07(1.10) mmol/L in patients with high calculated risk or SA. LLT was recommended to all patients with high calculated risk, but those with SA were more likely to receive the medications from pharmacies (93.3% vs 60.0%, p = 0.006). ConclusionScreening the FDRs of patients with premature CAD is feasible, may have high diagnostic yield and impact risk factor management.

SSM-Net: Semi-supervised multi-task network for joint lesion segmentation and classification from pancreatic EUS images

Pancreatic cancer does not show specific symptoms, which makes the diagnosis of early stages difficult with established image-based screening methods and therefore has the worst prognosis among all cancers. Although endoscopic ultrasonography (EUS) has a key role in diagnostic algorithms for pancreatic diseases, B-mode imaging of the pancreas can be affected by confounders such as chronic pancreatitis, which can make both pancreatic lesion segmentation and classification laborious and highly specialized. To address these challenges, this work proposes a semi-supervised multi-task network (SSM-Net) to leverage unlabeled and labeled EUS images for joint pancreatic lesion classification and segmentation. Specifically, we first devise a saliency-aware representation learning module (SRLM) on a large number of unlabeled images to train a feature extraction encoder network for labeled images by computing a contrastive loss with a semantic saliency map, which is obtained by our spectral residual module (SRM). Moreover, for labeled EUS images, we devise channel attention blocks (CABs) to refine the features extracted from the pre-trained encoder on unlabeled images for segmenting lesions, and then devise a merged global attention module (MGAM) and a feature similarity loss (FSL) for obtaining a lesion classification result. We collect a large-scale EUS-based pancreas image dataset (LS-EUSPI) consisting of 9,555 pathologically proven labeled EUS images (499 patients from four categories) and 15,500 unlabeled EUS images. Experimental results on the LS-EUSPI dataset and a public thyroid gland lesion dataset show that our SSM-Net clearly outperforms state-of-the-art methods.

Pharmacological modulation of developmental and synaptic phenotypes in human SHANK3 deficient stem cell-derived neuronal models

Phelan-McDermid syndrome (PMDS) arises from mutations in the terminal region of chromosome 22q13, impacting the SHANK3 gene. The resulting deficiency of the postsynaptic density scaffolding protein SHANK3 is associated with autism spectrum disorder (ASD). We examined 12 different PMDS patient and CRISPR-engineered stem cell-derived neuronal models and controls and found that reduced expression of SHANK3 leads to neuronal hyperdifferentiation, increased synapse formation, and decreased neuronal activity. We performed automated imaging-based screening of 7,120 target-annotated small molecules and identified three compounds that rescued SHANK3-dependent neuronal hyperdifferentiation. One compound, Benproperine, rescued the decreased colocalization of Actin Related Protein 2/3 Complex Subunit 2 (ARPC2) with ß-actin and rescued increased synapse formation in SHANK3 deficient neurons when administered early during differentiation. Neuronal activity was only mildly affected, highlighting Benproperine’s effects as a neurodevelopmental modulator. This study demonstrates that small molecular compounds that reverse developmental phenotypes can be identified in human neuronal PMDS models.

A novel in vitro high-content imaging assay for the prediction of drug-induced lung toxicity

The development of inhaled drugs for respiratory diseases is frequently impacted by lung pathology in non-clinical safety studies. To enable design of novel candidate drugs with the right safety profile, predictive in vitro lung toxicity assays are required that can be applied during drug discovery for early hazard identification and mitigation. Here, we describe a novel high-content imaging-based screening assay that allows for quantification of the tight junction protein occludin in A549 cells, as a model for lung epithelial barrier integrity. We assessed a set of compounds with a known lung safety profile, defined by clinical safety or non-clinical in vivo toxicology data, and were able to correctly identify 9 of 10 compounds with a respiratory safety risk and 9 of 9 compounds without a respiratory safety risk (90% sensitivity, 100% specificity). The assay was sensitive at relevant compound concentrations to influence medicinal chemistry optimization programs and, with an accessible cell model in a 96-well plate format, short protocol and application of automated imaging analysis algorithms, this assay can be readily integrated in routine discovery safety screening to identify and mitigate respiratory toxicity early during drug discovery. Interestingly, when we applied physiologically-based pharmacokinetic (PBPK) modelling to predict epithelial lining fluid exposures of the respiratory tract after inhalation, we found a robust correlation between in vitro occludin assay data and lung pathology in vivo, suggesting the assay can inform translational risk assessment for inhaled small molecules.

Revolutionizing vascular health through the temporal convolutional transformer for drug screening and model evolution

Vascular illnesses include a spectrum of disorders affecting the blood vessels and continue to be major global health challenges. Vascular disorders are major worldwide health concerns that require ongoing development of drug screening models in order to find efficient treatment approaches. The ever-changing field of vascular health demands novel ways to medication screening due to the enduring difficulties presented by a variety of illnesses. This abstract explores a paradigm shift, highlighting the innovative role of the Temporal Convolutional Transformer (TCT) in redefining drug screening techniques and promoting model evolution. With its seamless integration of transformer mechanisms and temporal convolutional layers, the TCT is a revolutionary hybrid algorithm. Because of this special combination, the model can simultaneously identify spatial subtleties and capture temporal dynamics in complicated biological information. The temporal convolutional layers of the TCT are particularly good at revealing the sequential dependencies that are present in the development of vascular disorders. The model ensures a detailed knowledge of the developing nature of these diseases by analyzing temporal dynamics with skill. Simultaneously, the TCT’s transformer mechanisms enable a thorough examination of long-range interdependence and global interconnections. The TCT’s dual capability establishes it as a comprehensive solution by offering a nuanced perspective on spatial patterns that are essential for identifying the complexity of vascular health. The TCT is hybrid in the sense that it can learn features at several scales, supporting different temporal and spatial resolutions. This flexibility guarantees that the model may extract pertinent information at various scales, enhancing the characterization of the complex nature of vascular illnesses. The TCT is positioned as a holistic solution since its transformer mechanisms allow for a thorough investigation of long-range dependencies and global interconnections at the same time. This combined capacity offers a sophisticated viewpoint on spatial patterns that is essential for understanding the intricacies of vascular health. Moreover, the remarkable flexibility of the TCT to dynamic temporal scales is consistent with the dynamic character of biological processes, providing a strong and adaptable framework for drug screening. The TCT is used in many different fields of vascular disease research. The model performs well in image-based screening, collecting sequential dependencies over time and processing spatial information from medical imaging datasets. The TCT’s capacity to identify long-range dependencies is essential for molecular interaction research since it helps to clarify the complex pathways associated with vascular disorders. Its adaptability also extends to personalized treatment approaches, which allow for the efficient modeling of patient-specific data, including genetic information. The TCT is a revolutionary tool that will continue to evolve as drug screening models do. By integrating temporal convolutional and transformer mechanisms, its all-encompassing methodology provides a sophisticated and comprehensive approach to deciphering the intricacies of vascular disorders. Across all evaluated metrics, the proposed method achieves near-perfect accuracy of 98.45%, sensitivity of 93.45%, specificity of 98.43%, precision of 84.25, AUC of 99 % and F1 score values 96.01%, respectively. This abstract sheds light on how the TCT has significantly changed drug development approaches, opening the door for more focused and effective therapies in the field of vascular health.

Abstract 4130: Targeting epigenetically-driven inflammation for PDAC metastasis prevention in high-risk individuals

Abstract Background: Pancreatic ductal adenocarcinoma (PDAC) is a lethal cancer with a dire prognosis, in large part due to the late stage at which it is diagnosed in most patients. High-risk individuals (HRIs) identified through family history or germline genetic testing are candidates for imaging-based screening, yet even with intensive surveillance, metastasis often occurs between scans. Our study aims to address this critical clinical challenge by investigating the epigenetic determinants of PDAC metastasis and developing novel strategies for metastasis prevention in HRIs. Methods: Innovative transplantation models of PDAC liver and peritoneal metastasis incorporating DNA barcodes for subclonal tracking were developed to facilitate the identification of subclones with high or low metastatic potential. An in-depth epigenomic characterization and comparison of these subclones was performed utilizing assay for transposase-accessible chromatin with sequencing (ATAC-seq) and RNA-seq. Results: Metastasis-high subclones demonstrate widespread chromatin accessibility differences relative to their metastasis-low peers. These differences are linked to expression changes for genes involved in motility, differentiation, survival, and particularly inflammation. Top epigenetically activated genes in metastasis-high subclones include the pro-inflammatory cytokines IL-1α and IL-23α and the pro-inflammatory cytokine receptor IL-18R1. Discussion: These findings highlight the role of epigenetic changes in PDAC metastasis, especially activation of inflammatory pathways resulting from altered chromatin accessibility. These insights provide a foundation for exploring anti-inflammatory agents for PDAC metastasis prevention. Targeting inflammation, alongside imaging-based screening, could significantly improve outcomes for HRIs. Citation Format: Jesse Handler, Reza Kalhor. Targeting epigenetically-driven inflammation for PDAC metastasis prevention in high-risk individuals [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 4130.

Abstract 4919: AI-enabled hit selection of drug screening on human pancreatic cancer organoids

Abstract Cancer remains one of the leading causes of death in the 21st century. Despite the latest advances in oncology, most cancer patients lack tailored therapeutic approaches with lasting benefit. Measuring the impact of anticancer compounds and their combinations is only possible on ex vivo assays. To this end, patient-derived organoids (PDOs) have been proposed as viable and efficient models for ex vivo testing. PDOs show long-term expansion potential while retaining tumor histopathology as well as cancer gene mutations. However, the translation of organoids in screening applications has so far been hampered by the lack of homogeneity and difficulties in handling and automation. Moreover, organoids are typically randomly distributed across the culture which complicates imaging and images analyses. To overcome these challenges, we set up a compound screening workflow with PDOs using Gri3D platform, comprising plates with micro-cavities suitable for high-throughput and reproducible organoid culture. Based on a standard 96 microtiter plate, each well contains a microwell array patterned in a cell repellent hydrogel. On Gri3D®, organoids are robustly generated in the microwells and are located in the same imaging plane. This greatly facilitates quantitative analyses in high content image-based screens. Furthermore, the pipetting port enables automation of cell seeding, media exchange and compound incubation with liquid-handlers, thus increasing assay reproducibility. In the presented work, we exposed human pancreatic cancer PDOs to a panel of anti-cancer compounds at different doses and followed their response to the drugs using viability dyes Calcein AM (live stain) and Ethidium Homodimer-1 (dead stain) with high-content confocal imaging. Using an AI-based approach, we efficiently detected each single organoid and extracted phenotypic features from all three channels (TL, live stain and dead stain) which correlate with cytotoxicity. The extracted features (more than 100) were first dimensionally reduced to 3 components using either PCA or UMAP, the treatments were then visualized in 3D scatter plots, ranked and clustered using machine learning. The data suggests that the compound treatment Palbociclib 50uM has significant cytotoxicity effects similar to the positive control, which is consistent with the traditional live/dead analysis. The AI approach demonstrates feasibility to perform drug screening in a robust and un-biased data analysis approach. Citation Format: Zhisong Tong, Angeline Lim, Marine Meyer, Maria Clapes, Oksana Sirenko, Nathalie Brandenberg. AI-enabled hit selection of drug screening on human pancreatic cancer organoids [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 4919.

Abstract 4722: Enhanced phenotypic screening: A kinetic, multiplexed approach to analyzing drug effects using Incucyte® Live-Cell Analysis System

Abstract Phenotypic screening is a valuable tool for analysis of the complex mix of effects that drug treatment has on cells. Traditionally image-based screening uses fixed cells stained with reagents at a pre-chosen end point, however our technique combines kinetic, label-free readouts with non-perturbing fluorescence reporters to enable information-rich quantification of compound effects. A549 lung cancer cells were treated with a library of over 800 FDA-approved drugs while images were acquired in the Incucyte® Live-Cell Analysis System every two hours for four days. Cell growth and viability were measured using label-free readouts of confluence and dead cell count while cell cycle, cell death and Akt pathways were investigated using multiplexed fluorescent readouts. Cytotoxicity was measured using label-free Incucyte® AI Cell Health Analysis Software Module which uses two neural networks to segment individual cells and classify them as live or dead. Assay robustness was quantified using the time course of positive and negative controls (camptothecin and vehicle, respectively). Z’ reached a maximal value >0.9 between 48h and 72h; at 72h 4.4% of the compounds within the library induced >50% cell death. Correlation between % Live cells and % Confluence indicated 3 types of effect: cytotoxic (low viability, low confluence), cytostatic (high viability, low confluence), and none (high viability, high confluence). Morphological changes were observed in a multitude of wells with 29 compounds inducing cell enlargement >1800 µm2, suggesting induction of senescence. Hits from fluorescent readouts were identified as the compounds which perturbed the quantified readout more than 3 standard deviations away from the mean value of the vehicle. Mechanisms of cell death were examined by correlative analysis of the total dead cell population (label-free cell death identification), caspase-active (caspase 3/7-reagent positive) and Annexin V positive cells. This enabled rapid identification of compounds which induced apoptosis by caspase-independent pathways, including the serotonin-selective reuptake inhibitor Sertraline. Kinetic plots of cell cycle stage (% S|G2|M or G1) revealed contrasting mechanisms. Mycophenolic acid induced cell cycle arrest, observed as a consistent increase in the percentage of cells in G1 over time. However flumazenil displayed repeated, transient peaks in the percentage of cells in S|G2|M over the same 72h period, indicating synchronisation behavior. Incucyte® Live-Cell Analysis System as a phenotypic screening platform provides valuable kinetic data from cells within the physiologically relevant environment of a standard incubator. Label-free readouts yield direct cellular information and can be combined with fluorescent data to provide vastly more information than a single endpoint readout. Citation Format: Kirsty McBain, Gillian Lovell, Jasmine Trigg, Nicola Bevan, Daniel Appledorn. Enhanced phenotypic screening: A kinetic, multiplexed approach to analyzing drug effects using Incucyte® Live-Cell Analysis System [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 4722.

Abstract 7546: Targeting NSD2 to reverse lineage plasticity and drug resistance in neuroendocrine prostate cancer and subtypes of mCRPC

Abstract Treatment with highly potent anti-androgens, such as enzalutamide and abiraterone promotes lineage plasticity in metastatic castration-resistant prostate cancer (mCRPC), which results in intra-tumor heterogeneity and emergence of mCRPC subtypes. The histological transformation from adenocarcinoma to aggressive neuroendocrine prostate cancer (CRPC-NE) has been associated with a loss of dependency on lineage-survival signals, leading to targeted drug resistance. Epigenomic reprogramming might be a fundamental driver of lineage plasticity. To determine CRPC-NE vulnerabilities, we have performed image-based screening using a small library of antibodies targeting different histone marks on CRPC-NE organoids derived from genetically engineered mice. We find that the histone mark H3K36me2 and the histone methyltransferase NSD2 play important roles in maintain the CRPC-NE epigenetic state. Knockout of NSD2 or ablation of H3K36me2 using an oncohistone mutant H3.3K36M reverted CRPC-NE to an adenocarcinoma phenotype. Simultaneous profiling of the transcriptome and epigenome from single cells verified this lineage conversion and the generation of cell states with canonical AR signaling. Moreover, H3K36me2 or NSD2 depleted mouse and human CRPC-NE organoids responded to enzalutamide treatment in vitro and in vivo, suggesting a reversal of castration-resistance. Most importantly, a small molecule inhibitor of NSD2 in combination with enzalutamide leads to growth suppression or apoptosis of mouse and human CRPC-NE organoids, as well as organoids of other CRPC subtypes. In conclusion, inhibition of NSD2 reverts the epigenomic state of CRPC-NE, reversing lineage plasticity and restoring anti-androgen sensitivity. Thus, we suggest that the combination of NSD2 inhibition with AR inhibition may represent a novel therapeutic approach for patients with CRPC-NE or even other CRPC subtypes. Citation Format: Jia J. Li, Alessandro Vasciaveo, Dimitrios Karagiannis, Xiao Chen, Chen Yu, Andrea Califano, Chao Lu, Michael M. Shen. Targeting NSD2 to reverse lineage plasticity and drug resistance in neuroendocrine prostate cancer and subtypes of mCRPC [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 7546.

Root system architecture change in response to waterlogging stress in a 448 global collection of rapeseeds (Brassica napus L.).

A novel image-based screening method for precisely identifying genotypic variations in rapeseed RSA under waterlogging stress was developed. Five key root traits were confirmed as good indicators of waterlogging and might be employed in breeding, particularly when using the MFVW approach. Waterlogging is a vital environmental factor that has detrimental effects on the growth and development of rapeseed (Brassica napus L.). Plant roots suffer from hypoxia under waterlogging, which ultimately confers yield penalty. Therefore, it is crucially important to understand the genetic variation of root system architecture (RSA) in response to waterlogging stress to guide the selection of new tolerant cultivars with favorable roots. This research was conducted to investigate RSA traits using image-based screening techniques to better understand how RSA changes over time during waterlogging at the seedling stage. First, we performed a t-test by comparing the relative root trait value between four tolerant and four sensitive accessions. The most important root characteristics associated with waterlogging tolerance at 12 h are total root length (TRL), total root surface area (TRSA), total root volume (TRV), total number of tips (TNT), and total number of forks (TNF). The root structures of 448 rapeseed accessions with or without waterlogging showed notable genetic diversity, and all traits were generally restrained under waterlogging conditions, except for the total root average diameter. Additionally, according to the evaluation and integration analysis of 448 accessions, we identified that five traits, TRL, TRSA, TRV, TNT, and TNF, were the most reliable traits for screening waterlogging-tolerant accessions. Using analysis of the membership function value (MFVW) and D-value of the five selected traits, 25 extremely waterlogging-tolerant materials were screened out. Waterlogging significantly reduced RSA, inhibiting root growth compared to the control. Additionally, waterlogging increased lipid peroxidation, accompanied by a decrease in the activities of superoxide dismutase (SOD), peroxidase (POD), and catalase (CAT). This study effectively improves our understanding of the response of RSA to waterlogging. The image-based screening method developed in this study provides a new scientific guidance for quickly examining the basic RSA changes and precisely predicting waterlogging-tolerant rapeseed germplasms, thus expanding the genetic diversity of waterlogging-tolerant rapeseed germplasm available for breeding.

Screening approaches for lung cancer by blood-based biomarkers: Challenges and opportunities.

Lung cancer (LC) is one of the leading causes for cancer-related deaths in the world, accounting for 28% of all cancer deaths in Europe. Screening for lung cancer can enable earlier detection of LC and reduce lung cancer mortality as was demonstrated in several large image-based screening studies such as the NELSON and the NLST. Based on these studies, screening is recommended in the US and in the UK a targeted lung health check program was initiated. In Europe lung cancer screening (LCS) has not been implemented due to limited data on cost-effectiveness in the different health care systems and questions on for example the selection of high-risk individuals, adherence to screening, management of indeterminate nodules, and risk of overdiagnosis. Liquid biomarkers are considered to have a high potential to address these questions by supporting pre- and post- Low Dose CT (LDCT) risk-assessment thereby improving the overall efficacy of LCS. A wide variety of biomarkers, including cfDNA, miRNA, proteins and inflammatory markers have been studied in the context of LCS. Despite the available data, biomarkers are currently not implemented or evaluated in screening studies or screening programs. As a result, it remains an open question which biomarker will actually improve a LCS program and do this against acceptable costs. In this paper we discuss the current status of different promising biomarkers and the challenges and opportunities of blood-based biomarkers in the context of lung cancer screening.

Pooled CRISPR screening of high-content cellular phenotypes using ghost cytometry

Recent advancements in image-based pooled CRISPR screening have facilitated the mapping of diverse genotype-phenotype associations within mammalian cells. However, the rapid enrichment of cells based on morphological information continues to pose a challenge, constraining the capacity for large-scale gene perturbation screening across diverse high-content cellular phenotypes. In this study, we demonstrate the applicability of multimodal ghost cytometry-based cell sorting, including both fluorescent and label-free high-content phenotypes, for rapid pooled CRISPR screening within vast cell populations. Using the high-content cell sorter operating in fluorescence mode, we successfully executed kinase-specific CRISPR screening targeting genes influencing the nuclear translocation of RelA. Furthermore, using the multiparametric, label-free mode, we performed large-scale screening to identify genes involved in macrophage polarization. Notably, the label-free platform can enrich target phenotypes without requiring invasive staining, preserving untouched cells for downstream assays and expanding the potential for screening cellular phenotypes even when suitable markers are absent.

A deep learning dataset for sample preparation artefacts detection in multispectral high-content microscopy

High-content image-based screening is widely used in Drug Discovery and Systems Biology. However, sample preparation artefacts may significantly deteriorate the quality of image-based screening assays. While detection and circumvention of such artefacts could be addressed using modern-day machine learning and deep learning algorithms, this is widely impeded by the lack of suitable datasets. To address this, here we present a purpose-created open dataset of high-content microscopy sample preparation artefact. It consists of high-content microscopy of laboratory dust titrated on fixed cell culture specimens imaged with fluorescence filters covering the complete spectral range. To ensure this dataset is suitable for supervised machine learning tasks like image classification or segmentation we propose rule-based annotation strategies on categorical and pixel levels. We demonstrate the applicability of our dataset for deep learning by training a convolutional-neural-network-based classifier.

Deep Learning in Breast Cancer Imaging: A Decade of Progress and Future Directions.

Breast cancer has reached the highest incidence rate worldwide among all malignancies since 2020. Breast imaging plays a significant role in early diagnosis and intervention to improve the outcome of breast cancer patients. In the past decade, deep learning has shown remarkable progress in breast cancer imaging analysis, holding great promise in interpreting the rich information and complex context of breast imaging modalities. Considering the rapid improvement in deep learning technology and the increasing severity of breast cancer, it is critical to summarize past progress and identify future challenges to be addressed. This paper provides an extensive review of deep learning-based breast cancer imaging research, covering studies on mammograms, ultrasound, magnetic resonance imaging, and digital pathology images over the past decade. The major deep learning methods and applications on imaging-based screening, diagnosis, treatment response prediction, and prognosis are elaborated and discussed. Drawn from the findings of this survey, we present a comprehensive discussion of the challenges and potential avenues for future research in deep learning-based breast cancer imaging.