Artificial intelligence-driven high-content imaging decodes for NLRP7-mutant recurrent hydatidiform moles.

Artificial intelligence-driven high-content imaging decodes for NLRP7-mutant recurrent hydatidiform moles - PMC Skip to main content An official website of the United States government Here's how you know Here's how you know Official websites use .gov A .gov website belongs to an official government organization in the United States. Secure .gov websites use HTTPS A lock ( Lock Locked padlock icon ) or https:// means you've safely connected to the .gov website. Share sensitive information only on official, secure websites. Search Log in Dashboard Publications Account settings Log out Search… Search NCBI Primary site navigation Search Logged in as: Dashboard Publications Account settings Log in Search PMC Full-Text Archive Search in PMC Journal List User Guide PERMALINK Copy As a library, NLM provides access to scientific literature. Inclusion in an NLM database does not imply endorsement of, or agreement with, the contents by NLM or the National Institutes of Health. Learn more: PMC Disclaimer | PMC Copyright Notice iScience . 2026 Mar 26;29(4):115479. doi: 10.1016/j.isci.2026.115479 Search in PMC Search in PubMed View in NLM Catalog Add to search Artificial intelligence-driven high-content imaging decodes for NLRP7-mutant recurrent hydatidiform moles Jiayang Wan Jiayang Wan 1 Department of Gynecology, the First Affiliated Hospital, Zhejiang University School of Medicine, Hangzhou, China Find articles by Jiayang Wan 1 , Tianliang Li Tianliang Li 2 Materials Genome Institute, Shanghai Engineering Research Center for Integrated Circuits and Advanced Display Materials, and Shanghai Engineering Research Center of Organ Repair, Shanghai University, Shanghai, China Find articles by Tianliang Li 2 , Limeng Cai Limeng Cai 1 Department of Gynecology, the First Affiliated Hospital, Zhejiang University School of Medicine, Hangzhou, China Find articles by Limeng Cai 1 , Yating Zhao Yating Zhao 1 Department of Gynecology, the First Affiliated Hospital, Zhejiang University School of Medicine, Hangzhou, China Find articles by Yating Zhao 1 , Jianhua Qian Jianhua Qian 1 Department of Gynecology, the First Affiliated Hospital, Zhejiang University School of Medicine, Hangzhou, China Find articles by Jianhua Qian 1, 3, ∗ Author information Article notes Copyright and License information 1 Department of Gynecology, the First Affiliated Hospital, Zhejiang University School of Medicine, Hangzhou, China 2 Materials Genome Institute, Shanghai Engineering Research Center for Integrated Circuits and Advanced Display Materials, and Shanghai Engineering Research Center of Organ Repair, Shanghai University, Shanghai, China ∗ Corresponding author qianjianhua@zju.edu.cn 3 Lead contact Received 2025 Sep 18; Revised 2026 Jan 23; Accepted 2026 Mar 23; Collection date 2026 Apr 17. © 2026 The Authors This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). PMC Copyright notice PMCID: PMC13091066  PMID: 42006372 Summary Recurrent hydatidiform moles (RHMs) are a gestational disorder primarily caused by maternal-effect loss-of-function mutations in NLRP7. This study established an in vitro model of NLRP7 mutation using patient-derived induced pluripotent stem cells (iPSCs) and integrated high-content imaging (HCI) with artificial intelligence (AI) to dissect mutation-induced multi-organellar dysfunction. Multiparametric HCI enabled synchronous capture of cellular and subcellular phenotypes, including mitochondrial function and lysosomal distribution. We developed a bio-inspired AI framework (BioVision-Segmentation) that combines a hybrid Transformer-Voronoi architecture for segmentation with mutual information analysis to quantify organelle interactions. Our findings reveal that NLRP7 mutations disrupt the lysosome-mitochondria crosstalk hub, leading to energy metabolism dysregulation, reactive oxygen species (ROS) accumulation, and organelle spatial distribution defects. Furthermore, transcriptome sequencing corroborated these findings. This study elucidates the organellar pathogenesis of RHMs and provides a technological platform for exploring therapeutic targets, facilitating a shift toward early embryo protection strategies. Subject areas: Biological sciences, Methodology in biological sciences, Machine learning Graphical abstract Open in a new tab Highlights • Integrates high-content imaging with AI to decode organelle phenotypes in iPSCs • Develops the BioVision-Segmentation framework to quantify organelle interactions • Reveals that NLRP7 mutation disrupts lysosome-mitochondria crosstalk • Provides a platform for targeting organelle dysfunction in gestational disorders Biological sciences; Methodology in biological sciences; Machine learning Introduction Recurrent hydatidiform moles (RHMs) are a gestational disorder characterized by early embryonic developmental arrest and abnormal trophoblast proliferation, clinically manifesting as recurrent preimplantation embryonic loss and miscarriage. 1 , 2 Effective interventions are currently lacking. Over 50% of RHMs cases are driven by loss-of-function mutations in the maternal-effect gene NLRP7. 3 , 4 , 5 The absence of a rodent ortholog for NLRP7 has hindered mechanistic studies using traditional animal models. 6 Therefore, constructing humanized in vitro models and elucidating the molecular pathways underlying NLRP7 mutation-induced pathogenesis are critical for understanding RHMs etiology. Patient-derived induced pluripotent stem cells (iPSCs), with their capacity for long-term expansion, directed differentiation, and retention of the disease genetic background, provide an ideal platform for investigating early developmental events. 7 , 8 , 9 , 10 Mitochondrial dysfunction is recognized as a central driver of embryonic developmental failure, regulating developmental homeostasis through energy metabolism, reactive oxygen species (ROS) balance, and organelle interaction networks. 11 For instance, loss of NLRP5 (a functional homolog of NLRP7) in mouse models disrupts mitochondrial positioning, impairs ATP synthesis, and causes ROS accumulation, culminating in developmental arrest at the blastocyst stage. 12 Similarly, deficiencies in Zbed3 or Nlrp4f cause organelle spatial distribution defects, further underscoring the critical role of organellar homeostasis in fertility. 13 , 14 Especially, mitochondrial oxidative damage plays important roles in the pathogenesis of complete hydatidiform mole. These findings suggest that investigating the impact of NLRP7 mutations on mitochondrial function using iPSCs models will establish a theoretical foundation for therapeutic strategies. 15 , 16 Lysosomes, as key degradative and signaling organelles, are increasingly implicated in early embryonic fitness through their roles in autophagy, metabolic coupling, and inflammasome regulation. Members of the NLRP family, including NLRP7, are hypothesized to interface with lysosomal pathways, potentially influencing lysosomal acidification, protease activity, or membrane stability, processes critical for organelle quality control and immune homeostasis. 17 , 18 , 19 Thus, investigating lysosomal status in NLRP7-deficient models is mechanistically warranted. High-content imaging (HCI), combining automated high-resolution microscopy with multi-channel fluorescence labeling, offers a powerful panoramic analysis platform by enabling high-throughput, multiparametric, synchronous capture of cellular morphology, subcellular structures, and functional molecules. 20 , 21 Notably, established multi-channel fluorescence profiling approaches, such as the Cell Painting assay, have demonstrated the utility of multiplexed organelle staining for systematic phenotypic screening. 22 , 23 In this study, we adopt a similar multi-channel imaging strategy tailored for iPSCs, focusing on organelles implicated in NLRP7-associated pathogenesis. However, traditional HCI faces significant bottlenecks: low efficiency in automated interpretation of massive image datasets, difficulties in precise organelle segmentation within complex biological contexts, and challenges in uncovering implicit relationships between multidimensional features, limiting its depth in mechanistic studies. 24 , 25 The integration of artificial intelligence (AI) technologies fundamentally overcomes these limitations. Deep learning-based segmentation models (e.g., U-Net and Transformer) can automatically identify cell boundaries and organelle spatial distributions; graph neural networks (GNNs) can quantify organelle interaction dynamics, such as mitochondria-lysosome contacts; and active learning strategies continuously optimize model accuracy in recognizing rare phenotypes (e.g., apoptotic bodies and autolysosomes). 26 , 27 , 28 , 29 , 30 Consequently, applying HCI to capture phenotypes in iPSCs-based NLRP7 mutant models and utilizing AI for effective image analysis to uncover underlying relationships is crucial for advancing RHMs therapeutics. This study leverages patient-derived NLRP7-mutant iPSCs combined with an HCI-AI integrated platform to systematically elucidate mutation-induced multi-organellar dysfunction. We employed multi-channel HCI to synchronously capture multidimensional phenotypes: cellular morphology, viability, mitochondrial function, and organelle interactions. We further developed a bio-inspired AI framework (BioVision-Segmentation), utilizing a hybrid Transformer-Voronoi architecture for multi-channel cell segmentation. Mutual information quantification was applied to analyze mitochondria-lysosome co-staining features, ultimately defining a lysosome-mitochondria crosstalk hub and revealing its role as a driver of developmental arrest. The organellar homeostasis scoring system established herein provides a platform for targeted interventions, promoting a shift in RHMs management from passive therapy toward early embryo protection ( Figure 1 ). Figure 1. Open in a new tab The scheme of the research AI-driven HCI decodes for NLRP7-mutant RHMs. Results Establishment of iPSCs-based in vitro models Given that NLRP7 is a human-specific gene, traditional animal models are inadequate for recapitulating its pathological mechanisms. 31 Patient-derived iPSCs, leveraging their capacity for long-term expansion, potential for directed differentiation, and retention of the disease genetic background, offer a powerful platform to construct humanized in vitro models for dissecting the molecular pathways underlying NLRP7 mutation-associated disease. The iPSCs utilized in this study were generated by reprogramming peripheral blood mononuclear cells (PBMCs) from a RHMs patient using non-integrating methods. The patient was a 39-year-old female who experienced four RHMs events; her sister also had a history of two RHMs, while her parents and brother had normal fertility ( Figure 2 ). Whole-exome sequencing (WES) of the patient’s blood sample revealed a homozygous mutation in the NLRP7 gene ( NM_001127255.2 ). This mutation is located in exon 5 (c.2179 G > A), resulting in an arginine-to-glutamine substitution at amino acid position 693 (p. Arg 693 Gln). This line is henceforth referred to as the R693Q iPSCs. The mutation site resides within the P loop motif (amino acids 680–688) of the NACHT domain, a region critical for ATP binding. We therefore confirmed the successful generation and characterization of patient-specific iPSCs, which serve as a cellular model for investigating the molecular mechanisms of NLRP7-associated RHMs. Figure 2. Open in a new tab Cell Sanger sequencing and patient clinical information (A) Sanger sequencing was used to analyze NLRP7 mutations in iPSCs, showing the Sanger sequencing peak diagram and sequence alignment for the R693Q iPSCs mutation site. (B) Sanger sequencing was used to analyze NLRP7 mutations in iPSCs, showing the Sanger sequencing peak diagram and sequence alignment for wild-type iPSCs as normal controls. (C) Clinical information and family genetic map of the patient from whom the R693Q iPSCs were derived. The proband (arrow) is a 39-year-old female with four RHMs episodes; her sister also had two RHMs, while her parents and brother are unaffected. Control iPSCs (WT iPSCs) were derived from a woman with normal fertility using identical methods. Sanger sequencing chromatograms confirming the genotypes of the R693Q iPSCs and WT iPSCs lines are presented in Figure 2 . Both the established WT iPSCs and R693Q iPSCs were cultured in mTeSR1 medium and used to construct humanized in vitro models harboring the NLRP7 gene. Data acquisition within iPSCs-based in vitro models To further evaluate the constructed humanized in vitro models harboring the NLRP7 gene, bright-field microscopy was performed to assess cellular morphological integrity. Acquired image parameters included: cell perimeter, area, and membrane smoothness (evaluating the presence of blebbing or vesicular structures); nucleus size and area (to determine potential nuclear pyknosis). As illustrated in Figure 3 A, R693Q iPSCs progressively lost aggregation integrity during culture. Over time, cell aggregation edges exhibited significant loosening, eventually fragmenting into multiple satellite clusters with visibly compromised intercellular connections compared with WT iPSCs. Figure 3. Open in a new tab Basic information on R693Q iPSCs and WT iPSCs (A) Brightfield images of R693Q iPSCs and WT iPSCs. (B) Live-cell staining images of R693Q iPSCs and WT iPSCs using Calcein AM. (C) Proliferation of R693Q iPSCs and WT iPSCs using Edu. (D) Immunofluorescence images of apoptosis-related caspase-3 antibodies in R693Q iPSCs and WT iPSCs. Scale bars, 100 μm. To further assess the growth status of the humanized in vitro models, cell counting was performed using Hoechst 33342 nuclear staining, and viable cells were identified via Calcein AM staining. 32 Calcein AM staining was used to analyze the impact of the NLRP7 mutation on cell viability. Calcein AM is a widely used fluorescent probe for labeling viable cells. This non-polar, cell-permeant compound is hydrolyzed by intracellular esterases upon entering the cell, cleaving its acetoxymethyl ester groups to produce calcein, a highly fluorescent polar molecule. Trapped within viable cells due to its membrane impermeability, calcein emits green fluorescence, enabling the detection and quantification of live cells. Parameters analyzed included: Hoechst 33342-stained nuclear count and intensity, the percentage of Calcein AM-positive viable cells, and the total area of viable cells. As shown in Figure 3 B, extensive green fluorescence was observed in both early culture images, indicating a high proportion of viable cells at this stage. However, the fluorescence of R693Q iPSCs is weaker than WT iPSCs with prolonged culture duration, increasing evidence of apoptosis appeared, potentially due to activation of apoptotic pathways within iPSCs under in vitro conditions. Furthermore, mitochondrial energy crisis might contribute to proliferative collapse. A reduction in EdU-labeled S-phase cells could indicate that mitochondrial dysfunction impacts cellular proliferation. 33 The Click-iT EdU assay was employed to evaluate the effect of the NLRP7 mutation on cell proliferation ( Figures 3 C and S1 ). Sustained mitochondrial dysfunction, oxidative stress, and potential cell-cycle arrest are expected to lead to slowed proliferation. During cell growth, early apoptotic events include loss of mitochondrial membrane potential and caspase activation. These are followed by morphological changes such as membrane blebbing and cell detachment, culminating in late-stage events like nuclear remodeling and eventual loss of sarcolemma integrity. Apoptosis was assessed using immunofluorescence analysis with an antibody against activated Caspase-3. 34 , 35 This approach targets cleaved Caspase-3, a well-established executioner protease in apoptotic pathways, providing specific molecular evidence of apoptosis beyond morphological changes. As depicted in Figures 3 D and S1 , R693Q iPSCs displayed apoptotic morphology, characterized by cell shrinkage and detachment. These cells exhibited enhanced fluorescence intensity for activated Caspase-3, localized to both the nucleus and perinuclear regions. In contrast, WT iPSCs, representing healthy proliferating cells, showed minimal apoptotic activity. Positive Caspase-3 signals in WT cells were predominantly nuclear and exhibited significantly lower intensity. Mitochondrial marker collection Mitochondria play a crucial role in early embryonic development. Mutations in NLRP7 may lead to impairments in mitochondrial morphology and function. 36 Advances in microscopy-based HCI technologies have enabled the visualization of mitochondrial structure and function within complex cellular environments. 37 MitoTracker staining was employed to assess mitochondrial mass, quantity, and morphology. 38 As depicted in Figures 4 D and S2 , R693Q iPSCs exhibited a reduction in mitochondrial number and abnormal morphology, with pronounced mitochondrial shrinkage. This decrease may result from the clearance of damaged mitochondria via mitophagy. In contrast, a healthy mitochondrial network was evident in the WT iPSCs group, maintaining normal quantity and morphology. Mitochondrial membrane potential (ΔΨm) was assessed using tetramethyl rhodamine methyl ester (TMRM) staining. TMRM is a lipophilic cationic dye that accumulates in the mitochondrial matrix in a membrane potential-dependent manner. 39 In the WT iPSC group, TMRM accumulated robustly within the mitochondrial matrix due to the electrochemical gradient, emitting strong fluorescence, indicative of normal mitochondrial status. Conversely, in R693Q iPSCs, TMRM accumulation was diminished, reflected by reduced fluorescence intensity, potentially associated with mitochondrial dysfunction ( Figures 4 A and S2 ). Figure 4. Open in a new tab Fluorescence images of organelle-related probes in R693Q iPSCs and WT iPSCs (A) TMRM staining images of mitochondrial membrane potential-related probes in R693Q iPSCs and WT iPSCs. (B) Fluorescence images of R693Q iPSC and WT iPSC stained with JC1, a mitochondrial membrane potential-related probe. (C) Fluorescence images of R693Q iPSCs and WT iPSCs stained with MitoSOX, a probe for mitochondrial ROS levels. (D) Co-tracking of mitochondria and lysosomes in R693Q iPSC and WT iPSC. Scale bars, 100 μm. Furthermore, we concurrently utilized JC-1 to characterize mitochondrial membrane potential. JC-1 analysis employs a dual-color approach based on aggregate (red)/monomer (green) fluorescence. 26 Under normal ΔΨm, JC-1 forms J-aggregates, yielding a high red/green fluorescence ratio. As shown in Figure 4 B, a bright green fluorescence appeared in R693Q iPSCs, indicating a change in membrane potential. The NLRP7 mutation likely causes mitochondrial functional impairment, with collapse of membrane potential being an early event. Prolonged culture of the R693Q iPSCs revealed alterations in mitochondrial morphology, characterized by shrinkage and reduced size. Concomitant changes in membrane potential were observed: compared to the red fluorescence in WT iPSCs, these mitochondria exhibited intense green fluorescence, indicative of altered membrane potential ( Figure 4 B). Mitochondrial ROS levels were monitored using MitoSOX Red staining (targeting mitochondrial superoxide). In the NLRP7 -mutant R693Q iPSCs, mitochondrial dysfunction (low ΔΨm) and consequent potential inflammasome-related stress led to excessive ROS production, manifested as elevated MitoSOX fluorescence intensity. Conversely, ROS generated by healthy mitochondria in the WT iPSCs group were effectively scavenged, maintaining low MitoSOX fluorescence intensity ( Figure 4 C). Co-tracking of mitochondria and lysosomes Mitochondrial and lysosomal activities are intimately linked. Lysosomes support mitochondrial metabolism by regulating iron homeostasis, and dysfunctional lysosomes can impair mitochondrial function, resulting in the accumulation of damaged mitochondria. Mitochondria and lysosomes exhibit both physical and functional interactions. In mammalian cells, lysosomes can directly contact mitochondria, thereby driving mitochondrial fission. To investigate the interplay within the mitochondria-lysosome axis triggered by the NLRP7 mutation, mitochondria were labeled using MitoTracker Green, while lysosomes were simultaneously tracked with LysoTracker ( Figures 4 D, S3 , and S4 ). In the NLRP7 -mutant R693Q group, mitochondria displayed reduced numbers and/or morphological abnormalities. Consequently, in response to mitochondrial damage (mitophagy) and potential protein aggregation, autophagy may be activated. This was manifested in lysosomes as intensified fluorescence intensity, indicative of increased lysosomal number or enlarged size. However, chronic cellular stress could ultimately compromise lysosomal function, evidenced by diminished fluorescence intensity or aberrant distribution. In contrast, the WT iPSC group exhibited healthy mitochondria maintaining normal morphology, with lysosomes displaying fluorescence intensity consistent with normal functional levels. In addition, the observed lysosomal alterations suggest that the NLRP7 mutation may directly disrupt lysosomal homeostasis, thereby contributing to the mitochondrial dysfunction and developmental arrest observed in our model. It is plausible that NLRP7 dysfunction directly impairs lysosomal acidification or hydrolase activity. Such a primary lysosomal defect would, in turn, compromise mitochondrial quality control via mitophagy, creating a vicious cycle of organellar stress and inflammatory activation. The establishment of BioVision-segmentation We employed an HCI system to acquire images of human iPSCs. To enhance image contrast and visualization, raw image data were preprocessed using contrast-limited adaptive histogram equalization (CLAHE) and pseudo-RGB conversion. During the segmentation stage, the DAPI-stained nuclear channel was processed with the SAM model, a vision transformer-based framework pre-trained on large-scale datasets, which enabled the automatic generation of precise nu-clear masks. To ensure high-fidelity segmentation, thresholds for predicted intersection-over-union (IoU) and stability scores were applied to filter out low-confidence results. To prevent excessive overlap, we further applied an area-based filtering strategy that preferentially retained smaller masks, thereby reducing false positives and noise. On the basis of the refined nuclear segmentation, a range of morphological descriptors was extracted—including area, circularity, smoothness, aspect ratio, and eccentricity—to quantify nuclear structural integrity and infer potential functional states. Subsequently, Voronoi diagrams were constructed from nuclear centroids to quantify spatial distribution and cell-to-cell proximity. A series of selection criteria was then introduced to optimize the representation of intercellular interactions, ensuring that only biologically meaningful nuclear regions were retained. The criteria were defined as follows in supplemental information. The primary purpose of these filtering criteria was to eliminate regions inconsistent with biological expectations, thereby enhancing the reliability of Voronoi-based segmentation while ensuring that the retained cellular regions adequately covered the cytoplasmic area without including excessive non-informative regions. Collectively, these procedures maximized the quality of the segmented cellular dataset. Following segmentation and filtering, texture descriptors were extracted from different imaging channels, including the gray level co-occurrence matrix (GLCM), local binary patterns (LBP), and entropy. These advanced features were computed to probe the textural characteristics of stained organelles in greater depth. Such descriptors capture the intrinsic complexity of cellular textures, thereby revealing the spatial distribution and potential functional states of subcellular structures. Specifically, they included: Contrast: quantifies the degree of intensity variation in the image texture. Contrast = ∑ i , j ‖ i − j ‖ 2 P ( i , j ) Homogeneity: measures the similarity of neighboring pixel pairs. Homogeneity = ∑ i , j P ( i , j ) 1 + ‖ i − j ‖ Correlation: characterizes the linear dependency of gray level gradients in the image. Correlation = ∑ i , j ( i , j ) P ( i , j ) − u i u j σ i σ j LBP: captures microstructural image features by encoding local neighborhood patterns into a histogram. In LBP, the value of the central pixel is compared with those of its surrounding pixels to generate a binary pattern, which is then converted into a decimal value to construct the histogram representation. Entropy: quantifies the complexity of texture patterns by measuring the randomness of pixel intensity distribution. Entropy = ∑ i p ( i ) log 2 p ( i ) In analyzing the aforementioned cell-level features, we established a systematic ML pipeline encompassing feature selection, model selection, model optimization, and evaluation. The final models were trained and refined through 10-fold cross-validation, which was employed both for hyperparameter tuning and for the selection of the optimal model that ultimately yielded the results reported herein. To evaluate the performance of established segmentation tools in our specific cellular context, we conducted a comparative analysis. As shown in Figure S5 , conventional DAPI-based object identification in CellProfiler showed limited accuracy in delineating individual cells within the dense, confluent iPSC colonies characteristic of our model. While deep learning-based CellPose provided robust nucleus segmentation from the DAPI channel ( Figure S6 ), its direct application to organelle-specific fluorescence channels (like MitoTracker) failed to generate reliable masks for whole-cell or organelle region analysis. These results substantiate the need for the tailored segmentation pipeline developed in this study, which combines precise nuclear identification with Voronoi-based cytoplasmic inference to enable accurate single-cell phenotyping in morphologically complex iPSC cultures. The analysis and prediction of BioVision-segmentation Cell segmentation Traditional cell segmentation methods typically rely on image intensity information to delineate cell boundaries, for example, by applying watershed algorithms to construct topological maps. However, when cellular contours are poorly defined, such methods often fail to accurately separate individual cells. To overcome this limitation, we performed nuclear segmentation using the DAPI channel, which provides unambiguous delineation of nuclei. Accurate identification of nuclear regions was achieved with the Segment Anything Model (SAM), which can effectively capture the complexity of nuclear boundaries even under challenging conditions, such as when heterogeneous marker distribution within the nucleus leads to blurred or ambiguous edges. Building on DAPI-stained nuclei, we employed Voronoi-based segmentation to construct cellular topologies. This approach effectively expands each nuclear region into a corresponding cellular compartment, ensuring that every nucleus defines a unique cell territory. Specifically, we first applied the SAM to the DAPI channel to identify nuclei and generate preliminary nuclear segmentation masks, which served as pseudo ground truth. Subsequently, Voronoi tessellation was used to partition the image space, assigning pixels to individual cells based on radial propagation from the nuclear centroid to the surrounding area, thereby delineating the boundaries of single-cell regions. To ensure the accuracy of the segmented regions, we applied a series of refined post-filtering strategies. Specifically, we evaluated key metrics such as the number of adjacent regions and the minimum internal angle, allowing us to discard regions that were likely invalid or biologically implausible. This direct filtering procedure maximized the retention of well partitioned iPSCs cell regions across multiple staining channels, thereby providing high quality input for subsequent machine learning-based feature analysis. Our strategy deliberately sacrifices a fraction of extractable cells in exchange for improved segmentation accuracy, which is critical for validating the mechanistic roles of iPSC-associated diseases. Representative segmentation results are illustrated in Figure 5 , showing successful delineation of individual cells in both TMRM (mitochondrial membrane potential) and LysoTracker (lysosomal) channels. The Voronoi-based approach effectively partitioned cells even in regions with weak membrane contrast, enabling reliable single-cell feature extraction. Figure 5. Open in a new tab The images of cell segmentation in R693Q iPSCs and WT iPSCs (A) Cell segmentation based on TMRM channel. (B) Cell segmentation based on Lyso staining. Scale bars, 100 μm. These design choices enabled us to accurately delineate cellular regions in complex and poorly differentiated cell samples, thereby providing essential data support for downstream analyses and for validating the potential mechanisms of RHMs. Through this optimization, we ensured both the reliability and applicability of the segmentation outcomes. Feature extraction and result analysis To comprehensively investigate potential inter-organelle dysfunction associated with NLRP7 mutations in RHM, we applied the aforementioned cell segmentation method, supported by the large number of cells provided by HCI. By comparing multiple features across fluorescence channels of different organelles between mutant and control groups, this approach not only suggested physical and biochemical alterations within cells but also provided preliminary insights into molecular pathways potentially implicated in the pathogenic effects of the mutation. We further extracted morphological, fluorescence intensity, and textural features from iPSCs. Leveraging the capacity of these features to indicate and quantify organelle dysfunction and cellular state changes, we sought to explore the mechanisms by which NLRP7 mutations may contribute to the pathogenesis of RHM. To further ensure optimal classification performance across distinct imaging channels, we systematically evaluated multiple machine learning algorithms and selected the most suitable model for each fluorescence channel based on accuracy maximization through 10-fold cross-validation. The TMRM and Caspase-3 channels were best modeled by CatBoost, leveraging its robustness against categorical feature distributions and strong performance in handling high-dimensional tabular data. For the MitoTracker channel, standard LightGBM achieved the most stable performance, while the JC-1 channel required the Light GBMXT variant to account for heterogeneous fluorescence distributions. The EdU proliferation channel was most effectively captured by a random forest classifier utilizing the Gini index for split criteria, reflecting its capability to balance interpretability and accuracy in proliferation-associated morphological features. Finally, the LysoTracker channel was optimally modeled by CatBoost, consistent with its strong performance on lysosomal feature textures. Hyperparameters for all models were rigorously optimized through nested grid search within the cross-validation framework, ensuring that the final model selection in each channel represented the highest attainable classification accuracy under standardized conditions. As shown in Table S2 , the classification accuracies obtained from binary training across different fluorescence channels indicate that the TMRM, MitoTracker, and JC-1 channels exhibit broadly comparable performance, and overall higher accuracy than the LysoTracker channel, which represents lysosomal activity. This suggests that mitochondrial impairment may occur further downstream within the pathogenic cascade. More results regarding different channel classification models can be found in Figures S7–S13 . In the ROC curve analysis, distinct classification accuracies were observed across features from different fluorescence channels, indicating differential discriminatory capacity for mutant cells ( Figure 6 A). Notably, the EDU and Caspase-3 channels, representing proliferation and apoptosis, exhibited relatively high predictive ability. However, their primary role lies in assessing the overall physiological impact of mutations on cell proliferation and apoptosis, rather than directly delineating the pathological processes or molecular pathways involved. By contrast, features of mitochondrial membrane potential (captured by JC-1, TMRM, and Mito SOX channels) played a central role in distinguishing mutant from control groups. Their ROC curves even outperformed those of the apoptosis-related Caspase-3 channel, further supporting the critical role of mitochondrial dysfunction in embryonic developmental arrest. Figure 6. Open in a new tab Machine learning model performance evaluation and interpretability analysis (A) ROC curves for different channel models. (B) SHAP analysis plots for different channel models. Through SHAP analysis ( Figure 6 B), we interpreted the contribution of model features and observed that the key determinants varied across different imaging channels. Although fluorescence intensity retained relatively high SHAP values in most channels, thereby validating the relevance of prior biological knowledge in cell image analysis, notable exceptions were observed at the mitochondrial level. In particular, in the JC-1 red fluorescence channel, SHAP analysis highlighted isolated brightspot features rather than homogeneous aggregates. Such a pattern may reflect residual mitochondria retaining high membrane potential within an otherwise depolarized population, consistent with mitochondrial heterogeneity and potential fragmentation under pathological conditions. Moreover, the marked elevation of MitoSOX fluorescence intensity in the mutant group corresponded to the pathological accumulation of mitochondrial superoxide, with its SHAP values supporting BLABLA as a core phenotype of mutant cells. In addition, SHAP analysis of Caspase-3 fluorescence signals highlighted their strong contribution to apoptosis detection, thereby providing clear evidence for the activation of apoptotic pathways in mutant cells. Although fluorescence features overall exhibited higher importance, texture features also played a critical role in capturing the structural complexity of cells. Beyond the aforementioned JC-1 red fluorescence isolated bright spot counts, mutant cells were characterized by an increased number of densely distributed Caspase-3 signal regions (lbp8), which was associated with lower Shannon entropy of fluorescence texture and may reflect apoptotic activation. Likewise, higher homogeneity in TMRM texture was more strongly associated with mitochondrial dysfunction, potentially reflecting mitochondrial fragmentation and loss of polarization. In addition, an increased number of isolated dark spots (lbp6) in the LysoTracker channel was suggestive of more frequent lysosomal impairment, consistent with pathological cell states. Collectively, these features not only support the reliability of machine learning-based classification of diseased cells when corroborated by expert knowledge, but also underscore the significance of texture analysis beyond fluorescence intensity after cell segmentation in revealing underlying disease mechanisms. In the multichannel analysis combining LysoTracker and MitoTracker staining, the model not only exhibited high classification performance but also revealed a notable correlation between their texture features. As shown in Figure 7 , the gray level gradient linear trend feature of the LysoTracker channel displayed a strong Spearman correlation with the same feature in the MitoTracker channel, and the two features shared the highest mutual information values among all cross-channel comparisons. These findings suggest that the disruption of lysosome-mitochondria interactions may represent a critical mechanism underlying impaired cellular homeostasis and potentially contribute to embryonic developmental failure. Figure 7. Open in a new tab Correlation and mutual information analysis of mitochondria and lysosome features (A) Heatmap of Spearman feature correlations for multichannel images co-stained with LysoTracker and MitoTracker. (B) The 10 features with the highest mutual information for MitoTracker gray gradient linear trend features. Based on this, to explore the causal relationship between the mixed channel features of MitoTracker and LysoTracker, on one hand, since the classification accuracy of the individual Mitotracker channel was higher compared to LysoTracker, we initially hypothesized that the MitoTracker channel is closer to the downstream outcome of cellular lesions. This suggests that the MitoTracker channel captures features more directly relevant to end-stage pathological outcomes, as downstream variables often manifest more specific impacts aligned with the final pathological state. Upstream factors, like those indicated by the LysoTracker, may incorporate intermediate processes or unaccounted downstream variables, which can dilute classification accuracy despite their role in initiating or propagating the pathological process. Furthermore, for the two texture features, Mito_glcm_correlation and Lyso_glcm_correlation, which exhibited high mutual information and correlation, we applied the double machine learning (DML) method to conduct a bidirectional causal analysis. The results showed that the average treatment effect (ATE) of Lyso_glcm_correlation on Mito_glcm_correlation and vice versa were 0.185 and 0.15, respectively. Considering the feedback loop between mitochondria and lysosomes, these findings provide some support for our hypothesis that mitochondrial damage occurs at a more downstream stage of the pathological process. Transcriptome sequencing analysis of R693Q iPSCs and WT iPSCs To elucidate the molecular mechanisms underlying the cellular phenotypes observed in R693Q iPSCs and machine learning prediction, we conducted comprehensive RNA-Seq analysis comparing genome-wide gene expression profiles between patient-derived R693Q iPSCs and WT iPSCs. Total RNA was extracted from three biologically independent replicates of each cell line, followed by library preparation and sequencing on the Illumina platform. Quality control assessment of the raw sequencing data revealed high quality reads with Q30 scores exceeding 95% for all samples. The cleaned reads were subsequently aligned to the human reference genome (GRCh38) using STAR aligner, and gene expression quantification was performed using feature counts. Differential expression analysis using DESeq2 identified a substantial number of genes dysregulated in R693Q iPSCs relative to WT iPSCs. At the gene expression level, 630 genes are uniquely expressed in R693Q iPSCs, while 933 genes are uniquely expressed in WT control ( Figure 8 A). Examination of sequencing error rate distribution in R693Q iPSCs and WT iPSCs were shown in Figure S14 . Differential expression analysis between the R693Q iPSCs and control groups identified 1,095 differentially expressed genes, of which 492 were upregulated and 603 downregulated ( Figure 8 B). Analysis of A/T/G/C content distribution was shown in Figure S15 . Furthermore, boxplot of gene expression distribution was shown in Figure S16 , indicating the difference of two groups within gene expression. Gene ontology (GO) enrichment analysis revealed significant functional categorizations among the DEGs ( Figure 8 C). The most significantly enriched biological processes (BP) were associated with “homophilic cell adhesion via plasma.” Cellular component (CC) terms were highly enriched for “collagen-containing extracellular matrix,” “postsynaptic membrane.” Molecular function (MF) categories included “extracellular matrix structural constituent,” “heparin binding,” “cytokine activity.” In addition, the significantly upregulated functional enrichment in GO functional enrichment analysis included embryonic organ development, embryonic organ morphogenesis. The significantly downregulated GO functions were enriched in regulation of cell killing and cytokine receptor binding ( Figures S19 and S20 ). Figure 8. Open in a new tab Transcriptome sequencing analysis of R693Q iPSCs and WT iPSCs (A) Venn diagram of co-expressed genes between R693Q iPSCs and WT iPSCs. (B) Volcano plot of differentially expressed genes between R693Q iPSCs and WT iPSCs. (C) Bar chart of GO enrichment analysis between R693Q iPSCs and WT iPSCs. (D) Heatmap of differentially expressed genes in R693Q iPSCs and WT iPSCs groups. (E) KEGG pathway enrichment bubble plot of R693Q iPSCs and WT iPSCs groups (the top 20 most significant KEGG pathways are displayed). Unsupervised hierarchical clustering of the DEGs clearly distinguished R693Q from WT iPSCs samples, demonstrating the profound impact of the NLRP7 R693Q mutation on the transcriptional landscape ( Figures 8 D and S19 – S21 ). The heatmap visualization revealed consistent expression patterns across biological replicates, indicating high reproducibility of our transcriptomic data. Notably, genes involved in inflammatory responses formed distinct clusters with elevated expression in R693Q iPSCs, while genes related to mitochondrial energy metabolism showed coordinated downregulation. KEGG pathway analysis further illuminated the biological pathways affected by the mutation. Figure 8 E shows the top 20 significant pathways. According to Figures S17 and S18 , the “MAPK signaling pathway” is upregulated. The “cytokine-cytokine receptor interaction,” “ECM-receptor interaction,” and “calcium signaling pathway” pathways are downregulated. 40 , 41 The regulation of trophoblast invasion and maternal-fetal interface adhesion is critical for normal placentation, and its dysregulation directly contributes to the hyperinvasive phenotype characteristic of hydatidiform moles. Key adhesion receptors, such as integrins and selectins, rely on mitochondrial-derived ATP for energy and lysosomal proteases (e.g., cathepsins) for extracellular matrix degradation. Notably, NLRP7 mutations may disrupt this finely tuned process by promoting inflammatory signaling, which interferes with the end lysosomal trafficking and recycling of adhesion receptors—a mechanism essential for maintaining controlled invasion. Furthermore, the core proliferation and stress pathways, activated by IGF2 and NLRP7-associated inflammatory factors, drive abnormal trophoblast proliferation. These pathways regulate mitochondrial apoptosis by activating JNK/p38 to induce immediate mitochondrial permeabilization and lysosomal autophagy through ERK/p38-dependent phosphorylation of autophagy-related proteins. Mitochondrial ROS further amplifies MAPK signaling, creating a feedforward loop of cellular stress. 42 Additionally, NLRP7 mutation disrupts the immune microenvironment at the maternal-fetal interface, leading to an imbalance of proinflammatory cytokines such as IL-1β and TNF-α, which further promotes aberrant trophoblast expansion. Cytokine signaling also modulates mitochondrial metabolic adaptation and lysosomal functions including antigen presentation and autophagic degradation. These processes exhibit extensive crosstalk with MAPK and PI3K pathways, synergistically amplifying a vicious cycle of inflammation and proliferation. 43 The molecular pathogenesis involves a coordinated cascade beginning with enhanced cellular adhesion and ECM degradation (promoting invasion), which subsequently activates MAPK/PI3K signaling pathways to drive aberrant proliferation. This is further modulated by calcium-mediated signaling that coordinates inter-organellar communication, ultimately shaping a proinflammatory cytokine microenvironment. These processes converge on mitochondrial and lysosomal executioners regulating energy supply, degradation capacity, and apoptotic commitment a mechanistic framework strongly supported by ML-based functional predictions. Discussion This study establishes a critical link between NLRP7 mutations and dysregulated mitochondrial-lysosomal crosstalk in the pathogenesis of RHM. Using a patient-specific iPSCs model combined with AI-enhanced HCI, we demonstrated that the R693Q mutation disrupts organellar homeostasis, leading to metabolic stress, ROS accumulation, and ultimately trophoblast dysfunction. Our findings reveal that NLRP7 deficiency impairs the coordinated functioning of mitochondria and lysosomes—two essential organelles for cellular energy management and quality control. The observed disruption in their spatial and functional interaction provides a mechanistic explanation for the excessive trophoblast invasion characteristic of RHM. Furthermore, transcriptomic data confirm the activation of inflammatory and stress-response pathways, creating a self-perpetuating cycle of cellular damage. The integration of AI-driven image analysis with traditional molecular approaches represents a significant methodological advance, enabling precise quantification of subtle organellar dynamics that would otherwise remain undetected. This multi-modal framework not only validates the pathological role of NLRP7 mutations but also establishes a platform for investigating other human-specific reproductive disorders. These insights open avenues for therapeutic intervention, suggesting that strategies targeting mitochondrial-lysosomal coordination or mitigating oxidative stress may help prevent disease progression. Future studies will focus on validating these targets and exploring their translational potential in clinical settings. Limitations of the study This study also has several limitations. First, given the rarity of RHMs, future studies should include additional NLRP7 variants and larger patient cohorts to ensure generalizability. Second, while our AI model achieved high classification accuracy, its interpretability remains partially dependent on feature engineering; incorporating explainable AI techniques could further elucidate decision pathways. Third, the in vitro iPSC model, though informative, cannot fully recapitulate the dynamic maternal-fetal microenvironment in vivo . Validation using trophoblast-derived organoids or blastocyst-like models would strengthen the pathological relevance. Conclusion In this study, we contrasted a bioinspired AI framework (BioVision-Segmentation), utilizing a hybrid Transformer-Voronoi architecture for multichannel cell segmentation though HCI. This study demonstrates that the NLRP7 R693Q mutation drives RHMs by disrupting mitochondrial-lysosomal function, amplifying inflammatory signaling, and impairing trophoblast regulation. Using patient-derived iPSCs, we established a humanized model that revealed aberrant organelle dynamics, activation of pro-apoptotic and stress pathways, and dysregulation of immune metabolic crosstalk. These findings provide mechanistic insight into NLRP7-related gestational disorders and underscore the value of iPSCs-based models in studying human-specific disease mechanisms. Resource availability Lead contact Requests for further information and resources should be directed to and will be fulfilled by the lead contact, Jianhua Qian ( qianjianhua@zju.edu.cn ). Materials availability This study did not generate new unique reagents. Data and code availability • The raw transcriptomics data of the RNA-Seq supporting the results of this article have been deposited at the GEO database (GEO: GSE309480 ) and are publicly available at: https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE309480 . • All original code is available in this paper’s supplemental information. • Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request. Acknowledgments This study was financially supported by the National Natural Science Foundation of China (no. 82371669), the National Natural Science Foundation of China (no. 82071665), and the Zhejiang Provincial Key Research and Development Program (no. 2025C02115). Author contributions Conceptualization, J.W. and J.Q.; methodology, J.W., T.L., and L.C.; investigation, J.W., T.L., L.C., and Y.Z.; writing – original draft, J.W.; writing – review and editing, J.W., T.L., L.C., Y.Z., and J.Q.; funding acquisition, J.Q.; resources, J.Q.; supervision, J.Q. Declaration of interests The authors declare no competing interests. STAR★Methods Key resources table REAGENT or RESOURCE SOURCE IDENTIFIER Antibodies Cleaved Caspase-3 Rabbit mAb Proteintech Cat# 25128-1-AP; RRID: AB_3073913 Goat Anti-Rabbit IgG (H + L), FITC Abcam Cat# ab6717; RRID: AB_955238 Chemicals, peptides, and recombinant proteins MitoTracker™ Green Thermo Fisher Scientific Cat# M22426 LysoTracker™ Deep Red Thermo Fisher Scientific Cat# L12492 Tetramethylrhodamine, methyl ester (TMRM) Thermo Fisher Scientific Cat# T668 MitoSOX™ Red Mitochondrial Superoxide Indicator Thermo Fisher Scientific Cat# M36008 JC-1 Thermo Fisher Scientific Cat# T3168 Calcein-AM Thermo Fisher Scientific Cat# L3224 Hoechst 33342 Thermo Fisher Scientific Cat# H3570 Critical commercial assays Click-iT™ EdU Alexa Fluor™ 488 Imaging Kit Thermo Fisher Scientific Cat# C10337 FastPure Cell/Tissue DNA Isolation Mini Kit Vazyme Cat# DC102-01 Deposited data Raw RNA-seq data This paper GSE309480 Experimental models: Cell lines Human induced pluripotent stem cells (iPSCs), NLRP7 R693Q homozygous mutant This paper hPSCreg: FAHZJUi001-A Software and algorithms ImageXpress® Micro Confocal High-Content Imaging System Molecular Devices LLC ImageXpress Micro Confocal MetaXpress® Software Molecular Devices LLC Version 7.0 Open in a new tab Experimental model and study participant details Cell lines Cell source and informed consent The induced pluripotent stem cells (iPSCs) line used in this study was derived and characterized from a patient with familial recurrent hydatidiform mole (FRHMs) harboring a homozygous NLRP7 gene mutation (c.2179 G > A; p. Arg693Gln). This FRHMs-iPSCs line underwent validation including mutation analysis, pluripotency characterization, and trilineage differentiation potential. The cell line is registered in the hPSCreg database ( https://hpscreg.eu/ ) under the identifier FAHZJUi001-A. The study protocol was approved by the Review Board of the First Affiliated Hospital, Zhejiang University School of Medicine (Approval No: 20240743). Written informed consent was obtained from the patient. Relevant clinical information was sourced from outpatient and inpatient medical records, supplemented by direct patient interviews. Culture of iPSCs lines iPSCs were cultured in mTeSR1 medium (STEMCELL Technologies, #85850) at 37°C under 5% CO 2 . The medium was replaced daily. For passaging, cells were dissociated using Versene (Gibco, #15040066) and split at a ratio of 1:8 onto plates pre-coated with Matrigel (Corning, #354277). Following each passage, the culture medium was supplemented with 10 μM of the ROCK inhibitor Y-27632 (MedChemExpress, HY-10071) for approximately 24 h. Subsequently, the medium was replaced with fresh mTeSR1 lacking Y-27632. Method details Generated iPSC lines from peripheral blood mononuclear cells Freshly drawn blood specimens were obtained from participating subjects, and peripheral blood mononuclear cells (PBMCs) were separated using Lymphoprep (STEMCELL Technologies, 07801). A total of 5×10 5 cells were subsequently cultured for one week in an erythroid expansion medium, which consisted of StemSpan SFEM II medium (Stem Cell Technologies, 09605) supplemented with 1X StemSpan Erythroid Expansion Supplement (Stem Cell Technologies, 02692). Expanded erythroid progenitor cells were transfected via electroporation using the P3 Primary Cell 4D-Nucleofector X Kit (Lonza V4XP-3012) on a 4D-Nucleofector system (Lonza) under program EO115. Each sample received 1 μg of OriP/EBNA-1–based episomal plasmids (pCXLE-hOCT3/4, pCXLE-hSK, pCXLE-hUL). Post-electroporation, cells were maintained for 4 h in RPMI 1640 medium (Gibco, 22400089) containing 20% fetal bovine serum (ExCell Bio, FSD500), after which they were transferred to wells of a 6-well plate pre-coated with 3% Geltrex (Gibco, A1413301) and cultured in erythroid expansion medium for the initial 72 h. From day 4 to day 8, 1 mL of ReproTeSR medium (STEMCELL Technologies, 05920) was added every two days. Starting on day 9, the medium was fully replaced with ReproTeSR daily until colony formation was visible microscopically. Thereafter, the culture medium was switched to mTeSR1 (STEMCELL Technologies, 05850) and refreshed daily. Approximately three weeks later, iPSC colonies were manually isolated. Specifically, approximately 8 iPSC clones were generated from R693Q PBMCs. Approximately 12 iPSC clones were generated from WT PBMCs. Based on their typical iPSC morphology and robust growth ability, three clones were randomly selected from each group for subsequent experiments. The iPSC markers OCT3/4, SOX2, and NANOG were also detected to confirm their identity. Genomic DNA extraction and sanger sequencing Genomic DNA was isolated from the iPSCs using the FastPure Cell/Tissue DNA Isolation Mini Kit (Vazyme, #DC102-01). The region encompassing the NLRP7 mutation was amplified by PCR using specific primers (see Table S1 ). Sanger sequencing and chromatogram analysis were performed by Tsingke Biotech Co., Ltd. EdU proliferation assay iPSCs were seeded in 6-well plates. After 24 h, pre-warmed 2X EdU working solution (20 μM EdU in culture medium) was added in equal volume to the existing culture medium, resulting in a final 1X EdU concentration (10 μM). Cells were incubated for 2 h at 37°C with 5% CO 2 . Following EdU labeling, the medium was removed, and cells were fixed with 4% paraformaldehyde (PFA) for 15 min at room temperature. Fixative was aspirated, and cells were washed thrice with 1 mL wash buffer 3% BSA in PBS) per well for 5 min. Cells were permeabilized with 1 mL permeabilization buffer (0.5% Triton X-100 in PBS) per well for 15 min at room temperature. Then, permeabilization buffer was removed, and cells were washed twice with wash buffer. The Click-iT reaction cocktail was prepared according to the manufacturer’s instructions (Click-iT EdU Alexa Fluor 488 Imaging Kit, Thermo Fisher Scientific, C10337 ). Briefly, for the C0071S kit component, one vial of click additive was dissolved in 1.3 mL deionized water to prepare the Click Additive Solution. The complete reaction cocktail (containing Alexa Fluor 488, CuSO 4 , reaction buffer, and Click Additive Solution) was added to each well and incubated for 30 min at room temperature, protected from light. The reaction cocktail was aspirated, and cells were washed thrice with wash buffer. Cells were counterstained with Hoechst 33342 (5 μg/mL in PBS) for 10 min at room temperature, protected from light, followed by three washes with PBS. HCI was performed immediately using fluorescence channels for EdU (Alexa Fluor 488; excitation: 495 nm, emission: 519 nm) and Hoechst 33342 (excitation: 405 nm, emission: 460 nm). Exposure time, light source power, and focal height were optimized. Images were acquired from selected wells and fields of view and exported for analysis. Cleaved caspase-3 immunofluorescence iPSCs were seeded in 24-well plates at 1 x 10 5 cells per well and cultured for 24 h. Culture medium was aspirated, and cells were washed thrice with PBS. Cells were fixed with 4% PFA for 15 min at room temperature, washed thrice with PBS, and permeabilized with 0.5% Triton X-100 in PBS for 15 min. After three PBS washes, cells were blocked with blocking buffer (e.g., 5% normal goat serum in PBS) for 1 h at room temperature. Cells were incubated overnight at 4°C with primary antibody against cleaved Caspase-3 (Proteintech, 25128-1-AP, Rabbit, 1:200 dilution in blocking buffer). The primary antibody solution was aspirated, and cells were washed thrice with PBS (5 min per wash). Cells were incubated for 1 h at room temperature, protected from light, with FITC-conjugated secondary antibody (e.g., Goat Anti-Rabbit IgG (H + L), FITC, Abcam, ab6717; 1:500 dilution in blocking buffer). The secondary antibody solution was aspirated, and cells were washed thrice with PBS. Nuclei were counterstained with Hoechst 33342 (5 μg/mL in PBS) for 10 min at room temperature, protected from light, followed by three final PBS washes. HCI was performed immediately using fluorescence channels for FITC (excitation: 494 nm, emission: 518 nm) and Hoechst 33342. Instrument settings (exposure time, power, focal height) were optimized, images were acquired, and exported. Mitochondrial staining (MitoTracker green) iPSCs were seeded in 24-well plates at a density of 1 x 10 5 cells per well and cultured for 24 h. The culture medium (mTeSR1) was aspirated and replaced with fresh mTeSR1 containing 100 nM MitoTracker Green (Thermo Fisher Scientific, M22426 ). Cells were incubated at 37°C with 5% CO 2 , protected from light, for 20 min. The staining solution was then aspirated, and cells were washed three times with pre-warmed phosphate-buffered saline (PBS) under light-protected conditions. Subsequently, cells were incubated with mTeSR1 containing Hoechst 33342 (Thermo Fisher Scientific, H3570; 1:1000 dilution) for 10 min, protected from light. Following three additional PBS washes, HCI was performed immediately. Image acquisition settings included fluorescence channels for MitoTracker Green (excitation: 490 nm, emission: 516 nm) and Hoechst 33342 (excitation: 405 nm, emission: 460 nm). Exposure time, light source power, and focal height were optimized. Wells and fields of view were selected, and images were acquired using the HCI system. Acquired images were exported for analysis. Mitochondrial membrane potential staining (TMRM) iPSCs were seeded after 24 h, medium was aspirated and replaced with mTeSR1 containing 50 nM TMRM (Thermo Fisher Scientific, T668). Incubation proceeded at 37°C with 5% CO 2 , protected from light, for 30 min. The staining solution was aspirated, and cells were washed three times with PBS (protected from light). Cells were then stained with Hoechst 33342 (as in 1.1.1) for 10 min, followed by three PBS washes. HCI was performed immediately using channels for TMRM (excitation: 552 nm, emission: 575 nm) and Hoechst 33342. Instrument settings (exposure time, power, focal height) were adjusted, images were acquired and exported. Cell viability staining iPSCs were seeded as described. After 24 h, medium was replaced with mTeSR1 containing a 1 μM working solution of Calcein-AM (Thermo Fisher Scientific, L3224). Cells were incubated at 37°C with 5% CO 2 , protected from light, for 15 min. The staining solution was aspirated, and cells were washed three times with PBS (protected from light). HCI was performed immediately using channels for Calcein-AM (excitation: 494 nm, emission: 517 nm). Instrument settings were optimized, images were acquired and exported. Lysosomal staining (LysoTracker deep red) iPSCs were seeded as described. After 24 h, medium was replaced with mTeSR1 containing 100 nM LysoTracker Deep Red (Thermo Fisher Scientific, L12492 ). Cells were incubated at 37°C with 5% CO 2 , protected from light, for 20 min. The staining solution was aspirated, and cells were washed three times with PBS (protected from light). Cells were then counterstained with Hoechst 33342 (as in 1.1.1) for 10 min, followed by three PBS washes. HCI was performed immediately using channels for LysoTracker Deep Red (excitation: 577 nm, emission: 590 nm; Note: Verify instrument compatibility for UV excitation) and Hoechst 33342. Instrument settings were optimized, images were acquired and exported. Mitochondrial membrane potential staining (JC-1) iPSCs were seeded as described. After 24 h, medium was replaced with mTeSR1 containing JC-1 (10 μg/mL; Thermo Fisher Scientific, T3168) diluted from a DMSO stock solution. Cells were incubated at 37°C with 5% CO 2 , protected from light, for 20 min. The staining solution was aspirated, and cells were washed three times with PBS (protected from light). HCI was performed immediately. Image acquisition utilized two distinct channels: JC-1 monomers (excitation: 514 nm, emission: 529 nm) and JC-1 aggregates (J-aggregates; excitation: 580 nm, emission: 595 nm). Instrument settings (exposure time, power, focal height) were adjusted, images were acquired and exported. The ratio of J-aggregate (red) to monomer (green) fluorescence intensity was used to assess mitochondrial membrane potential. Mitochondria-lysosome co-staining iPSCs were seeded as described. After 24 h, medium was replaced with mTeSR1 containing 100 nM MitoTracker Green. Cells were incubated at 37°C with 5% CO 2 , protected from light, for 20 min. The staining solution was aspirated, and cells were washed three times with PBS (protected from light). Subsequently, cells were incubated with mTeSR1 containing 100 nM LysoTracker Deep Red at 37°C with 5% CO 2 , protected from light, for 20 min. The staining solution was aspirated, and cells were washed three times with PBS (protected from light). HCI was performed immediately using channels for MitoTracker Green (excitation: 490 nm, emission: 516 nm) and LysoTracker Deep Red (excitation: 577 nm, emission: 590 nm). Instrument settings were optimized, images were acquired and exported for co-localization analysis. High-content imaging acquisition Cell phenotype imaging was performed using the ImageXpress Micro Confocal High-Content Imaging System (Molecular Devices LLC, San Jose, CA, USA). This laser-based confocal system was employed for all quantitative analyses of cellular phenotypes, including viability (Calcein-AM), proliferation (EdU), mitochondrial membrane potential (TMRM, JC-1), mitochondrial mass/distribution (Mito Tracker), lysosomal distribution (Lyso Tracker), mitochondria-lysosome co-localization, and apoptosis (cleaved Caspase-3). Images were acquired using a 20x or 40× objective. For each well, a minimum of 15 non-overlapping fields of view were captured to ensure statistical robustness. Acquired multi-channel images were analyzed using the MetaXpress Software Package (Version 7.0, Molecular Devices). Specific application modules (Multi Wavelength Cell Scoring, Mitochondrial Potential, Translocation, Co-localization) were utilized for feature segmentation and quantification based on predefined fluorescence parameters and morphological criteria. For all binary classification tasks, cross-entropy was adopted as the loss function to maintain consistent optimization criteria. Specifically, the mixed-channel dataset (Mito Tracker + Lyso Tracker) was modeled using LightGBM with an extended tree (LightGBMXT) framework, which demonstrated superior capacity to capture complex inter-channel feature dependencies. RNA transcriptome sequencing analysis Quantification of gene expression level Gene expression quantification was performed using featureCounts (v2.0.6) to assign reads to genomic features. Gene expression levels were subsequently normalized and reported as FPKM (Fragments Per Kilobase of transcript per Million mapped reads), which accounts for both sequencing depth and gene length, providing a standardized measure for cross-comparison of transcript abundance. Differential expression analysis Differential expression analysis between experimental conditions was conducted using DESeq2 (v1.42.0) for datasets with biological replicates. This method employs a negative binomial generalized linear model to test for statistical significance of expression changes, with multiple testing correction performed using the Benjamini-Hochberg procedure to control the false discovery rate. Genes meeting the threshold of adjusted p -value (padj) ≤ 0.05 and absolute log2 fold change ≥1 were considered significantly differentially expressed. For analyses without biological replicates, the edgeR package (v4.0.16) was implemented. Read counts were first normalized using scaling factors to account for variations in library sizes, followed by differential expression testing. The resulting p -values were adjusted using the Benjamini-Hochberg method, with the same significance thresholds applied. GO and KEGG enrichment analysis of differentially expressed genes Functional enrichment analysis of differentially expressed genes was performed through Gene Ontology (GO) and KEGG pathway analyses using the clusterProfiler package (v4.8.1). GO analysis included correction for gene length bias, with terms showing corrected p -values <0.05 deemed significantly enriched. KEGG pathway analysis examined the representation of differentially expressed genes in curated biological pathways from the Kyoto Encyclopedia of Genes and Genomes database, with statistical significance determined using hypergeometric testing followed by multiple test correction. Quantification and statistical analysis All statistical tests were performed using GraphPad Prism 9 or R (v4.2.1). Data are presented as mean ± standard deviation (SD) unless otherwise stated. Comparisons between two groups were evaluated using two-tailed unpaired Student’s t test when normality was confirmed (Shapiro–Wilk test); otherwise, the Mann–Whitney U test was applied. For multiple comparisons, one-way ANOVA with Tukey’s post hoc test was used. A p -value <0.05 was considered statistically significant. Statistical significance was defined as ∗ p < 0.05, ∗∗ p < 0.01, ∗∗∗ p < 0.001. Published: March 26, 2026 Footnotes Supplemental information can be found online at https://doi.org/10.1016/j.isci.2026.115479 . Supplemental information Document S1. Figures S1–S21 and Tables S1 and S2 mmc1.pdf (5.1MB, pdf) Data S1. Original code for high-content imaging analysis mmc2.zip (161.4KB, zip) References 1. Fisher R.A., Hodges M.D., Newlands E.S. Familial recurrent hydatidiform mole - A review. J. Reprod. Med. 2004;49:595–601. [ PubMed ] [ Google Scholar ] 2. Gonzalez J., Popp M., Ocejo S., Abreu A., Bahmad H.F., Poppiti R. Gestational Trophoblastic Disease: Complete versus Partial Hydatidiform Moles. Diseases. 2024;12:159. doi: 10.3390/diseases12070159. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 3. Chen Z., Jiang L., Su M., Zeng Q., Luo P., Chu L. NLRP7 maintains the genomic stability during early human embryogenesis via mediating alternative splicing. Commun. Biol. 2025;8:125. doi: 10.1038/s42003-025-07571-5. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 4. Riccio S., Galanti F., Scudo M., Di Troia L., Ferrillo M.G., Manzara F., Ianiri P., Battaglia F.A. Recurrent Hydatidiform Moles: A Clinical Challenge-A Case Report and an Update on Management and Therapeutical Strategies. Case Rep. Obstet. Gynecol. 2023;2023 doi: 10.1155/2023/3752274. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 5. Abi Nahed R., Elkhoury Mikhael M., Reynaud D., Collet C., Lemaitre N., Michy T., Hoffmann P., Sergent F., Marquette C., Murthi P., et al. Role of NLRP7 in Normal and Malignant Trophoblast Cells. Biomedicines. 2022;10:252. doi: 10.3390/biomedicines10020252. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 6. Allias F., Mechtouf N., Gaillot-Durand L., Hoffner L., Hajri T., Devouassoux-Shisheboran M., Massardier J., Golfier F., Bolze P.A., Surti U., Slim R. A novel NLRP7 protein-truncating mutation associated with discordant and divergent p57 immunostaining in diploid biparental and triploid digynic moles. Virchows Arch. 2020;477:309–315. doi: 10.1007/s00428-020-02769-w. [ DOI ] [ PubMed ] [ Google Scholar ] 7. Echeverría-Altamar K., Barreto-Gamarra C., Domenech-García M., Resto-Irizarry P. Prediction of cardiac differentiation in human induced pluripotent stem cell-derived cardiomyocyte supernatant using surface-enhanced Raman spectroscopy and machine learning. Biosens. Bioelectron. 2025;283 doi: 10.1016/j.bios.2025.117528. [ DOI ] [ PubMed ] [ Google Scholar ] 8. Wang J., Huang S., Li K., Li J., Huang X. Global research dynamics in the induced pluripotent stem cell and diabetes: A bibliometric analysis of the past twenty years. Regen. Ther. 2025;30:341–350. doi: 10.1016/j.reth.2025.06.015. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 9. Hockemeyer D., Soldner F., Beard C., Gao Q., Mitalipova M., DeKelver R.C., Katibah G.E., Amora R., Boydston E.A., Zeitler B., et al. Efficient targeting of expressed and silent genes in human ESCs and iPSCs using zinc-finger nucleases. Nat. Biotechnol. 2009;27:851–857. doi: 10.1038/nbt.1562. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 10. Tomizawa S.i., Sasaki H. Genomic imprinting and its relevance to congenital disease, infertility, molar pregnancy and induced pluripotent stem cell. J. Hum. Genet. 2012;57:84–91. doi: 10.1038/jhg.2011.151. [ DOI ] [ PubMed ] [ Google Scholar ] 11. Divvela S.S.K., Gallorini M., Gellisch M., Patel G.D., Saso L., Brand-Saberi B. Navigating redox imbalance: the role of oxidative stress in embryonic development and long-term health outcomes. Front. Cell Dev. Biol. 2025;13 doi: 10.3389/fcell.2025.1521336. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 12. Fernandes R., Tsuda C., Perumalsamy A.L., Naranian T., Chong J., Acton B.M., Tong Z.-B., Nelson L.M., Jurisicova A. NLRP5 mediates mitochondrial function in mouse oocytes and embryos. Biol. Reprod. 2012;86:138. doi: 10.1095/biolreprod.111.093583. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 13. Gao Z., Zhang X., Yu X., Qin D., Xiao Y., Yu Y., Xiang Y., Nie X., Lu X., Liu W., et al. Zbed3 participates in the subcortical maternal complex and regulates the distribution of organelles. J. Mol. Cell Biol. 2018;10:74–88. doi: 10.1093/jmcb/mjx035. [ DOI ] [ PubMed ] [ Google Scholar ] 14. Qin D., Gao Z., Xiao Y., Zhang X., Ma H., Yu X., Nie X., Fan N., Wang X., Ouyang Y., et al. The subcortical maternal complex protein Nlrp4f is involved in cytoplasmic lattice formation and organelle distribution. Development. 2019;146 doi: 10.1242/dev.183616. [ DOI ] [ PubMed ] [ Google Scholar ] 15. Gong T., Wang H., Huang B., Wang H., Zhang P., Qian J. Reprogramming of human peripheral blood mononuclear cells from a patient suffering from recurrent hydatidiform mole to an iPSC line FAHZUi001-A carrying a homozygous p.Gln421Ter mutation in NLRP7 gene. Stem Cell Res. 2021;53 doi: 10.1016/j.scr.2021.102361. [ DOI ] [ PubMed ] [ Google Scholar ] 16. Alici-Garipcan A., Özçimen B., Süder I., Ülker V., Önder T.T., Özören N. NLRP7 plays a functional role in regulating BMP4 signaling during differentiation of patient-derived trophoblasts. Cell Death Dis. 2020;11:658. doi: 10.1038/s41419-020-02884-1. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 17. Gu A., Wang B., Zhang Y., Sun C., Wei Y., Yan J., Yang Y., Wang Y., Liu R., Zhang R., et al. SENP6 Restrains NLRP3 Inflammasome Activation via DeSUMOylation-Driven K48-Linked Ubiquitination of NLRP3 in Acute Lung Injury. Research (Wash D C) 2026;9:1069. doi: 10.34133/research.1069. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 18. Savini M., Zhao Q., Wang M.C. Lysosomes: Signaling Hubs for Metabolic Sensing and Longevity. Trends Cell Biol. 2019;29:876–887. doi: 10.1016/j.tcb.2019.08.008. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 19. Bednash J.S., Weathington N., Londino J., Rojas M., Gulick D.L., Fort R., Han S., McKelvey A.C., Chen B.B., Mallampalli R.K. Targeting the deubiquitinase STAMBP inhibits NALP7 inflammasome activity. Nat. Commun. 2017;8 doi: 10.1038/ncomms15203. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 20. Tietze L., Urbano L., Eisenmann S., Schwarzinger J., Kollan J., Forbes B., Dailey L.A., Hädrich G. High Content Image Analysis of Cellular Responses of the Murine J774A.1 Cell Line and Primary Human Cells Alveolar Macrophages to an Extended Panel of Pharmaceutical Agents. Pharm. Res. 2025;42:93–108. doi: 10.1007/s11095-024-03806-y. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 21. Behr N.J., Pierre S., Ickelsheimer T., Ziegler N., Luckhardt S., Kannt A., Pinter A., Geisslinger G., Schäfer S.M.G., König A., Scholich K. High content imaging shows distinct macrophage and dendritic cell phenotypes for psoriasis and atopic dermatitis. Sci. Rep. 2025;15 doi: 10.1038/s41598-025-99727-w. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 22. Seal S., Trapotsi M.-A., Spjuth O., Singh S., Carreras-Puigvert J., Greene N., Bender A., Carpenter A.E. Cell Painting: a decade of discovery and innovation in cellular imaging. Nat. Methods. 2025;22:254–268. doi: 10.1038/s41592-024-02528-8. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 23. Serrano E., Peters J., Wagner J., Graham R.E., Chen Z., Feng B., Miranda G., Kalinin A.A., Vulliard L., Tomkinson J., et al. Progress and new challenges in image-based profiling. arXiv. 2025 doi: 10.48550/arXiv.2508.05800. Preprint at. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 24. Iskandarani L., Bayen S., Hales B.F., Robaire B. High-content imaging and transcriptomic analyses of the effects of bisphenol S and alternative color developers on KGN granulosa cells. Toxicol. Sci. 2025;207:401–414. doi: 10.1093/toxsci/kfaf096. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 25. Herter S., Emperador M., Smyrilli K., Kocher D., Celikyürekli S., Zeiser C., Gerloff X., Kreth S., Henrich K.O., Maaß K.K., et al. High content-imaging drug synergy screening identifies specific senescence-related vulnerabilities of mesenchymal neuroblastomas. Cell Death Dis. 2025;16:644. doi: 10.1038/s41419-025-07933-1. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 26. Li T., Cao B., Su T., Lin L., Wang D., Liu X., Wan H., Ji H., He Z., Chen Y., et al. Machine Learning-Engineered Nanozyme System for Synergistic Anti-Tumor Ferroptosis/Apoptosis Therapy. Small. 2025;21 doi: 10.1002/smll.202408750. [ DOI ] [ PubMed ] [ Google Scholar ] 27. Li T., Cao B., Wang Y., Lin L., Chen L., Su T., Song H., Ren Y., Zhang L., Chen Y., et al. Interpretable Active Learning Identifies Iron-Doped Carbon Dots With High Photothermal Conversion Efficiency for Antitumor Synergistic Therapy. Aggregate. 2025;6 doi: 10.1002/agt2.70060. [ DOI ] [ Google Scholar ] 28. Li T., Chen L., Cao B., Liu S., Lin L., Li Z., Chen Y., Li Z., Zhang T.y., Feng L. Optimize the quantum yield of G-quartet-based circularly polarized luminescence materials via active learning strategy-BgoFace. Materials Genome Engineering Advances. 2025;3 doi: 10.1002/mgea.70031. [ DOI ] [ Google Scholar ] 29. Li H., Seada H., Madnick S., Zhao H., Chen Z., Li F., Zhu F., Hall S., Boekelheide K. Machine learning-assisted high-content imaging analysis of 3D MCF7 microtissues for estrogenic effect prediction. Sci. Rep. 2024;14:2999. doi: 10.1038/s41598-024-53323-6. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 30. Winfree S., McNutt A.T., Khochare S., Borgard T.J., Barwinska D., Sabo A.R., Ferkowicz M.J., Williams J.C., Jr., Lingeman J.E., Gulbronson C.J., et al. Integrated Cytometry With Machine Learning Applied to High-Content Imaging of Human Kidney Tissue for In Situ Cell Classification and Neighborhood Analysis. Lab. Invest. 2023;103 doi: 10.1016/j.labinv.2023.100104. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 31. Zhou Z., Wang J., Plott S., Gorkin D. Loss of the Mendelian disease gene BICRA alters the developmental potential of human iPSCs. J. Biol. Chem. 2025;301 doi: 10.1016/j.jbc.2025.108816. [ DOI ] [ Google Scholar ] 32. Bratosin D., Mitrofan L., Palii C., Estaquier J., Montreuil J. Novel fluorescence assay using calcein-AM for the determination of human erythrocyte viability and aging. Cytometry. A. 2005;66:78–84. doi: 10.1002/cyto.a.20152. [ DOI ] [ PubMed ] [ Google Scholar ] 33. Yamakoshi H., Dodo K., Okada M., Ando J., Palonpon A., Fujita K., Kawata S., Sodeoka M. Imaging of EdU, an Alkyne-Tagged Cell Proliferation Probe, by Raman Microscopy. J. Am. Chem. Soc. 2011;133:6102–6105. doi: 10.1021/ja108404p. [ DOI ] [ PubMed ] [ Google Scholar ] 34. Zhang Z., Lin X., Yang Y., Wang X., Wang Y., Huang X., Hong M., Gao W., He H., You M.J., et al. Caspase 3-specific cleavage of ubiquitin-specific peptidase 48 enhances drug-induced apoptosis in AML. Cancer Biol. Ther. 2025;26 doi: 10.1080/15384047.2025.2459426. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 35. Yu A.G., Wei X.L., Zito E., Zheng H., Zhong C.C., Gong Y.L., Luo Z. Oxidative stress and caspase 3/Gsdme-dependent pyroptosis contributes to high fat diet induced-intestinal inflammation and lipotoxicity via Srebp1 cleavage at D444 site by caspase 3. J. Nutr. Biochem. 2025;145 doi: 10.1016/j.jnutbio.2025.110033. [ DOI ] [ PubMed ] [ Google Scholar ] 36. Moore C.B., Bergstralh D.T., Duncan J.A., Lei Y., Morrison T.E., Zimmermann A.G., Accavitti-Loper M.A., Madden V.J., Sun L., Ye Z., et al. NLRX1 is a regulator of mitochondrial antiviral immunity. Nature. 2008;451:573–577. doi: 10.1038/nature06501. [ DOI ] [ PubMed ] [ Google Scholar ] 37. Kolusari A., Cebi A., Akgun E., Alp H.H., Bakan E. May supplementation of coenzyme Q10 help prevent development of hydatidiform mole? Clin. Exp. Obstet. Gynecol. 2017;44:398–402. doi: 10.12891/ceog3472.2017. [ DOI ] [ PubMed ] [ Google Scholar ] 38. Zhou M., Zheng M., Liang S., Li M., Ma J., Zhang S., Song X., Hu Y., Lyu Y., Ou X., Yue C. Inherent potential of mitochondria-targeted interventions for chronic neurodegenerative diseases. Neural Regen. Res. 2026;21:1409–1427. doi: 10.4103/nrr.Nrr-d-24-01507. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 39. Vasilev A., Sofi R., Smith S.J., Rahman R., Teschemacher A.G., Kasparov S. Feasibility of Photodynamic Therapy for Glioblastoma with the Mitochondria-Targeted Photosensitizer Tetramethylrhodamine Methyl Ester (TMRM) Biomedicines. 2021;9:1453. doi: 10.3390/biomedicines9101453. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 40. Xu Q., Lin Y., Yu L., Zhang Y., Kong N., Yan G., Sun H., Cao G. KHDC3 affects the lineage differentiation of early embryo by regulating the differential distribution of YAP signal. Theriogenology. 2025;245 doi: 10.1016/j.theriogenology.2025.117527. [ DOI ] [ PubMed ] [ Google Scholar ] 41. Reynaud D., Alfaidy N., Collet C., Lemaitre N., Sergent F., Miege C., Soleilhac E., Assi A.A., Murthi P., Courtois G., et al. NLRP7 Enhances Choriocarcinoma Cell Survival and Camouflage in an Inflammasome Independent Pathway. Cells. 2023;12:857. doi: 10.3390/cells12060857. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 42. Reynaud D., Abi Nahed R., Lemaitre N., Bolze P.A., Traboulsi W., Sergent F., Battail C., Filhol O., Sapin V., Boufettal H., et al. NLRP7 Promotes Choriocarcinoma Growth and Progression through the Establishment of an Immunosuppressive Microenvironment. Cancers. 2021;13:2999. doi: 10.3390/cancers13122999. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 43. Anvar Z., Jafarpour F., Jahromi B.N., Riccio A., Nasr-Esfahani M.H., Cubellis M.V. A Maternal Loss-of-Function Variant in KHDC3L Gene Causes a Range of Adverse Pregnancy Outcomes: A Case Report. Mol. Genet. Genomic Med. 2025;13 doi: 10.1002/mgg3.70051. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Associated Data This section collects any data citations, data availability statements, or supplementary materials included in this article. Supplementary Materials Document S1. Figures S1–S21 and Tables S1 and S2 mmc1.pdf (5.1MB, pdf) Data S1. Original code for high-content imaging analysis mmc2.zip (161.4KB, zip) Data Availability Statement • The raw transcriptomics data of the RNA-Seq supporting the results of this article have been deposited at the GEO database (GEO: GSE309480 ) and are publicly available at: https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE309480 . • All original code is available in this paper’s supplemental information. • Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request. Articles from iScience are provided here courtesy of Elsevier ACTIONS View on publisher site PDF (13.9 MB) Cite Collections Permalink PERMALINK Copy RESOURCES Similar articles Cited by other articles Links to NCBI Databases Cite Copy Download .nbib .nbib Format: AMA APA MLA NLM Add to Collections Create a new collection Add to an existing collection Name your collection * Choose a collection Unable to load your collection due to an error Please try again Add Cancel Follow NCBI NCBI on X (formerly known as Twitter) NCBI on Facebook NCBI on LinkedIn NCBI on GitHub NCBI RSS feed Connect with NLM NLM on X (formerly known as Twitter) NLM on Facebook NLM on YouTube National Library of Medicine 8600 Rockville Pike Bethesda, MD 20894 Web Policies FOIA HHS Vulnerability Disclosure Help Accessibility Careers NLM NIH HHS USA.gov Back to Top

Related documents

• BioHackSWAT4HCLS25 report: Towards an interactive mapping experience for data owners#462510
• DivideFold+: an AI-based tool for RNA secondary structure prediction with subdomains identification and visualization and data augmentation#268471
• A VNIR spectral database of terrestrial magmatic rocks for the interpretation of Martian plagioclase hyperspectral detections#417951
• Leveraging UAV hyperspectral imaging for crop physiology and biochemistry: A comprehensive review of feature extraction and selection methods.#138635
• CHM-Net: Center Heatmap-driven Macro-Micro Modeling Network for MRI-based Microbial Density Stratification#630073