Systems Analytics (United States)

The stability of therapeutic proteins is crucial for maintaining their efficacy and safety. Misfolded or aggregated proteins can accelerate aggregation by a seeding effect, altering protein stability and interactions. Heat-induced aggregates can pose a risk because they can have perturbed structures that expose aggregate-prone regions. This study explored seeding by heat-induced protein aggregates on the stability and aggregation of two therapeutic proteins, human growth hormone (hGH) and a monoclonal Antibody (Antibody A), during storage of two weeks at three relevant temperatures of 4 °C, room temperature (RT, 21-23 °C), and 40 °C. Grwoth of aggregates were seen in the sub-micron-sized aggregates by Dynamic Light Scattering at 4 °C or RT. Furthermore, heated Antibody A showed signs of reversibility of higher-order structure after storage. Samples stored at 4 °C and RT showed similar stability and aggregation behaviour. While two weeks of storage at 40 °C accelerated the structural degradation of both proteins, it did not accelerate aggregation or particle formation for either hGH or Antibody A.

Deepfake-enabled identity abuse has moved from a peripheral cyber-risk to an operational threat for U.S. banking, especially in remote onboarding, account recovery, contact-center authentication, and high-risk payment authorization. The problem is no longer limited to obvious synthetic media. Financial institutions increasingly face blended attacks that combine forged identity documents, face or voice cloning, injected video streams, social-engineering pressure, mule accounts, and adaptive retries across channels. This paper develops a publication-ready research framework for machine learning systems that identify deepfake-driven customer impersonation, authentication evasion, and identity abuse in U.S. banking environments. The study synthesizes regulatory guidance, public fraud data, biometric spoofing research, deepfake detection literature, and banking fraud analytics to propose a multimodal detection architecture spanning document forensics, face presentation attack detection, deepfake-video detection, speaker anti-spoofing, behavioral biometrics, device and network telemetry, graph-based entity resolution, and risk-calibrated decisioning. Real public evidence motivates the design: FTC data show reported fraud losses reached $12.5 billion in 2024, including $2.95 billion in imposter-scam losses, while FinCEN reported increasing suspicious activity narratives involving deepfake media targeting financial institutions. The proposed methodology treats identity abuse as a sequential, multimodal, and adversarial classification problem rather than a single-screen biometric check. The paper argues that the most effective defense is not one detector but an explainable ensemble with smart friction, human escalation, and governance aligned to NIST identity-proofing standards, anti-money-laundering expectations, and consumer-protection obligations. By linking technical detection with operational controls, the study provides a practical blueprint for authentication, resilient fraud prevention, and more trustworthy digital banking systems.

Purpose There are currently no standardized medical protocols for mass gathering events. This study sought to close that gap by creating a disaster preparedness program for the Boston Marathon, including bleeding control training and bleeding control kits for all medical tents. Design/methodology/approach All medical volunteers at the Boston Marathon are encouraged to complete online training prior to the race. Medical volunteers for the 2025 Boston Marathon completed a pre- and post-video trauma first aid training knowledge test. Medical volunteers were also graded on two simulation scenarios. Descriptive statistics, the Wilcoxon signed rank and McNemar's tests were performed. Findings Of the 1800 medical volunteers, 879 chose to complete the DLP training. Of those, 56 (6.4%) participated in the video-training study, but only 32 completed both the pre- and post-video surveys. After watching the video, overall knowledge scores increased from 8 (IQR 6, 9) to 10 (IQR 9, 10) (p-value <0.001). Of the 23 who participated in the simulation, 11 (48%) had completed the DLP trauma training video. Those who watched the training video scored better on the tourniquet (p = 0.035) and direct pressure simulation (p = 0.18). Originality/value Implementation of trauma first aid training and bleeding control kits to events such as the Boston Marathon is not only feasible but also increases the disaster preparedness confidence and knowledge of medical volunteers. This type of disaster preparedness in event medicine is crucial in mitigating future terrorist attacks at mass gatherings.

Artificial Intelligence (AI) is revolutionizing contemporary marketing by bringing strategic choice from intuition-driven campaigns to data-driven, constantly evolving systems. This document provides an overview of artificial intelligence applications that are recently being utilized. These involve predictive segmentation, real-time sentiment, conversational bots, and automated media buying. In addition, critical issues around data heterogeneity, compliance with privacy, and openness of models are highlighted. Tooffer the Predictive Personalisation Engine (PPE), a modular platform that combines deep sequential learning and reinforcement learning to offer real-time content suggestions which take into account the context in which they are presented. A multi-modal consumer dataset is fed through PPE, audiences segmented by K-Means++ clustering, future behavior is predicted by LSTM networks, and best offers are decided through an epsilon-greedy contextual bandit.

Abstract Critical Infrastructures (CI) are increasingly interdependent, forming complex dependency networks where disruptions can propagate across sectors and amplify systemic risk. Conventional risk assessment methods are limited by evaluating components in isolation and neglecting how interdependencies shape cascading exposure. This paper introduces the Risk Impact Pathway Analysis (RIPA) framework–a dynamic, dependency-aware risk assessment model that explicitly traces and quantifies risk propagation pathways within interdependent CI systems. RIPA incorporates three key propagation factors: Linkage Intensity (LI), measuring dependency strength; Resilience (RE), acting as a risk shield; and Criticality, functioning as a risk amplifier. Unlike traditional methods focused solely on initial risk (R 0 ), RIPA calculates Total Systemic Risk, revealing hidden high-risk nodes and propagation routes. The framework is validated through a multi-sector case study of the LAMAD-LLC integrated infrastructure system, demonstrating its applicability and effectiveness. Results demonstrate that incorporating interdependency criteria into risk assessment reveals significant hidden vulnerabilities, with certain infrastructure components exhibiting risk increases compared to traditional single-point assessments. This amplification effect illustrates how failures or vulnerabilities propagate through interconnected systems, creating compound risks not captured by conventional methods. Sensitivity analysis reveals that resilience enhancement interventions achieve approximately twice the risk reduction compared to equivalent modifications in either criticality scores or linkage intensity parameters. These findings validate RIPA’s capacity to provide a pathway-centric, topology-aware understanding of systemic exposure, enabling more targeted and effective infrastructure risk management strategies than traditional approaches.

The integration of high-voltage DC (HVDC) systems alongside traditional AC grids presents both challenges and opportunities for modern power networks. Central to analyzing these systems is the power flow methodology. However, existing approaches for hybrid AC/DC grids often fall short, particularly in accurately modeling the diverse control possibilities offered by converters and efficiently solving the complete system. This paper introduces a unified Generalized AC/DC Power Flow method that addresses these limitations by incorporating key features: rich converter modeling possibilities, controls integration, and most notably a unified approach that handles AC and DC grids as a single system and achieves quadratic convergence. In addition, this methodology solves the convergence issues typically encountered in other approaches. We achieve this by offering a fundamental rethinking of the role of bus types in the power flow formulation. The proposed method is validated using large systems, including the Polish transmission network (which is an openly-available 3120-bus system) and other modified IEEE test systems, demonstrating its robustness and scalability. The algorithm outperforms the benchmark by converging to the solution between 70.21% to 91.36% faster depending on the test case, and has also been fully integrated into VeraGrid, a popular open-source tool for power systems analysis.

Abstract Background GLPG3970 targeted inhibition of salt-inducible kinases (SIK) 2/3 combines anti-inflammatory and immunoregulatory activity.1,2 GLPG3970 showed positive effects on histology, endoscopy and faecal calprotectin in a 6-week phase 2a signal-detection study, consistent with preclinical findings of pro-inflammatory cytokines suppression and goblet cell protection, although without differentiation from placebo for total Mayo Clinic Score.2-5 Building on our previous ex vivo comparison of GLPG3970 with approved UC therapy tofacitinib,6 this study aimed to elucidate the mechanistic effects of GLPG3970 and delineate response patterns in bio-naive and bio-experienced patients. Methods Inflamed UC biopsies (bio-naive n = 5, bio-experienced n = 5; eMayo ≥ 2) were treated ex vivo for 24h with GLPG3970 (60 µM), vehicle, or tofacitinib (10 µM), followed by supernatant profiling (MSD 54 V-PLEX) and bulk transcriptomics by Illumina NovaSeq 6000 (Salmon + DESeq2; FoldChange &amp;gt; 1.5, FDR &amp;lt; 0.05). Weighted gene co-expression analysis (WGCNA) was run on DMSO and GLPG3970 samples only along Gene ontology (GO) pathways, with multiple testing correction (FDR&amp;lt;0.05) (Fig1A). RNA–protein correlation was assessed in matched samples (Log2FC, Spearman R, p-values). Results GLPG3970 exposure significantly reduced pro-inflammatory cytokine (TNFα, GM-CSF, IFN-γ) and chemokine (MCP-1, MIP1α, MIP1β) release in supernatant, as well as transcriptomic activity of NF-κB driven inflammatory mediators (e.g. CSF2, MMP9, CCL4), antigen presentation, collagen formation and inflammasome signalling (Fig1B-D). Transcriptomic and proteomic responses were strongly correlated (R = 0.70, p = 1.3 × 10−7), while GLPG3970 exposure could partially reverse a UC transcriptomic disease signature7 (ρ:−0.25) (Fig1E-F). WGCNA identified 18 gene expression modules, 7 of which were significantly associated with bio-experienced status [turquoise module (extracellular structure and remodelling; r = 0.58, FDR 3.6x10−5; green module (r = 0.34, FDR 0.031) Fig2A-B]. Upon GLPG3970 exposure, the green module correlation was reversed (r=-0.41, FDR 0.028) encompassing angiogenesis, cell motility and connective tissue pathways with TGFB1 identified as central hub gene (Fig2B). Modulation of this and the purple module – linked to RNA and mRNA processing - was driven by bio-experienced patients, and absent in bio-naive patients (Fig2C). Conclusion GLPG3970 effectively suppressed inflammatory signalling in UC ex vivo biopsies. Network analysis showed enhanced extracellular remodelling in bio-experienced patients, with GLPG3970 reversing connective tissue-related processes in this group. These findings emphasize the importance of considering prior treatment exposure in patient stratification and preclinical study design.

Inflation forecasting became harder after the pandemic because price formation turned faster, more uneven, and more exposed to energy shocks, supply disruptions, and changes in household spending. This paper develops a framework for real-time nowcasting and short-horizon forecasting of U.S. inflation by combining predictive analytics, artificial intelligence, and machine learning with high-frequency retail price and energy-market information observed before official CPI release dates. The empirical design uses publicly available data, centered on BLS consumer price and average-price series and EIA daily and weekly energy prices. Monthly CPI inflation is linked to within-month signals from retail gasoline prices, electricity and utility-gas measures, weekly retail motor fuel prices, and daily crude-oil benchmarks. The modeling architecture integrates econometric benchmarks with elastic net, random forest, and gradient boosting models under pseudo-real-time evaluation. Results from the empirical illustration show that feature-rich machine learning models, especially penalized regression that controls overfitting while exploiting mixed-frequency signals, can outperform naïve and autoregressive benchmarks for 2023-2024 monthly nowcasts. The findings also indicate that gasoline and broader energy signals are most valuable when inflation is turning and when conventional lag structures respond too slowly. Beyond predictive gains, the paper contributes a transparent workflow for mixed-frequency feature engineering, model governance, and explainability. High-frequency inflation surveillance should not replace official statistics or professional judgment; it should complement them by offering earlier, more adaptive evidence for central banks, financial institutions, retailers, and policy analysts. Practical implications and future research directions for richer scanner and web-scraped price systems are discussed.