Prototypical Patterns Research Articles

Out-of-Distribution Detection with Memory-Augmented Variational Autoencoder

This paper proposes a novel method capable of both detecting OOD data and generating in-distribution data samples. To achieve this, a VAE model is adopted and augmented with a memory module, providing capacities for identifying OOD data and synthesising new in-distribution samples. The proposed VAE is trained on normal data and the memory stores prototypical patterns of the normal data distribution. At test time, the input is encoded by the VAE encoder; this encoding is used as a query to retrieve related memory items, which are then integrated with the input encoding and passed to the decoder for reconstruction. Normal samples reconstruct well and yield low reconstruction errors, while OOD inputs produce high reconstruction errors as their encodings get replaced by retrieved normal patterns. Prior works use memory modules for OOD detection with autoencoders, but this method leverages a VAE architecture to enable generation abilities. Experiments conducted with CIFAR-10 and MNIST datasets show that the memory-augmented VAE consistently outperforms the baseline, particularly where OOD data resembles normal patterns. This notable improvement is due to the enhanced latent space representation provided by the VAE. Overall, the memory-equipped VAE framework excels in identifying OOD and generating creative examples effectively.

Rethinking prediction-based video anomaly detection from local–global normality perspective

Video anomaly detection (VAD) has been intensively studied for years because of its potential applications in intelligent video systems. Prediction-based VAD methods have gained a lot of attention in recent years due to their simplicity and effectiveness, which learn normality through a frame prediction tasks. However, we find that they often consider only local or global normality in the temporal dimension thus are difficult to generalize for various scenarios. Specifically, some of them focus on learning local spatiotemporal representations from consecutive frames to enhance the representation for normal events while the other methods are devoted to memorizing prototypical normal patterns of whole training videos to weaken the generalization for anomalies. In this paper, we rethink these two types of methods and find that the former methods are difficult to adapt to simple scenarios due to powerful representation allows these methods to represent some anomalies and causes miss detection while the latter struggle with complex scenarios due to the limitations for diverse normal patterns and causes false alarm. To verify our discoveries, we design a two-branch model, Local–Global Normality Network (LGN-Net), to simultaneously learn local and global normality. We verified the effect of local and global normality on the VAD model by ablation study and find that the model can be more adaptive to different scenarios when accounting for both normality. The code is available online: https://github.com/Myzhao1999/LGN-Net.

Data-driven unsupervised anomaly detection of manufacturing processes with multi-scale prototype augmentation and multi-sensor data

Accurate anomaly detection (AD) of machine tools is crucial to ensure the quality and efficiency of the manufacturing processes. Due to the lack of tool anomaly information, it is difficult for AD model to precisely capture the distribution of health states and then obtain a discriminative decision boundary. Current methods try to reconstruct the normal data distribution without restricting the abnormal, resulting in the unacceptable overlap between normal and abnormal regions and finally leading to high false alarm rate. To tackle these issues, a hierarchical augmented autoencoder is proposed for AD of machine tools during manufacturing. First, a skip-connected autoencoder is built to basically learn the normal representations of multi-sensor data in an unsupervised manner. Then, to improve further emphasis the reconstruction on normality and suppress that on anomalies, we propose hierarchical memory modules to store multi-scale normal prototypical patterns, using them as a prior to guide the reconstruction with preference. Finally, A compound metric loss function is designed to measure data similarity considering both distance and angle perspectives, which can restrain noise interference and enhance model robustness. Extensive experiments are conducted on real-world CNC machine tool datasets, the proposed method achieves better performance for unsupervised AD compared with other typical methods.

Individualized generation of women's prototype based on the classification of body shape

The traditional prototype method only relies on the basic dimensions of the human body such as bust girth and back length to make the prototype pattern, ignoring the differences of human body details dimensions. In order to improve the fit of garments, the research on intelligent generation technology of individualized garment patterns has become one of the hot spots in the field of clothing. Based on the 3D body scanning technology, we proposed a method for generating prototype pattern based on body shape classification. The proposed method takes as inputs the body parameters (width, thickness and angle), while the output of the method is one of the four pattern generation rules-"Y" shape, "V" shape, "I" shape, or "X" shape. 207 subjects were divided into four categories based on the body angle parameters, namely "Y" shape, "V" shape, "I" shape and "X" shape, and a shape recognition model was built. The pattern database is required to obtain pattern generation rules by 3D point cloud reconstruction and flattening. Combined with the human body shape parameters, the mathematical formulas of the feature points on the pattern landmarks are built. The accuracy of body shape recognition model is 97.4%. The error analysis of the predicted pattern parameters shows that 90% of the pattern feature points have a goodness of fit above 0.7. In the 5 main landmarks, the proportion of pattern within the absolute error range is more than 80%, indicating that the prediction effect of the pattern is good. This method can be applied to the process of automatic pattern generation system based 2D measurement to improve work efficiency.

Recognizing gender from images with facial makeup

<p>Recognizing the sex of an individual is a difficult task due to pose variation, occlusion, illumination effect, facial expression, plastic surgery, and makeup. In this manuscript, a novel approach for gender recognition with facial makeup is proposed. A novel Log-Gabor COSFIRE (LG-COSFIRE) filter is a shape-selective filter that is trained with prototype patterns of interest. The geometrical structure of the faces is acquired using the dual-tree complex wavelet transform (DT-CWT). Dense SIFT descriptor extracts the shape attributes of an image by building local histograms of gradient orientation. Finally, least square support vector machine (LS-SVM) is utilized to recognize the gender of an individual. The experiment was performed on self-built facial makeup for male and female (FMMF) database and achieves 89.7% accuracy.</p>

Consolation Through Argumentation? Prototypical and Stereotypical Argumentative Patterns in Secular Eulogies

Consolation Through Argumentation? Prototypical and Stereotypical Argumentative Patterns in Secular Eulogies

“So I know how to do this”

Abstract Debates are important events during presidential elections in the U.S.A. Candidates are juxtaposed and engage with each other on a wide range of issues. This poses the question how disagreement between the two candidates and the public is managed. The aim of this paper is to articulate the prototypical argumentative pattern used by candidates which shows that to defend that the public should vote for them, candidates recurringly make three central claims. Specifically, they claim that some political action has to happen, they will do that action if elected, while their opponent will not. This basic argument scheme – which could be referred to as campaign promise argumentation – is further expanded by candidates by responding to six distinct critical questions, resulting in a prototypical argumentative pattern designed to deal with potential criticisms against a bid to become president.

Dominant-while-speaking: How bilingual Norwegian–English children conceptualize goal-oriented motion events

Aim: This study investigates how simultaneous bilingual Norwegian–English children conceptualize goal-oriented motion events in their two languages, which have different and partly conflicting language-specific conceptualization patterns (the so-called holistic, endpoint-oriented perspective vs. a phasal perspective with focus on ongoingness). Design: The experiment combined three different methodologies to measure attention to endpoints: elicitation of spoken data, eye-tracking, and a subsequent memory test. A total of 23 bilingual Norwegian–English children participated on a separate day for each language. The comparison groups comprised monolingual Norwegian children ( n = 21), first language (L1) Norwegian adults ( n = 30), and L1 English adults ( n = 20). Data and analysis: The statistical analysis included calculations of endpoints mentioned and endpoints remembered in each language and for both sessions, and an event-related analysis was conducted to establish the total length of all fixations on the areas of interest (AoIs [the endpoints]), in each language and for both sessions. Findings/conclusions: The results showed an effect of the language of operation on conceptualization, independent of language dominance, but depending on the situation. In their first session, the bilingual children showed a strong awareness of the prototypical conceptualization pattern in the language they were speaking. In their second session, the children’s conceptualization was influenced by their first. Hence, we can talk about a flexible conceptual dominance linked to language, the situation, and to previous experience. Originality: This study for the first time explores simultaneous bilingual children’s conceptualization of goal-oriented motion events in their two languages and with different methodologies. Significance/implications: This study brings new insights into our understanding of early bilingualism at a conceptual level, with implications for the field, but also for parents, caregivers, and stakeholders, who at times need to be reminded about children’s unique capacity for language learning.

HMGB1: A New Target for Ischemic Stroke and Hemorrhagic Transformation.

Stroke in China is distinguished by its high rates of morbidity, recurrence, disability, and mortality. The ultra-early administration of rtPA is essential for restoring perfusion in acute ischemic stroke, though it concurrently elevates the risk of hemorrhagic transformation. High-mobility group box 1 (HMGB1) emerges as a pivotal player in neuroinflammation after brain ischemia and ischemia-reperfusion. Released passively by necrotic cells and actively secreted, including direct secretion of HMGB1 into the extracellular space and packaging of HMGB1 into intracellular vesicles by immune cells, glial cells, platelets, and endothelial cells, HMGB1 represents a prototypical damage-associated molecular pattern (DAMP). It is intricately involved in the pathogenesis of atherosclerosis, thromboembolism, and detrimental inflammation during the early phases of ischemic stroke. Moreover, HMGB1 significantly contributes to neurovascular remodeling and functional recovery in later stages. Significantly, HMGB1 mediates hemorrhagic transformation by facilitating neuroinflammation, directly compromising the integrity of the blood-brain barrier, and enhancing MMP9 secretion through its interaction with rtPA. As a systemic inflammatory factor, HMGB1 is also implicated in post-stroke depression and an elevated risk of stroke-associated pneumonia. The role of HMGB1 extends to influencing the pathogenesis of ischemia by polarizing various subtypes of immune and glial cells. This includes mediating excitotoxicity due to excitatory amino acids, autophagy, MMP9 release, NET formation, and autocrine trophic pathways. Given its multifaceted role, HMGB1 is recognized as a crucial therapeutic target and prognostic marker for ischemic stroke and hemorrhagic transformation. In this review, we summarize the structure and redox properties, secretion and pathways, regulation of immune cell activity, the role of pathophysiological mechanisms in stroke, and hemorrhage transformation for HMGB1, which will pave the way for developing new neuroprotective drugs, reduction of post-stroke neuroinflammation, and expansion of thrombolysis time window.

Parallel Chords: an audio-visual analytics design for parallel coordinates

AbstractOne of the commonly used visualization techniques for multivariate data is the parallel coordinates plot. It provides users with a visual overview of multivariate data and the possibility to interactively explore it. While pattern recognition is a strength of the human visual system, it is also a strength of the auditory system. Inspired by the integration of the visual and auditory perception in everyday life, we introduce an audio-visual analytics design named Parallel Chords combining both visual and auditory displays. Parallel Chords lets users explore multivariate data using both visualization and sonification through the interaction with the axes of a parallel coordinates plot. To illustrate the potential of the design, we present (1) prototypical data patterns where the sonification helps with the identification of correlations, clusters, and outliers, (2) a usage scenario showing the sonification of data from non-adjacent axes, and (3) a controlled experiment on the sensitivity thresholds of participants when distinguishing the strength of correlations. During this controlled experiment, 35 participants used three different display types, the visualization, the sonification, and the combination of these, to identify the strongest out of three correlations. The results show that all three display types enabled the participants to identify the strongest correlation — with visualization resulting in the best sensitivity. The sonification resulted in sensitivities that were independent from the type of displayed correlation, and the combination resulted in increased enjoyability during usage.

Artificial intelligence in conceptual prototyping design

The treasures of world architecture, reflecting theories and concepts, principles and techniques of design, are an inexhaustible base for the growth and development of young architects, searching for ideas to create the image and form of an object. Artificial intelligence, in addition to developing floor plans, structural layouts, patterns for environmental and organisational decisions, etc., can effectively generate conceptual and prototypical images using the diffusion method. Artificial intelligence can significantly reduce the time of the creative process by offering a wide range of options for creative and practical solutions based on the prototype patterns and their descriptions offered to it.

Chromophobe Renal Cell Carcinoma With Extensive Retraction Artifact: A Potential Diagnostic Pitfall From Micropapillary Urothelial Carcinoma.

In addition to "classic" and eosinophilic subtype, chromophobe renal cell carcinoma (RCC) is well-known to demonstrate various morphological patterns including adenomatoid, microcystic, pigmented, multicystic, papillary, neuroendocrine-like, and small cell-like, all of which are important to appreciate for accurate diagnosis. Herein, we expand on a unique chromophobe RCC morphology not previously described consisting of tumor cells with extensive stromal retraction, mimicking upper urothelial tract micropapillary carcinoma (MPC). Twelve MPC-like chromophobe RCC nephrectomies were reviewed with clinicopathological features recorded; molecular testing was performed on 7 of 12 tumors. Patients were mostly men (n=10) with a mean age of 65 years. Mean tumor size was 6.4 cm with pathological stage distribution as follows: 4 (33%) T1a, 2 (17%) T1b, 1 (8%) T2b, and 3 (25%) T3a. The extent of MPC-like chromophobe RCC foci ranged from 10% to 40% (mean=26%; there was no correlation between the extent of MPC-like chromophobe RCC foci and tumor stage). Other chromophobe RCC morphological patterns were not identified. When performed, all (100%) tumors depicted prototypic chromophobe RCC staining pattern of KIT positivity/KRT7 positivity. Molecular showed 6 of 7 (86%) with multiple chromosomal losses. Clinically significant mutations were identified in NF1, TP53, FLCN (likely somatic), CHEK2, and ZFHX3 genes. Follow up available in 9 patients showed no evidence of disease (mean=23 months). Although the etiology behind the extensive stromal retraction in our tumors is unknown, this may likely be artifactual in nature. Nonetheless, it is important to include MPC-like chromophobe RCC in the spectrum of "variant" morphologies to avoid diagnostic pitfalls from micropapillary carcinoma.

Classification of radio galaxies with trainable COSFIRE filters

ABSTRACT Radio galaxies exhibit a rich diversity of morphological characteristics, which make their classification into distinct types a complex challenge. To address this challenge effectively, we introduce an innovative approach for radio galaxy classification using COSFIRE filters. These filters possess the ability to adapt to both the shape and orientation of prototype patterns within images. The COSFIRE approach is explainable, learning-free, rotation-tolerant, efficient, and does not require a large training set. To assess the efficacy of our method, we conducted experiments on a benchmark radio galaxy data set comprising of 1180 training samples and 404 test samples. Notably, our approach achieved an average accuracy rate of 93.36 per cent. This achievement outperforms contemporary deep learning models, and it is the best result ever achieved on this data set. Additionally, COSFIRE filters offer better computational performance, ∼20 × fewer operations than the DenseNet-based competing method (when comparing at the same accuracy). Our findings underscore the effectiveness of the COSFIRE filter-based approach in addressing the complexities associated with radio galaxy classification. This research contributes to advancing the field by offering a robust solution that transcends the orientation challenges intrinsic to radio galaxy observations. Our method is versatile in that it is applicable to various image classification approaches.

Abstract 346: Trop-2 induces apoptosis via transcriptional activation of TRAIL, FAS-FASL, CD40-TNFRSF5

Abstract Trop-2 is a transmembrane signal transducer that activates cancer growth and metastatic diffusion (1). However, this does not occur in all tumor types, as Trop-2 is associated to better disease outcome in some cancers, e.g. breast and lung tumors. A prototypic pattern of Trop-2 expression in normal multistratified epithelia, e.g. epidermis, esophagus, is that of preferential expression in non-proliferating, terminally differentiated cells. The corresponding expression gradient parallels that of induction of programmed cell death, e.g. in the corneum layer of skin keratinocytes. This led us to explore Trop-2 function as a trigger of apoptosis. We thus assessed NS-0 myeloma Trop-2 transfectants for apoptosis induction by dynamic cell morphometry and quantification of membrane blebbing and cell fragmentation. This was shown to be preceded by phosphatidyl-serine membrane exposure, and was followed by nuclear pyknosis and karyorrhexis. DNA fragmentation was shown by TUNEL and DNA laddering, and was accompanied by a dramatic increase of apoptotic death of Trop-2-expressing cells. Transcriptomic analysis at early time points of apoptosis induction indicated overexpression of TRAIL, FAS/FASL and CD40/TNFRSF5 transmembrane death receptors. This paralleled upregulation of downstream death-inducing signaling complex components and of executioner Caspases 8 and 3, suggesting Trop-2 as a trigger of extrinsic apoptosis pathways. Apoptosis induction translated into diminished tumor growth in vivo, which was shown to be proportional to Trop-2 expression levels. However, apoptosis induction was found to parallel stimulation of NS-0 cell growth and of defence mechanisms from apoptosis, as increased levels of inhibitory transcription factors, Rb-binding protein, survivin, Myd-118. These findings suggest a previously unappreciated balance between stimulation of cell growth and triggering of cell death by Trop-2, and may pave the way for selective anticancer therapy approaches in Trop-2-expressing cancers.(1) Trerotola M. et al., Neoplasia 23: 415-428 (2021). Citation Format: Saverio Alberti, Emanuela Guerra, Antonino Moschella, Martina Ceci, Ludovica Pantalone, Roberta Di Pietro, Marco Trerotola. Trop-2 induces apoptosis via transcriptional activation of TRAIL, FAS-FASL, CD40-TNFRSF5 [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 346.

Stylometry and Deep Learning: A case study on Milan Kundera's Le Livre du rire et de l'oubli

This study aims to uncover the prototypical linguistic elements and patterns of Kundera's prose in his
 novel Le Livre du rire et de l'oubli (1979, Gallimard). The exploration employs statistical and machine
 learning techniques, including the application of Hyperbase in both its web and standard versions.
 Hyperbase provides deep learning features for text classification tasks (Savoy 2015; Tuzzi & Cortelazzo
 2018), based on convolutional neural networks (Kalchbrenner et al. 2014; Kim 2014) which go beyond
 the process of convolution and incorporate an innovative deconvolution mechanism that extracts key
 linguistic markers essential for classification purposes (Vanni et al. 2018; Mayaffre & Vanni 2021). The
 training of the Hyperbase deep learning model involves an extensive corpus containing novels by 36
 authors, including Kundera, thus encompassing the French literature landscape from 1960-2014. The
 study leads to the identification of linguistic markers related to vocabulary, morphosyntax, lexical and
 grammatical patterns, and lexico-grammatical structures. These markers are then examined to reveal
 the underlying aesthetic intentions of the author. The conclusion focuses on the contribution of deep
 learning and statistics in the context of this qualitative linguistic study of a literary text.

Age-Dependent and Aβ-Induced Dynamic Changes in the Subcellular Localization of HMGB1 in Neurons and Microglia in the Brains of an Animal Model of Alzheimer's Disease.

HMGB1 is a prototypical danger-associated molecular pattern (DAMP) molecule that co-localizes with amyloid beta (Aβ) in the brains of patients with Alzheimer's disease. HMGB1 levels are significantly higher in the cerebrospinal fluid of patients. However, the cellular and subcellular distribution of HMGB1 in relation to the pathology of Alzheimer's disease has not yet been studied in detail. Here, we investigated whether HMGB1 protein levels in brain tissue homogenates (frontal cortex and striatum) and sera from Tg-APP/PS1 mice, along with its cellular and subcellular localization in those regions, differed. Total HMGB1 levels were increased in the frontal cortices of aged wildtype (7.5 M) mice compared to young (3.5 M) mice, whereas total HMGB1 levels in the frontal cortices of Tg-APP/PS1 mice (7.5 M) were significantly lower than those in age-matched wildtype mice. In contrast, total serum HMGB1 levels were enhanced in aged wildtype (7.5 M) mice and Tg-APP/PS1 mice (7.5 M). Further analysis indicated that nuclear HMGB1 levels in the frontal cortices of Tg-APP/PS1 mice were significantly reduced compared to those in age-matched wildtype controls, and cytosolic HMGB1 levels were also significantly decreased. Triple-fluorescence immunohistochemical analysis indicated that HMGB1 appeared as a ring shape in the cytoplasm of most neurons and microglia in the frontal cortices of 9.5 M Tg-APP/PS1 mice, indicating that nuclear HMGB1 is reduced by aging and in Tg-APP/PS1 mice. Consistent with these observations, Aβ treatment of both primary cortical neuron and primary microglial cultures increased HMGB1 secretion in the media, in an Aβ-dose-dependent manner. Our results indicate that nuclear HMGB1 might be translocated from the nucleus to the cytoplasm in both neurons and microglia in the brains of Tg-APP/PS1 mice, and that it may subsequently be secreted extracellularly.

Characterizing Circular and Open Business Models in a profit-driven environment through Business Model Patterns

In recent years, multiple crises have shown the fragility of global supply chains. In particular, climate change sheds light on the never-ending waste of resources caused by the linear economic model. In response, circular economy emerges as a promising alternative. Companies must rethink and innovate their business models, product life cycles, and underlying mindsets for a systemic transition. Cross-sectoral collaboration in networks is necessary to redesign, reuse, recycle, and remanufacture products and close the loop of material flows. First, projects and research on the concepts of open source hardware, open innovation, and open design show that openness can be a game-changer for the circular economy and circular business models. Although there has been growing interest, there is still a lack of comprehensive research examining possible synergies of openness and circularity, as well as identifying patterns for integrating these concepts into business models. This paper explores possible ways, prerequisites, and obstacles of open source concepts enabling circular value creation in industry. We combine explorative data from a literature review and more than 10 interviews with open source and circular economy experts. Furthermore, a total of 38 case studies of circular or open or open & circular business models were identified and selected for analysis. The empirical results were triangulated with existing business model patterns from the literature. The study presents a framework to describe open and circular business models and four prototypical patterns. In an exploratory approach, we identify necessary lines of future research in order to facilitate open source movements toward a circular economy. The paper provides insights for academics and practitioners from production and supply chain management. It combines interdisciplinary perspectives to unlock the potential of open source concepts and collaboration for enabling a circular economy.

When do leaders feel stressed by their followers? An examination of face-to-face, virtual and written interactions

This study examines in which interactions and why followers’ behaviors can be perceived as stressful by leaders. Therefore, leaders were interviewed about particularly stressful interactions with a follower in face-to-face, virtual and written interactions. We find leaders’ stressors at work to fit three categories: leader’s loss of control, (potential) loss of the leader’s valuable resources, and threats to leader’s role. Moreover, we identify diverse daily as well as prototypical patterns of how leaders deal with some of them. Additionally, we identify undesirable work-life events and find a prototypical resilient leader characterized by feeling nearly no stress. Our study reduces the current research gap of mostly considering only leader behavior to be stressful (for followers).

HMGB1 induces hepcidin upregulation in astrocytes and causes an acute iron surge and subsequent ferroptosis in the postischemic brain

Dysregulation of brain iron levels causes functional disturbances and damages neurons. Hepcidin (a peptide hormone) plays a principal role in regulating intracellular iron levels by modulating ferroportin (FPN, the only known iron exporter) through triggering its internalization and lysosomal degradation. We observed a significant and rapid iron surge in the cortices of ischemic hemispheres at 3 h after cerebral ischemia (middle cerebral artery occlusion, MCAO) that was maintained until 4 d post-MCAO. We showed upregulation of hepcidin expression in the brain as early as 3 h post-MCAO, mainly in astrocytes, and significant hepcidin accumulation in serum from 6 h post-MCAO, and these inductions were maintained for 1 day and 7 days, respectively. High mobility group box 1 (HMGB1), a prototypic danger-associated molecular pattern, accumulates markedly after transient MCAO and plays critical roles in damage aggravation via its proinflammatory effects. Here, we demonstrated that treatment with recombinant HMGB1 stimulated astrocytes to induce hepcidin expression in a TLR4- and CXCR4-dependent manner. Furthermore, hepcidin-mediated intracellular iron accumulation in neurons was confirmed by an experiment using N-methyl-D-aspartate (NMDA)-conditioned medium-treated primary astrocytes and fresh primary cortical neurons treated with hepcidin-containing astrocyte-conditioned medium. Moreover, HMGB1-mediated local hepcidin upregulation and subsequent local iron surge were found to cause ferroptosis in the postischemic brain, which was suppressed by the functional blocking of HMGB1 using intranasally administered HMGB1 A box or anti-HMGB1 antibody. These findings show that HMGB1 serves as a ferroptosis inducer by upregulating hepcidin in astrocytes and thus aggravates acute damage in the postischemic brain.

Purine metabolism-related genes and immunization in thyroid eye disease were validated using bioinformatics and machine learning

Thyroid eye disease (TED), an autoimmune inflammatory disorder affecting the orbit, exhibits a range of clinical manifestations. While the disease presentation can vary, cases that adhere to a prototypical pattern typically commence with mild symptoms that subsequently escalate in severity before entering a phase of stabilization. Notably, the metabolic activity of cells implicated in the disease substantially deviates from that of healthy cells, with purine metabolism representing a critical facet of cellular material metabolism by supplying components essential for DNA and RNA synthesis. Nevertheless, the precise involvement of Purine Metabolism Genes (PMGs) in the defensive mechanism against TED remains largely unexplored. The present study employed a bioinformatics approach to identify and validate potential PMGs associated with TED. A curated set of 65 candidate PMGs was utilized to uncover novel PMGs through a combination of differential expression analysis and a PMG dataset. Furthermore, GSEA and GSVA were employed to explore the biological functions and pathways associated with the newly identified PMGs. Subsequently, the Lasso regression and SVM-RFE algorithms were applied to identify hub genes and assess the diagnostic efficacy of the top 10 PMGs in distinguishing TED. Additionally, the relationship between hub PMGs and clinical characteristics was investigated. Finally, the expression levels of the identified ten PMGs were validated using the GSE58331 and GSE105149 datasets. This study revealed ten PMGs related with TED. PRPS2, PFAS, ATIC, NT5C1A, POLR2E, POLR2F, POLR3B, PDE3A, ADSS, and NTPCR are among the PMGs. The biological function investigation revealed their participation in processes such as RNA splicing, purine-containing chemical metabolism, and purine nucleotide metabolism. Furthermore, the diagnostic performance of the 10 PMGs in differentiating TED was encouraging. This study was effective in identifying ten PMGs linked to TED. These findings provide light on potential new biomarkers for TED and open up possibilities for tracking disease development.