Becton Dickinson (Canada)

Xanthomonas campestris is a diverse bacterial plant pathogen that causes disease on many critical crops. However, the mechanisms underlying its virulence have remained elusive. We paired a quantitative virulence screen with a bacterial genome-wide association study (GWAS) to uncover the genetic underpinnings of virulence on this model pathogen. We found that a diverse panel of 67 X. campestris strains varied substantially in their ability to grow and cause disease on the model host A. thaliana Col-0. However, its virulence on this host was not a monophyletic trait. This opened the door to the application of a GWAS to test for convergent genetic loci underlying virulence in X. campestris. After correcting for phylogenetic structure, only four accessory gene families were positively associated with X. campestris virulence: a xopJ family effector, a hypothetical gene, and two genes associated with transposable element mobility. These four genes mapped to a Tn3-family transposon that appears to have been horizontally transferred across multiple distinct X. campestris lineages, enabling the emergence of divergent pathogens in these lineages with the same underlying genetic architecture. Further analysis revealed that these transposons were often nested within larger transposons that also contained additional virulence genes, including effectors xopE, avrBs3, and xopAH, which may further contribute to host specific adaptation within independent lineages. This study demonstrates that the horizontal transfer of an effector harbouring transposon is a primary determinant of X. campestris virulence on the model host A. thaliana. It also highlights the utility of GWASs for identifying bacterial loci associated with virulence, which has far-reaching implications for pathogen surveillance and the engineering of resistant crops.

We introduce SC-MGM, an innovative framework for spatiotemporal community recognition within video scene graphs. Our objective is to find cohesive communities of objects that endure and evolve over time inside a succession of video frames represented as scene graphs. SC-MGM incorporates three complementary elements: a Graph Neural Network (GNN) for encoding intra-frame relational structures, a differentiable soft clustering module for seamless and trainable community assignment, and a Markov transition module that represents the temporal evolution of community prototypes. This integrated architecture facilitates comprehensive optimization of structural precision and temporal coherence. We assess SC-MGM using two extensive datasets, Action Genome and VIDOR, and exhibit consistent enhancements over current baselines regarding modularity, clustering accuracy, and interpretability. Qualitative representations demonstrate that our model accurately represents fluid community trajectories in accordance with semantic dynamics. The suggested method offers a systematic and comprehensible basis for dynamic scene comprehension, with possible applications in action analysis, video condensation, and relational inference.

Suboptimal blood volume collection negatively affects pathogen recovery and time to pathogen detection in clinical blood culture samples. This study evaluated the performance of the ultrathin-wall cannula device of the BD Vacutainer UltraTouch Push Button Blood Collection Set (UltraTouch; BD Life Sciences, Franklin Lakes, NJ) in increasing the blood volume collected compared to regular cannulas. In this retrospective study, the volume of blood collected in negative BD BACTEC Plus Aerobic/F Culture Vials prior to and following blind conversion to UltraTouch was measured from blood background metabolic activity signal acquired during the initial incubation period in the BD BACTEC FX instrument using a custom software application (patent pending). Differences in average fill volume between pre- and post-conversion were evaluated overall and by month. A separate logistic regression was also performed to model the relationship between monthly average blood volume increase (mL) and sample positivity rate. Prevalence rates for identified organisms were also assessed. The blood fill volume of 13,356 pre-conversion and 119,971 post-conversion bottles was compared. The average post-conversion fill volume was significantly higher than that of pre-conversion (7.56 mL [95% CI, 7.54-7.59] vs. 5.68 mL [95% CI, 5.62-5.74], respectively). This 1.88 mL increase in collected blood volume increased sample positivity from 7.0% to 7.7% (P = 0.002; odds ratio [OR], 1.11; 95% CI, 1.04-1.19). This study demonstrated that using an ultrathin-wall cannula blood collection device can significantly increase blood culture volume and sample positivity, supporting improved identification of patients suffering from bloodstream infections.IMPORTANCESubstandard blood fill volume practices in clinical settings negatively affect pathogen recovery which, in turn, may impact patients' care. Whereas professional education regarding the importance of adhering to clinical guidelines has been helpful in partly remediating the issue, the role of blood collection set technologies in optimizing blood culture volume has not been studied as thoroughly. Our study assessed the performance of the BD Vacutainer UltraTouch Push Button Blood Collection Set, which is designed with an ultrathin-wall cannula, to improve blood culture volume.

This study investigated EEG microstate dynamics in trigeminal neuralgia (TN) patients to understand the central nervous system's contribution to this neuropathic pain condition. Despite TN's traditional classification as a peripheral neuropathy, altered brain network organization may play a critical role in pain chronification and treatment resistance, making EEG microstates a valuable tool for capturing these dynamic neural signatures. We analyzed resting-state EEG recordings from 14 healthy individuals and 36 TN patients through a systematic analytical pipeline. After preprocessing with a fifth-order Butterworth band-pass filter (10-40Hz), we employed k-means clustering to identify four distinct microstate configurations (4-7 states). From these configurations, we extracted temporal parameters (duration, occurrence, coverage, and mean global field power) and constructed transition probability matrices to characterize brain state dynamics. These features were then evaluated using ANOVA and utilized in machine learning classification models to assess their discriminative potential. TN patients demonstrated distinct microstate abnormalities, including dramatically increased durations in specific microstates (5-6 times longer than controls) and consistently reduced global field power (0.03 vs. 0.35). Transition probability analyses revealed striking differences between groups: healthy subjects exhibited balanced bidirectional transitions (particularly B↔C at ~ 31-33%), whereas TN patients showed highly asymmetric patterns with strong directional flows (B→A: 33.5%, C→A: 35.2%, D→A: 34.4% in 4-state model). Most notably, state E functioned as a distinctive "sink" in TN patients, receiving significant transitions while exhibiting minimal outward flow (only 2.8-3.6% in 7-state model), suggesting trapped neural processing. Machine learning classification achieved exceptional discrimination between groups (91.9% accuracy with SVM), with optimal performance using four features in simpler 4-state models. Our findings establish EEG microstate analysis as a promising neurophysiological framework for understanding TN pathophysiology, revealing objective biomarkers that reflect altered brain network dynamics rather than simply peripheral nerve dysfunction. These distinctive microstate patterns align with contemporary pain processing theories and offer potential applications in diagnosis, treatment monitoring, and development of novel therapeutic approaches targeting the central mechanisms of TN.

PurposeThis paper aims to discuss the ethical permissibility of using publicly available online texts in the training of AI. This practice has facilitated and accelerated the growth of AI technology, but it has also drawn accusations of exploiting authors and violating their intellectual property rights.Design/methodology/approachThe discussion is based on the Kantian theory of ethics, a theory that is grounded in the duty to preserve and empower the autonomy of the human will.FindingsFollowing from the duty to support human autonomy, the article makes two claims: First, AI should be granted the right to learn from public texts because the more AI learns, the better it assists humans in expanding their will. Second, AI’s right to learn should be conditional, not absolute. The article articulates three conditions that should be met prior to granting an AI system the right to learn from public texts: AI that uses public data must be freely available to the public, AI (and its developers) ought to educate the public on the uses and misuses of AI, and AI (and its developers) must remove personal information from the training data.Originality/valueUnder what conditions is it justifiable to use someone’s text without their permission? This very question is currently being examined in courts in the many lawsuits filed against AI companies. An ethical analysis of this question helps bridge the gap between the perspective of AI companies that claim to be using public texts for the public good and the perspectives accusing them of violating authors’ copyrights.

Software testing is very important in software development to ensure its quality and reliability. As software systems have become more complex, the number of test cases has increased, which presents the challenge of executing all the tests in a limited time frame. Various test case prioritization techniques have been introduced to solve this problem. These methods aim to identify and implement the most critical tests first. In this paper, we propose an implementation of a dynamic test case prioritization strategy to improve software quality by increasing code coverage with special attention to edge case handling. Edge case test prioritization is a technique that improves test efficiency by selecting extreme case scenarios that can reveal critical bugs or unexpected behavior early in development, improving overall software reliability and dependability. In order to prioritize test cases, this paper presents a regression-based method that makes use of machine learning algorithms. The approach leverages previous performance data to optimize regression testing efficiency by examining variables like test time and execution status. Performance evaluations, when compared against industry standards and cutting-edge techniques, show how effective these algorithms are at correctly prioritizing test cases and identifying faults. This study offers simplified yet reliable solutions for regression testing optimization by shedding light on the efficacy of regression algorithms, such as Random Forest and decision trees.