Transcription factors are frequent cancer driver genes, exhibiting noted specificity based on the precise cell of origin. We demonstrate that ZIC1 exhibits loss-of-function (LOF) somatic events in group 4 (G4) medulloblastoma through recurrent point mutations, subchromosomal deletions and mono-allelic epigenetic repression (60% of G4 medulloblastoma). In contrast, highly similar SHH medulloblastoma exhibits distinct and diametrically opposed gain-of-function mutations and copy number gains (20% of SHH medulloblastoma). Overexpression of ZIC1 suppresses the growth of group 3 medulloblastoma models, whereas it promotes the proliferation of SHH medulloblastoma precursor cells. SHH medulloblastoma ZIC1 mutants show increased activity versus wild-type ZIC1, whereas G4 medulloblastoma ZIC1 mutants exhibit LOF phenotypes. Distinct ZIC1 mutations affect cells of the rhombic lip in diametrically opposed ways, suggesting that ZIC1 is a critical developmental transcriptional regulator in both the normal and transformed rhombic lip and identifying ZIC1 as an exquisitely context-dependent driver gene in medulloblastoma.

This study presents a novel and interpretable, deployment-ready framework for predicting cybersecurity incidents through item-level behavioral, cognitive, and dispositional indicators. Based on survey data from 453 professionals across countries and sectors, we developed 72 logistic regression models across twelve self-reported incident outcomes—from account lockouts to full device compromise—within six analytically stratified layers (Education, IT, Hungary, UK, USA, and full sample). Drawing on five theoretically grounded domains—cybersecurity behavior, digital literacy, personality traits, risk rationalization, and work–life boundary blurring—our models preserve the full granularity of individual responses rather than relying on aggregated scores, offering rare transparency and interpretability for real-world applications. This approach reveals how stratified models, despite smaller sample sizes, often outperform general ones by capturing behavioral and contextual specificity. Moderately prevalent outcomes (e.g., suspicious logins, multiple mild incidents) yielded the most robust predictions, while rare-event models, though occasionally high in “Area Under the Receiver Operating Characteristic Curve” (AUC), suffered from overfitting under cross-validation. Beyond model construction, we introduce threshold calibration and fairness-aware integration of demographic variables, enabling ethically grounded deployment in diverse organizational contexts. By unifying theoretical depth, item-level precision, multilayer stratification, and operational guidance, this study establishes a scalable blueprint for human-centric cybersecurity. It bridges the gap between behavioral science and risk analytics, offering the tools and insights needed to detect, predict, and mitigate user-level threats in increasingly blurred digital environments.

Our genome-wide association study identified single-nucleotide polymorphisms (SNPs) associated with estimated breeding values (EBVs) for udder traits and longevity in Holstein-Friesian cows. While no SNP was individually associated with multiple EBVs, the functional profiles of the associated genes revealed overlapping biological processes across traits, including cell signaling, transcription regulation, immune response, metabolism, and cellular maintenance. Notably, nearby SNPs BTB-01738708 and ARS-BFGL-NGS-111478 were associated with EBVlongevity and EBVudder and located near numerous genes, including GPR85, BMT2, IFRD1, and DOCK4, suggesting a potential for shared genetic influence on these traits. Our findings provide insights into the complex genetic architecture of these economically important traits and highlight the need for further research, including fine-mapping and functional genomics, to elucidate the specific variants and their effects.

The rapid integration of AI-based tools in education calls for a critical assessment of how these technologies impact student engagement. This study explores the perceived effects of AI chat tools by analyzing pre- and post-semester survey data from 724 to 642 students, respectively, across diverse disciplines and demographic groups. Initially, students reported high levels of engagement in key areas such as academic self-efficacy, autonomy, interest, and self-regulation. However, by the semester's end, all these areas showed a noticeable decline. This suggests that while AI tools offer initial benefits, their long-term effectiveness in maintaining engagement may be limited due to challenges in integrating the tools consistently or the novelty effect fading. To capture and quantify these trends effectively we used four latent engagement factors—Academic Self-Efficacy and Preparedness, Autonomy and Resource Utilization, Interest and Engagement, and Self-Regulation and Goal Setting—identified in our previous comprehensive factor analysis. Despite the overall decline, certain groups of students and specific conditions led to not only maintaining but even improving engagement levels. This study highlights the importance of tailored strategies that consider aspects such as age, discipline, usage frequency, duration of use, quality of teacher support, and the type of AI used to maximize the benefits of AI in education. These strategies can sometimes counterbalance the decline, ensuring the long-term effectiveness of AI-enhanced learning environments.

The primary goal of this research was to empirically identify and validate the factors influencing student engagement in a learning environment where AI-based chat tools, such as ChatGPT or other large language models (LLMs), are intensively integrated into the curriculum and teaching–learning process. Traditional educational theories provide a robust framework for understanding diverse dimensions of student engagement, but the integration of AI-based tools offers new personalized learning experiences, immediate feedback, and resource accessibility that necessitate a contemporary exploration of these foundational concepts. Exploratory Factor Analysis (EFA) was utilized to uncover the underlying factor structure within a large set of variables, and Confirmatory Factor Analysis (CFA) was employed to verify the factor structure identified by EFA. Four new factors have been identified: “Academic Self-Efficacy and Preparedness”, “Autonomy and Resource Utilization”, “Interest and Engagement”, and “Self-Regulation and Goal Setting.” Based on these factors, a new engagement measuring scale has been developed to comprehensively assess student engagement in AI-enhanced learning environments.

As the digital age permeates higher education, the cybersecurity awareness of university students has emerged as a pressing concern. This study examines the behavioral factors influencing students’ cybersecurity practices, developing a robust, empirically validated survey. Our research applies a comprehensive framework employing both exploratory and confirmatory factor analyses (EFA; CFA) to affirm the survey’s ability to capture the intricate dimensions of students’ cybersecurity awareness. A structural equation model (SEM) has been developed to delineate and scrutinize five key dimensions of cybersecurity behaviors within the student body. Post-validation, we utilized this model to conduct a thorough comparative analysis of cybersecurity behaviors among members of the varied student demographic that participated in the survey. The investigation included an examination of behavior across genders, age groups, academic disciplines, and cultural backgrounds, shedding light on the diverse cybersecurity behaviors that define the modern student experience. Our research ultimately strives to contribute to the enhancement of digital security in educational environments, aligning student online practices with robust security measures and nurturing a cybersecurity-aware culture in academia.

Higher education institutions are facing a major issue with student dropout rates, which is a global phenomenon that affects a significant portion of enrolled students, particularly those in their first year. The challenge is how to retain students who do not meet requirements during their first year and are at high risk of dropping out, which can have significant economic and social consequences as well as personal ramifications for the students themselves. Universities must prioritize identifying at-risk students and providing targeted assistance to prevent them from leaving the system. Machine learning (ML) models have proven effective in identifying students at risk of dropping out with a high degree of accuracy. In this study, we aim to construct a machine learning model using data extracted from the administration system (Neptun) to predict student dropout rates in the Business Informatics BSc course at the Faculty of Finance and Accounting of Budapest Business School.

Supplementary Figures S1-S3; Supplementary Tables S1-S2.

The process for building effective machine learning models that predict the learning success of university students, the competences of the actors involved in model building, and the main factors and conditions that influence the reliability of the predictions are reviewed in this paper. It is shown that, in addition to the site-level and course-level indicators commonly used in the literature for prediction, significantly more accurate predictions can be made by introducing so-called chapter-level indicators. These chapter-level indicators are closely linked to the content structure of the subject under study, the hierarchy of its chapters and the learning resources and student activities used in them.Specifically, we make suggestions to the course instructor about the conditions under which there is a hope of obtaining reliable predictions of the student's chances of success or failure. We show how the use of a previously trained model for a newly started similar course requires caution. Even relatively small differences, such as a change in the minimum score to be achieved, a difference in the form of study (full-time or correspondence course), or a difference in the number of compulsory mid-term tests, can cast doubt on the validity of our predictions. We also discuss details of model training that affect the goodness of fit of machine learning models.

We investigated the effect of deep brain stimulation on dynamic balance during gait in Parkinson's disease with motion sensor measurements and predicted their values from disease-related factors. We recruited twenty patients with Parkinson's disease treated with bilateral subthalamic stimulation for at least 12 months and 24 healthy controls. Six monitors with three-dimensional gyroscopes and accelerometers were placed on the chest, the lumbar region, the two wrists, and the shins. Patients performed the instrumented Timed Up and Go test in stimulation OFF, stimulation ON, and right- and left-sided stimulation ON conditions. Gait parameters and dynamic balance parameters such as double support, peak turn velocity, and the trunk's range of motion and velocity in three dimensions were analyzed. Age, disease duration, the time elapsed after implantation, the Hoehn-Yahr stage before and after the operation, the levodopa, and stimulation responsiveness were reported. We individually calculated the distance values of stimulation locations from the subthalamic motor center in three dimensions. Sway values of static balance were collected. We compared the gait parameters in the OFF and stimulation ON states and controls. With cluster analysis and a machine-learning-based multiple regression method, we explored the predictive clinical factors for each dynamic balance parameter (with age as a confounder). The arm movements improved the most among gait parameters due to stimulation and the horizontal and sagittal trunk movements. Double support did not change after switching on the stimulation on the group level and did not differ from control values. Individual changes in double support and horizontal range of trunk motion due to stimulation could be predicted from the most disease-related factors and the severity of the disease; the latter also from the stimulation-related changes in the static balance parameters. Physiotherapy should focus on double support and horizontal trunk movements when treating patients with subthalamic deep brain stimulation.

Sir, We would like to present a deep brain stimulation (DBS) surgery case wherein we had to reconsider the target for lead implantation due to a developmental venous anomaly (DVA). Our 45-year-old male patient was diagnosed with Parkinson's disease ten years prior to surgery. He was selected for subthalamic DBS implantation, and he underwent the standard examination protocol. The preoperative MRI revealed a DVA in the region of the left basal ganglia. As the majority of the vessels were found in the trajectory running toward the subthalamic nucleus, based on the major symptoms of the patient, we decided to place the leads into the internal globus pallidus (GPi) as GPi DBS has long-term efficacy that is comparable with STN DBS.[1,2,3] Trajectories were selected in a way so as to avoid the branches of the DVA. On the left side, only one safe trajectory could be defined for microelectrode recording and final lead placement [Figure 1]; on the right side, we applied the usual 5-channel microelectrode trajectory planning. The implantation of the right side lead was conducted in the usual way using 5-channel microelectrode recording.Figure 1: Preoperative planning avoiding the branches of the DVAAfter surgery, the patient was observed in the general ward. No neurological deficit occurred. Native thin-slice CT was done postoperatively revealing no surgical adverse event, such as bleeding or lead misplacement. The stimulation started approximately one month after implantation with good results regarding the patient's symptoms. Follow-up MRI showed the electrode passing safely between the branches of the DVA [Figure 2].Figure 2: Postoperative MRI revealing the lead position related to the branches of the DVA (white arrows)In this case, a DVA made surgical planning difficult. DVAs can be present in various locations, including the cerebrum, cerebellum, and brainstem,[4] and they are quite common accidental findings. The presence of a DVA in the trajectory or target sites of DBS surgeries can jeopardize the feasibility of the planned operation. However, we could find a safe way to place the DBS lead avoiding conflict with the branches of the DVA; in turn, we had to modify our initial plan from STN DBS to GPi DBS. Our case proves that even if in certain cases unexpected or unusual structures might act as a contraindication for DBS surgery, using meticulous planning and precise devices some of these difficulties can be overcome. Financial support and sponsorship Nil. Conflicts of interest There are no conflicts of interest.

Background Balance impairment in Parkinson’s disease is multifactorial and its changes due to subthalamic stimulation vary in different studies. Objective We aimed to analyze the combination of predictive clinical factors of balance impairment in patients with Parkinson’s disease treated with bilateral subthalamic stimulation for at least one year. Methods We recruited 24 patients with Parkinson’s disease treated with bilateral subthalamic stimulation and 24 healthy controls. They wore an Opal monitor (APDM Inc.) consisting of three-dimensional gyroscopes and accelerometers in the lumbar region. We investigated four stimulation conditions (bilateral stimulation OFF, bilateral stimulation ON, and unilateral right- and left-sided stimulation ON) with four tests: stance on a plain ground with eyes open and closed, stance on a foam platform with eyes open and closed. Age, disease duration, the time elapsed after implantation, levodopa, and stimulation responsiveness were analyzed. The distance of stimulation location from the subthalamic motor center was calculated individually in each plane of the three dimensions. We analyzed the sway values in the four stimulation conditions in the patient group and compared them with the control values. We explored factor combinations (with age as confounder) in the patient group predictive for imbalance with cluster analysis and a machine‐learning‐based multiple regression method. Results Sway combined from the four tasks did not differ in the patients and controls on a group level. The combination of the disease duration, the preoperative levodopa responsiveness, and the stimulation responsiveness predicted individual stimulation-induced static imbalance. The more affected patients had more severe motor symptoms; primarily, the proprioceptive followed by visual sensory feedback loss provoked imbalance in them when switching on the stimulation. Conclusions The duration of the disease, the severity of motor symptoms, the levodopa responsiveness, and additional sensory deficits should be carefully considered during preoperative evaluation to predict subthalamic stimulation-induced imbalance in Parkinson’s disease.

Sonic hedgehog medulloblastoma encompasses a clinically and molecularly diverse group of cancers of the developing central nervous system. Here, we use unbiased sequencing of the transcriptome across a large cohort of 250 tumors to reveal differences among molecular subtypes of the disease, and demonstrate the previously unappreciated importance of non-coding RNA transcripts. We identify alterations within the cAMP dependent pathway (GNAS, PRKAR1A) which converge on GLI2 activity and show that 18% of tumors have a genetic event that directly targets the abundance and/or stability of MYCN. Furthermore, we discover an extensive network of fusions in focally amplified regions encompassing GLI2, and several loss-of-function fusions in tumor suppressor genes PTCH1, SUFU and NCOR1. Molecular convergence on a subset of genes by nucleotide variants, copy number aberrations, and gene fusions highlight the key roles of specific pathways in the pathogenesis of Sonic hedgehog medulloblastoma and open up opportunities for therapeutic intervention.

In this paper the issue of bias-variance trade-off in building and operating Moodle Machine Learning (ML) models are discussed to avoid traps of get-ting unreliable predictions. Moodle is one of the world’s most popular open source Learning Management System (LMS) with millions of users. Although since Moodle 3.4 release it is possible to create ML models within the LMS system very few studies have been published so far about the conditions of its proper application. Using these models as black boxes hold serious risks to get unreliable predictions and false alarms. From a comprehensive study of differently built machine learning models elaborated at the University of Dunaújváros in Hungary, one specific issue is addressed here, namely the in-fluence of the size and the row-column ratio of the predictor matrix on the goodness of the predictions. In the so-called Time Splitting Method in Moo-dle Learning Analytics the effect of varying numbers of time splits and of predictors has also been studied to see their influence on the bias and the variance of the models. An Applied Statistics course is used to demonstrate the consequences of the different model set up.

The relation between an educational target and a set of predictors related to the learners and their learning activities in a given learning context can be investigated by predictive Machine Learning (ML) modelling. For courses with many students and with predictors that well reflect the specialties of the courses, the predictive power of “classical” ML models generally meets expectations. At the same time, even large universities have several courses where the small number of students does not allow the use of “classical” ML models, although the need to forecast student performance also appears for these courses. In this study, considering the research on the Applied Statistics course with the participation of 56 full-time students at the University of Dunaujvaros, we present various model building techniques that can be used to increase the predictive power of models. We systematically show different model building technics starting from less effective technics to more developed ones. These developed ones are applicable even for small or mid-sized university courses producing a monotonically increasing good performance metrics in time. The conditions and limits of their applicability are also discussed.

In today’s modern world, the pace of technological development can be con-sidered exponential. Education must constantly adapt to this dynamic devel-opment. It must be able to innovate, to use modern tools and methods that are effectively integrated into the learning process. Education should provide an appropriate learning environment to meet changing needs, one way of which is e-learning. This environment can be an excellent support for the learning process; however, it will hardly be effective without developing the right student learning attitude. The e-learning environment gives freedom and independence to the individual, at the same time. For the learner's individual endeavor to be successfully completed as expected, control and continuous feedback are needed. One way to do this is through self-quizzing. Self-quizzes, divided into units, related to the learning material, with appropriate difficulty and amount, can help to understand and engrave the processed material, and thus improve the effectiveness of learning. Self-quizzes create the opportunity for imme-diate feedback, which is a very important feature of an e-learning environ-ment. According to many scientific research, immediate feedback can greatly help maintain interest and motivation, and quizzes are suitable tools for this purpose. In our research, we sought to answer the question of how the continuous, self-monitoring practice opportunity provided by online quizzes affects stu-dent achievement. In the case of an online course held at the University of Dunaújváros in 2019, we examined whether students who continuously per-form self-quizzes will be more effective by the end of the learning process than their peers who are less receptive to independent self-quizzing. We also addressed the effect of the nature of time spent on self-reflexive quizzing on learning success.

Objective: Bradykinesia has been associated with beta and gamma band interactions in the basal ganglia-thalamocortical circuit in Parkinson's disease. In this present cross-sectional study, we aimed to search for neural networks with electroencephalography whose frequency-specific actions may predict bradykinesia. Methods: Twenty Parkinsonian patients treated with bilateral subthalamic stimulation were first prescreened while we selected four levels of contralateral stimulation (0: OFF, 1-3: decreasing symptoms to ON state) individually, based on kinematics. In the screening period, we performed 64-channel electroencephalography measurements simultaneously with electromyography and motion detection during a resting state, finger tapping, hand grasping tasks, and pronation-supination of the arm, with the four levels of contralateral stimulation. We analyzed spectral power at the low (13-20 Hz) and high (21-30 Hz) beta frequency bands and low (31-60 Hz) and high (61-100 Hz) gamma frequency bands using the dynamic imaging of coherent sources. Structural equation modelling estimated causal relationships between the slope of changes in network beta and gamma activities and the slope of changes in bradykinesia measures. Results: Activity in different subnetworks, including predominantly the primary motor and premotor cortex, the subthalamic nucleus predicted the slopes in amplitude and speed while switching between stimulation levels. These subnetwork dynamics on their preferred frequencies predicted distinct types and parameters of the movement only on the contralateral side. Discussion: Concurrent subnetworks affected in bradykinesia and their activity changes in the different frequency bands are specific to the type and parameters of the movement; and the primary motor and premotor cortex are common nodes.