EPOP Restricts PRC2.1 Targeting to Chromatin by Directly Modulating Enzyme Complex Dimerization.

Polycomb repressive complex 2 (PRC2) mediates developmental gene repression as two classes of holocomplexes, PRC2.1 and PRC2.2. EPOP is an accessory subunit specific to PRC2.1, which also contains PCL proteins. Unlike other accessory subunits that collectively facilitate PRC2 targeting, EPOP was implicated in an enigmatic inhibitory role, together with its interactor Elongin BC. We report an unusual molecular mechanism whereby EPOP regulates PRC2.1 by directly modulating its oligomerization state. EPOP disrupts the PRC2.1 dimer and weakens its chromatin association, likely by disabling the avidity effect conferred by the dimeric complex. Congruently, an EPOP mutant specifically defective in PRC2 binding enhances genome-wide enrichments of MTF2 and H3K27me3 in mouse epiblast-like cells. Elongin BC is largely dispensable for the EPOP-mediated inhibition of PRC2.1. EPOP defines a distinct subclass of PRC2.1, which uniquely maintains an epigenetic program by preventing the over-repression of key gene regulators along the continuum of early differentiation.

Cinnamaldehyde Alleviates Alveolar Epithelial Cell Injury in ALI by Inhibiting the CaMKII Pathway.

Alveolar epithelial cell injury plays a key role in acute lung injury (ALI) and is a vital determinant of its severity. Here, we aimed to assess the protective effects of cinnamaldehyde (CA) on lipopolysaccharide (LPS)-induced A549 cells and elucidate the underlying mechanisms. A549 cells were stimulated with 1 µg/mL LPS for 24 h to establish an alveolar epithelial cell injury model and subsequently treated with CA or Ca2+/calmodulin-dependent protein kinase II (CaMKII) inhibitor KN93. Flow cytometry, 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide, and lactate dehydrogenase release assays were used to evaluate apoptosis, cell viability, and lactate dehydrogenase activity, respectively. Levels of inflammatory cytokines (interleukin-6, interleukin-1ß, tumor necrosis tactor-α, and interferon-γ) and oxidative stress markers (reactive oxygen species, superoxide dismutase, catalase, and malondialdehyde) were determined using enzyme-linked immunosorbent assay and specific assay kits, respectively. Furthermore, levels of apoptosis-related proteins (cleaved caspase-3, Bcl-2-associated X, and Bcl-2) and CaMKII were assessed via western blotting. CA did not exhibit significant cytotoxicity in A549 cells. It dose-dependently improved the cell viability, suppressed apoptosis, decreased cleaved caspase-3 and Bcl-2-associated X levels, and increased Bcl-2 levels in LPS-treated A549 cells. It also inhibited inflammatory factor release and oxidative stress in LPS-induced A549 cells. Similar results were observed in the KN93- and CA-treated groups. Western blotting assay revealed that CA and KN93 inhibited CaMKII pathway activation, as indicated by the reduced p-CaMKII and p-phospholamban (PLN) levels and p-CaMKII/CaMKII and p-PLN/PLN ratios. Overall, CA alleviated alveolar epithelial cell injury by inhibiting the inflammatory response and oxidative stress and inducing cell apoptosis in LPS-induced A549 cells by regulating the CaMKII pathway, serving as a potential candidate for ALI prevention and treatment.

Fine-Tuned Large Language Model for Visualization System: A Study on Self-Regulated Learning in Education.

Large Language Models (LLMs) have shown great potential in intelligent visualization systems, especially for domainspecific applications. Integrating LLMs into visualization systems presents challenges, and we categorize these challenges into three alignments: domain problems with LLMs, visualization with LLMs, and interaction with LLMs. To achieve these alignments, we propose a framework and outline a workflow to guide the application of fine-tuned LLMs to enhance visual interactions for domain-specific tasks. These alignment challenges are critical in education because of the need for an intelligent visualization system to support beginners' self-regulated learning. Therefore, we apply the framework to education and introduce Tailor-Mind, an interactive visualization system designed to facilitate self-regulated learning for artificial intelligence beginners. Drawing on insights from a preliminary study, we identify self-regulated learning tasks and fine-tuning objectives to guide visualization design and tuning data construction. Our focus on aligning visualization with fine-tuned LLM makes Tailor-Mind more like a personalized tutor. Tailor-Mind also supports interactive recommendations to help beginners better achieve their learning goals. Model performance evaluations and user studies confirm that Tailor-Mind improves the self-regulated learning experience, effectively validating the proposed framework.

Highly Regular Layered Structure via Dual-Spatially-Confined Alignment of Nanosheets Enables High-Performance Nanocomposites.

Assembling ultrathin nanosheets into layered structure represents one promising way to fabricate high-performance nanocomposites. However, how to minimize the internal defects of the layered assemblies to fully exploit the intrinsic mechanical superiority of nanosheets remains challenging. Here, a dual-scale spatially confined strategy for the co-assembly of ultrathin nanosheets with different aspect ratios into a near-perfect layered structure is developed. Large-aspect-ratio (LAR) nanosheets are aligned due to the microscale confined space of a flat microfluidic channel, small-aspect-ratio (SAR) nanosheets are aligned due to the nanoscale confined space between adjacent LAR nanosheets. During this co-assembly process, SAR nanosheets can flatten LAR nanosheets, thus reducing wrinkles and pores of the assemblies. Benefiting from the precise alignment (orientation degree of 90.74%) of different-sized nanosheets, efficient stress transfer between nanosheets and interlayer matrix is achieved, resulting in layered nanocomposites with multiscale mechanical enhancement and superior fatigue durability (100 000 bending cycles). The proposed co-assembly strategy can be used to orderly integrate high-quality nanosheets with different sizes or diverse functions toward high-performance or multifunctional nanocomposites.

CKG-IMC: An inductive matrix completion method enhanced by CKG and GNN for Alzheimer's disease compound-protein interactions prediction.

Alzheimer's disease (AD) is one of the most prevalent chronic neurodegenerative disorders globally, with a rapidly growing population of AD patients and currently no effective therapeutic interventions available. Consequently, the development of therapeutic anti-AD drugs and the identification of AD targets represent one of the most urgent tasks. In this study, in addition to considering known drugs and targets, we explore compound-protein interactions (CPIs) between compounds and proteins relevant to AD. We propose a deep learning model called CKG-IMC to predict Alzheimer's disease compound-protein interaction relationships. CKG-IMC comprises three modules: a collaborative knowledge graph (CKG), a principal neighborhood aggregation graph neural network (PNA), and an inductive matrix completion (IMC). The collaborative knowledge graph is used to learn semantic associations between entities, PNA is employed to extract structural features of the relationship network, and IMC is utilized for CPIs prediction. Compared with a total of 16 baseline models based on similarities, knowledge graphs, and graph neural networks, our model achieves state-of-the-art performance in experiments of 10-fold cross-validation and independent test. Furthermore, we use CKG-IMC to predict compounds interacting with two confirmed AD targets, 42-amino-acid ß-amyloid (Aß42) protein and microtubule-associated protein tau (tau protein), as well as proteins interacting with five FDA-approved anti-AD drugs. The results indicate that the majority of predictions are supported by literature, and molecular docking experiments demonstrate a strong affinity between the predicted compounds and targets.

PFKFB3 controls acinar IP3R-mediated Ca2+ overload to regulate acute pancreatitis severity.

Acute pancreatitis (AP) is among the most common hospital gastrointestinal diagnoses; understanding the mechanisms underlying the severity of AP is critical for development of new treatment options for this disease. Here, we evaluate the biological function of phosphofructo-2-kinase/fructose-2,6-biphosphatase 3 (PFKFB3) in AP pathogenesis in 2 independent genetically engineered mouse models of AP. PFKFB3 was elevated in AP and severe AP (SAP), and KO of Pfkfb3 abrogated the severity of alcoholic SAP (FAEE-SAP). Using a combination of genetic, pharmacological, and molecular studies, we defined the interaction of PFKFB3 with inositol 1,4,5-trisphosphate receptor (IP3R) as a key event mediating this phenomenon. Further analysis demonstrated that the interaction between PFKFB3 and IP3R promotes FAEE-SAP severity by altering intracellular calcium homeostasis in acinar cells. Together, our results support a PFKFB3-driven mechanism controlling AP pathobiology and define this enzyme as a therapeutic target to ameliorate the severity of this condition.

Effects of phosphorous and antimony doping on thin Ge layers grown on Si.

Suppression of threading dislocations (TDs) in thin germanium (Ge) layers grown on silicon (Si) substrates has been critical for realizing high-performance Si-based optoelectronic and electronic devices. An advanced growth strategy is desired to minimize the TD density within a thin Ge buffer layer in Ge-on-Si systems. In this work, we investigate the impact of P dopants in 500-nm thin Ge layers, with doping concentrations from 1 to 50 × 1018 cm-3. The introduction of P dopants has efficiently promoted TD reduction, whose potential mechanism has been explored by comparing it to the well-established Sb-doped Ge-on-Si system. P and Sb dopants reveal different defect-suppression mechanisms in Ge-on-Si samples, inspiring a novel co-doping technique by exploiting the advantages of both dopants. The surface TDD of the Ge buffer has been further reduced by the co-doping technique to the order of 107 cm-2 with a thin Ge layer (of only 500 nm), which could provide a high-quality platform for high-performance Si-based semiconductor devices.

Hierarchical and reconfigurable interfibrous interface of bioinspired Bouligand structure enabled by moderate orderliness.

Introducing natural Bouligand structure into synthetics is expected to develop high-performance structural materials. Interfibrous interface is critical to load transfer, and mechanical functionality of bioinspired Bouligand structure yet receives little attention. Here, we propose one kind of hierarchical and reconfigurable interfibrous interface based on moderate orderliness to mechanically reinforce bioinspired Bouligand structure. The interface imparted by moderate alignment of adaptable networked nanofibers hierarchically includes nanofiber interlocking and hydrogen-bonding (HB) network bridging, being expected to facilitate load transfer and structural stability through dynamic adjustment in terms of nanofiber sliding and HB breaking-reforming. As one demonstration, the hierarchical and reconfigurable interfibrous interface is constructed based on moderate alignment of networked bacterial cellulose nanofibers. We show that the resultant bioinspired Bouligand structural material exhibits unusual strengthening and toughening mechanisms dominated by interface-microstructure multiscale coupling. The proposed interfibrous interface enabled by moderate orderliness would provide mechanical insight into the assembly of widely existing networked nanofiber building blocks toward high-performance macroscopic bioinspired structural assemblies.

LEVA: Using Large Language Models to Enhance Visual Analytics.

Visual analytics supports data analysis tasks within complex domain problems. However, due to the richness of data types, visual designs, and interaction designs, users need to recall and process a significant amount of information when they visually analyze data. These challenges emphasize the need for more intelligent visual analytics methods. Large language models have demonstrated the ability to interpret various forms of textual data, offering the potential to facilitate intelligent support for visual analytics. We propose LEVA, a framework that uses large language models to enhance users' VA workflows at multiple stages: onboarding, exploration, and summarization. To support onboarding, we use large language models to interpret visualization designs and view relationships based on system specifications. For exploration, we use large language models to recommend insights based on the analysis of system status and data to facilitate mixed-initiative exploration. For summarization, we present a selective reporting strategy to retrace analysis history through a stream visualization and generate insight reports with the help of large language models. We demonstrate how LEVA can be integrated into existing visual analytics systems. Two usage scenarios and a user study suggest that LEVA effectively aids users in conducting visual analytics.

[Effects of Platelet-Rich Plasma-Derived Exosomes on Proliferation and Migration of Tendon Stem/Progenitor Cell].

Objective To investigate the effects of platelet-rich plasma-derived exosomes (PRP-Exos) on the proliferation and migration of tendon stem/progenitor cell (TSPC).Methods PRP-Exos were extracted through the combination of polymer-based precipitation and ultracentrifugation.The morphology,concentration,and particle size of PRP-Exos were identified by transmission electron microscopy and nanoparticle tracking analysis.The expression levels of surface marker proteins on PRP-Exos and platelet membrane glycoproteins were determined by Western blot analysis.Rat TSPC was extracted and cultured,and the expression of surface marker molecules on TSPC was detected using flow cytometry and immunofluorescence staining.The proliferation of TSPC influenced by PRP-Exos was evaluated using CCK-8 assay and EdU assay.The effect of PRP-Exos on the migration of TSPC was evaluated by cell scratch assay and Transwell assay.Results The extracted PRP-Exos exhibit typical saucer-like structures,with a concentration of 4.9×1011 particles/mL,an average particle size of (132.2±56.8) nm,and surface expression of CD9,CD63 and CD41.The extracted TSPC expressed the CD44 protein.PRP-Exos can be taken up by TSPC,and after co-cultured for 48 h,concentrations of 50 and 100 µg/mL of PRP-Exos significantly promoted the proliferation of TSPC (both P<0.001),with no statistical difference between the two concentrations (P=0.283).Additionally,after co-cultured for 24 h,50 µg/mL of PRP-Exos significantly promoted the migration of TSPC (P<0.001).Conclusion Under in vitro culture conditions,PRP-Exos significantly promote the proliferation and migration of rat TSPC.

Ultrasonic shear wave elastography predicts the quality of the residual tendon before the rotator cuff repair.

BACKGROUND AND PURPOSE: Effective evaluation of rotator cuff tear residual tendon quality is the key to surgical repair. However, until now, the evaluation of rotator cuff tissue by ultrasonic shear wave elasticity (SWE) has been controversial. This prospective study analyzed the association between preoperative SWE and arthroscopic residual tendon quality scores. METHODS: The shear wave velocity (SWV) of the deltoid muscle, the supraspinatus tendon, and the supraspinatus muscle were measured in full-thickness rotator cuff tear patients. Tendon quality was scored according to tear size, tendon margin, tendon thickness, and footprint coverage during arthroscopy. The arthroscopic scores were used as the gold standard, and the SWV ratio of tendon and muscle (supraspinatus tendon/deltoid and supraspinatus muscle/deltoid) were calculated and correlated with the arthroscopic scores. RESULT: Eighty-nine patients (129 shoulders) were enrolled, including 89 operation shoulders and 40 control shoulders. In the group of operation shoulders, both the SWV ratios of tendon (SWV-RT) and the SWV ratio of muscle (SWV-RM) were negatively correlated with arthroscopic scores (The correlation coefficient (R) ranged from -0.722 to -0.884 and -0.569 to -0.689). The SWV-RT and SWV-RM of the operation shoulders were significantly lower than that of the control shoulders (p < 0.05). CONCLUSION: SWE could be used to predict the quality of the residual tendon before the rotator cuff repair. SWV of the supraspinatus tendon and muscle was a useful parameter to predict the quality of the residual tendon. CRITICAL RELEVANCE STATEMENT: Measuring the shear wave velocity of the supraspinatus tendon and muscle with SWE is useful for predicting the quality of the residual tendon which is one of the key factors for a successful rotator cuff repair. KEY POINTS: • Evaluating the quality of the residual tendon is important before surgery. • Elasticity measurements were negatively correlated with the arthroscopic score. • SWE is useful for predicting the quality of the residual tendon.

Viewpoint Recommendation for Point Cloud Labeling through Interaction Cost Modeling.

Semantic segmentation of 3D point clouds is important for many applications, such as autonomous driving. To train semantic segmentation models, labeled point cloud segmentation datasets are essential. Meanwhile, point cloud labeling is time-consuming for annotators, which typically involves tuning the camera viewpoint and selecting points with a lasso tool. To reduce the time cost of point cloud labeling, we propose a viewpoint recommendation approach to reduce annotators' labeling time costs. We adapt Fitts' law to model the time cost of lasso selection in point clouds. Using the modeled time cost, the viewpoint that minimizes the lasso selection time cost is recommended to the annotator. We build a data labeling system for semantic segmentation of 3D point clouds that integrates our viewpoint recommendation approach. The system enables users to navigate to recommended viewpoints for efficient annotation. Through a user study, we observed that our approach effectively reduced the data labeling time cost. We also qualitatively compare our approach with previous viewpoint selection approaches on different datasets.

Bottom-Up Film-to-Bulk Assembly Toward Bioinspired Bulk Structural Nanocomposites.

Biological materials, although composed of meager minerals and biopolymers, often exhibit amazing mechanical properties far beyond their components due to hierarchically ordered structures. Understanding their structure-properties relationships and replicating them into artificial materials would boost the development of bulk structural nanocomposites. Layered microstructure widely exists in biological materials, serving as the fundamental structure in nanosheet-based nacres and nanofiber-based Bouligand tissues, and implying superior mechanical properties. High-efficient and scalable fabrication of bioinspired bulk structural nanocomposites with precise layered microstructure is therefore important yet remains difficult. Here, one straightforward bottom-up film-to-bulk assembly strategy is focused for fabricating bioinspired layered bulk structural nanocomposites. The bottom-up assembly strategy inherently offers a methodology for precise construction of bioinspired layered microstructure in bulk form, availability for fabrication of bioinspired bulk structural nanocomposites with large sizes and complex shapes, possibility for design of multiscale interfaces, feasibility for manipulation of diverse heterogeneities. Not limited to discussing what has been achieved by using the current bottom-up film-to-bulk assembly strategy, it is also envisioned how to promote such an assembly strategy to better benefit the development of bioinspired bulk structural nanocomposites. Compared to other assembly strategies, the highlighted strategy provides great opportunities for creating bioinspired bulk structural nanocomposites on demand.

In-situ gelation of fibrin gel encapsulating platelet-rich plasma-derived exosomes promotes rotator cuff healing.

Although platelet-rich plasma-derived exosomes (PRP-Exos) hold significant repair potential, their efficacy in treating rotator cuff tear (RCT) remains unknown. In light of the potential for clinical translation of fibrin gel and PRP-Exos, we evaluated their combined impact on RCT healing and explored suitable gel implantation techniques. In vitro experiments demonstrated that PRP-Exos effectively enhanced key phenotypes changes in tendon stem/progenitor cells. Multi-modality imaging, including conventional ultrasound, shear wave elastography ultrasound, and micro-computed tomography, and histopathological assessments were performed to collectively evaluate the regenerative effects on RCT. The regenerated tendons exhibited a well-ordered structure, while bone and cartilage regeneration were significantly improved. PRP-Exos participated in the healing process of RCT. In-situ gelation of fibrin gel-encapsulated PRP-Exos at the bone-tendon interface during surgery proved to be a feasible gel implantation method that benefits the healing outcome. Comprehensive multi-modality postoperative evaluations were necessary, providing a reliable foundation for post-injury repair.

HucMSCs Delay Muscle Atrophy After Peripheral Nerve Injury Through Exosomes by Repressing Muscle-Specific Ubiquitin Ligases.

Cell therapy based on mesenchymal stem cells (MSCs) alleviate muscle atrophy caused by diabetes and aging; however, the impact of human umbilical cord mesenchymal stem cells on muscle atrophy following nerve injury and the underlying mechanisms remain unclear. In this study, we evaluated the therapeutic efficacy of human umbilical cord MSCs (hucMSCs) and hucMSC-derived exosomes (hucMSC-EXOs) for muscle atrophy following nerve injury and identified the underlying molecular mechanisms. Sciatic nerve crush injury in rats and the induction of myotubes in L6 cells were used to determine the ameliorating effect of hucMSCs and hucMSC-EXOs on muscle atrophy. Q-PCR and Western blot analyses were used to measure the expression of muscle-specific ubiquitin ligases Fbxo32 (Atrogin1, MAFbx) and Trim63 (MuRF-1). Dual-luciferase reporter gene experiments were conducted to validate the direct binding of miRNAs to their target genes. Local injection of hucMSCs and hucMSC-EXOs mitigated atrophy in the rat gastrocnemius muscle following sciatic nerve crush injury. In vitro, hucMSC-EXOs alleviated atrophy in L6 myotubes. Mechanistic analysis indicated the upregulation of miR-23b-3p levels in L6 myotubes following hucMSC-EXOs treatment. MiR-23b-3p significantly inhibited the expression of its target genes, Fbxo32 and Trim63, and suppressed myotube atrophy. Notably, an miR-23b-3p inhibitor reversed the inhibitory effect of miR-23b-3p on myotube atrophy in vitro. These results suggest that hucMSCs and their exosomes alleviate muscle atrophy following nerve injury. MiR-23b-3p in exosomes secreted by hucMSCs contributes to this mechanism by inhibiting the muscle-specific ubiquitination ligases Fbxo32 and Trim63.

Bioinspired polysaccharide-based nanocomposite membranes with robust wet mechanical properties for guided bone regeneration.

Polysaccharide-based membranes with excellent mechanical properties are highly desired. However, severe mechanical deterioration under wet conditions limits their biomedical applications. Here, inspired by the structural heterogeneity of strong yet hydrated biological materials, we propose a strategy based on heterogeneous crosslink-and-hydration (HCH) of a molecule/nano dual-scale network to fabricate polysaccharide-based nanocomposites with robust wet mechanical properties. The heterogeneity lies in that the crosslink-and-hydration occurs in the molecule-network while the stress-bearing nanofiber-network remains unaffected. As one demonstration, a membrane assembled by bacterial cellulose nanofiber-network and Ca2+-crosslinked and hydrated sodium alginate molecule-network is designed. Studies show that the crosslinked-and-hydrated molecule-network restricts water invasion and boosts stress transfer of the nanofiber-network by serving as interfibrous bridge. Overall, the molecule-network makes the membrane hydrated and flexible; the nanofiber-network as stress-bearing component provides strength and toughness. The HCH dual-scale network featuring a cooperative effect stimulates the design of advanced biomaterials applied under wet conditions such as guided bone regeneration membranes.

Acyl-coenzyme A: cholesterol acyltransferase inhibitor avasimibe suppresses tumorigenesis and induces G1-phase cell-cycle arrest by activating PPARγ signaling pathway in bladder cancer.

Reprogramming of energy metabolism is one of the most important characteristics of tumors. Bladder cancer (BLCA) cells contain higher levels of cholesterol content compared to normal cells, and acyl-coenzyme A (CoA): cholesterol acyltransferase-1 (ACAT1) plays a crucial role in the esterification of cholesterol. Avasimibe is a drug that has been used in the treatment of atherosclerosis, and it can effectively inhibit ACAT1. We observed that ACAT1 was significantly up-regulated in BLCA and positively correlated with tumor grade. By avasimibe administration, the proliferation and migration ability of BLCA cells were reduced, while the production of ROS was strongly increased, accompanied by the up-regulated expression of ROS metabolism-related proteins SOD2 and catalase. Furthermore, BLCA cell cycle was arrested at the G1 phase, accompanied by the downregulation of cell cycle-related proteins (CCNA1/2, CCND1, CDK2 and CDK4), while the PPARγ was found to be up-regulated at both transcriptional and protein levels after avasimibe treatment. Then we found that the PPARγ antagonist GW9662 could reverse the effect of avasimibe on the cell cycle. Moreover, xenograft and pulmonary metastasis models further demonstrated that avasimibe could inhibit tumor cell growth and metastasis in vivo. Taken together, our results for the first time revealed that avasimibe can inhibit BLCA progression and metastasis, and PPARγ signaling pathway may play a key role in regulation of cell cycle distribution induced by avasimibe.

OldVisOnline: Curating a Dataset of Historical Visualizations.

With the increasing adoption of digitization, more and more historical visualizations created hundreds of years ago are accessible in digital libraries online. It provides a unique opportunity for visualization and history research. Meanwhile, there is no large-scale digital collection dedicated to historical visualizations. The visualizations are scattered in various collections, which hinders retrieval. In this study, we curate the first large-scale dataset dedicated to historical visualizations. Our dataset comprises 13K historical visualization images with corresponding processed metadata from seven digital libraries. In curating the dataset, we propose a workflow to scrape and process heterogeneous metadata. We develop a semi-automatic labeling approach to distinguish visualizations from other artifacts. Our dataset can be accessed with OldVisOnline, a system we have built to browse and label historical visualizations. We discuss our vision of usage scenarios and research opportunities with our dataset, such as textual criticism for historical visualizations. Drawing upon our experience, we summarize recommendations for future efforts to improve our dataset.

TransforLearn: Interactive Visual Tutorial for the Transformer Model.

The widespread adoption of Transformers in deep learning, serving as the core framework for numerous large-scale language models, has sparked significant interest in understanding their underlying mechanisms. However, beginners face difficulties in comprehending and learning Transformers due to its complex structure and abstract data representation. We present TransforLearn, the first interactive visual tutorial designed for deep learning beginners and non-experts to comprehensively learn about Transformers. TransforLearn supports interactions for architecture-driven exploration and task-driven exploration, providing insight into different levels of model details and their working processes. It accommodates interactive views of each layer's operation and mathematical formula, helping users to understand the data flow of long text sequences. By altering the current decoder-based recursive prediction results and combining the downstream task abstractions, users can deeply explore model processes. Our user study revealed that the interactions of TransforLearn are positively received. We observe that TransforLearn facilitates users' accomplishment of study tasks and a grasp of key concepts in Transformer effectively.

Interpreting High-Dimensional Projections With Capacity.

Dimensionality reduction (DR) algorithms are diverse and widely used for analyzing high-dimensional data. Various metrics and tools have been proposed to evaluate and interpret the DR results. However, most metrics and methods fail to be well generalized to measure any DR results from the perspective of original distribution fidelity or lack interactive exploration of DR results. There is still a need for more intuitive and quantitative analysis to interactively explore high-dimensional data and improve interpretability. We propose a metric and a generalized algorithm-agnostic approach based on the concept of capacity to evaluate and analyze the DR results. Based on our approach, we develop a visual analytic system HiLow for exploring high-dimensional data and projections. We also propose a mixed-initiative recommendation algorithm that assists users in interactively DR results manipulation. Users can compare the differences in data distribution after the interaction through HiLow. Furthermore, we propose a novel visualization design focusing on quantitative analysis of differences between high and low-dimensional data distributions. Finally, through user study and case studies, we validate the effectiveness of our approach and system in enhancing the interpretability of projections and analyzing the distribution of high and low-dimensional data.