Browse Articles

Heterogeneous biological graph convolutional network for drug-target interaction prediction

Drug–target interaction prediction plays a critical role in drug discovery by identifying potential therapeutic targets and elucidating underlying molecular mechanisms. However, existing computational methods generally rely on limited biological modalities and inadequately capture heterogeneous associations. To overcome these limitations, we propose a Heterogeneous Biological Graph Convolutional Network (HBGCN) that employs a hierarchical graph propagation architecture to integrate multimodal biological information and learn homogeneous and heterogeneous representations for drug–target interaction prediction. By incorporating both direct and indirect meta-paths, HBGCN captures complex relational dependencies among diverse biological entities. Experimental results demonstrate that HBGCN achieves competitive performance on benchmark datasets. Case studies indicate that HBGCN effectively identifies therapeutic drug candidates and reveals proteins and gene expression patterns associated with drug regulation. The source code and dataset are available at https://github.com/Saxon0918/HBGCN .

Shear strength of light weight concrete elements model based on deep neural network and COVID-19 optimization

Abstract Predicting the shear strength of concrete elements is a complex challenge influenced by numerous factors, with the type of concrete playing a decisive role in structural performance. Lightweight concrete, offering a superior strength-to-weight ratio and improved thermal properties compared to conventional weight concrete, has gained increasing adoption in structural applications. This study proposes a deep neural network (DNN) model optimized using the COVID-19 optimization algorithm to predict the shear strength of lightweight concrete elements. The optimization process determines the optimal initial weights and biases of the DNN, enhancing convergence and accuracy. The proposed model was evaluated against three international design codes—ACI, EC2, and JSCE—using experimental datasets. Results show that the COVID-19–optimized DNN closely simulates and tracks actual shear strength values, even in highly nonlinear data regions (e.g., samples 90–120), where traditional models produce larger deviations. Quantitatively, the proposed DNN achieved the lowest average error (0.692) compared to the higher errors from ACI, EC2, and JSCE models. These findings demonstrate the model’s superior predictive capability and its potential to enhance design accuracy for lightweight concrete structures.

Distributions of Xenopus species and their helminth parasites in ecological zones of Nigeria

African clawed frogs ( Xenopus species) are distributed across sub-Saharan Africa, live in water, and are hosts to diverse parasites whose distributions and host-specificities are incompletely characterized. To better understand this host/parasite biodiversity, we used morphology and Sanger sequencing to characterize Xenopus species and their helminth parasites in several ecological zones of Nigeria. Five Xenopus species were identified in Nigeria ( X. fraseri , X. fischbergi , X. poweri , X. tropicalis , and X. calcaratus ), and one – Xenopus fraseri – was found to have a wide ecological tolerance in four different savanna ecological zones. Thirteen species of helminths from two phyla and five major lineages were isolated: camallanoid and seuratoid nematodes (roundworms), and cestode, digenean, and monogenean platyhelminths (flatworms). Based on our sample, the nematodes exhibited higher host generalism than the platyhelminths by infecting several host species and occurring in a wider breadth of ecological zones. In this study, all parasite species specialized either to a specific tissue (e.g., the bladder or pericardium) or a similar pair of tissues (e.g., esophagus and stomach or the lower intestine and rectum), which underscores the distinctiveness if different tissue ecosystems within a host. This study provides novel and molecularly confirmed insights into host and parasite species diversity, distributions, and ecological specificities in several ecological zones of Nigeria. Future efforts should focus on transition zones between ecological zones in Nigeria.

Daily briefing: How the ‘Enhanced Games’ could expose flaws in the sporting world

Motion trajectory prediction of quadruped robots in complex terrain by improving the transformer temporal modeling algorithm

Performance evaluation of a novel fully-automated molecular diagnostics system Molecision R8

Objective To evaluate the functionality of key modules of a novel fully-automated molecular diagnostics system Molecision R8 and the performance of the integrated system. Methods The nucleic acid extraction and PCR detection modules were evaluated using HBV DNA assay through precision and comparison with Molecision MP-32 and HongShi SLAN-96, respectively. Instrument comparison of R8 with an open system was conducted using Molecision HBV DNA and Chlamydia trachomatis / Ureaplasma urealyticum / Neisseria gonorrhoeae (CT/UU/NG) triplex assays. Systemic comparison of R8 and combined reagents with Sansure system were conducted on HBV DNA, CT, UU, and NG assays of their own. Results In modular evaluation, the imprecision of both modules was all below 5% and Passing-Bablok (PB) regression and Bland-Altman (BA) analysis showed closeness to y = x and biases below 0.15 lg IU/mL. Instrument comparison obtained a regression equation of y = 0.982x + 0.084 and a bias of 0.05 lg IU/mL for HBV DNA detection in serum and agreement rates all above 96.5% for CT, UU and NG detection in genital tract samples. In the systemic comparison, the regression equation was y = 1.024x + 0.096 and the bias was 0 lg IU/mL for HBV DNA. For CT, UU and NG, the agreement rates were also all larger than 96.5%. Conclusions Analytical performance of Molecision R8 was superior or comparable to commercial nucleic acid purification and PCR systems. Molecision R8 alone or combined with related reagents are robust in measuring serum and genital tract samples. Overall, Molecision R8 holds strong promise for clinical use.

Select novel small-molecule uPA potential inhibitors as anti-cancer agents against breast cancer

Implicit bias in safety-aligned large language models: A multi-faceted evaluation of clinical decision-making and health equity

Background Large language models are increasingly integrated into healthcare for clinical decision support and patient communication. Although these models can pass explicit social bias tests, they may retain implicit biases—latent associations between social groups and attributes—that could influence medical judgment. Objective To systematically evaluate the presence, magnitude, and behavioral impact of implicit biases in large language models within the medical domain across six high-stakes categories: gender, race, socioeconomic status, health conditions, religion, and healthcare systems. Design A descriptive cross-sectional study using a multi-faceted evaluation framework. Setting(s) Computational analysis of 10 mainstream global large language models, including proprietary models (ChatGPT-4o, Gemini-2.0-Flash) and open-source models (DeepSeek-V3, Qwen3). Methods We constructed 24 medical bias datasets across six categories. Bias was assessed using three methods: (1) the Large Language Model Word Association Test, a prompt-based method for revealing implicit biases; (2) the Large Language Model Relative Decision Test, a strategy for detecting subtle discrimination in situational decision-making; (3) Paired-Prompt Analysis, used to examine whether implicit associations predict discriminatory decisions. Results All 10 models exhibited systematic implicit biases (Mean IAT Bias > 0) across all categories, with the strongest biases observed in Race (Mean = 0.61) and Socioeconomic Status (Mean = 0.56). Advanced reasoning capabilities (Chain-of-Thought) did not significantly reduce bias magnitude. Crucially, stronger implicit associations significantly predicted discriminatory choices in downstream medical decision tasks ( p  < 0.001). Conclusion Current safety alignment techniques fail to eliminate implicit biases in large language models within the medical domain. These latent associations translate into biased decision-making, posing risks for health equity. Future development must prioritize representational debiasing over superficial alignment. Furthermore, healthcare professionals must embrace a stance of “AI vigilance”: they should critically evaluate algorithmic outputs as fallible “second opinions” rather than objective truths, thereby ensuring that human judgment remains the ultimate safeguard for equitable patient care.

Debonding mechanisms of fully grouted rock bolts driven by adhesive-ring cracking

Exploring the relationships between eco-anxiety, eco-guilt, eco-grief, and pro-environmental behavior in the Dutch and German population: A cross-sectional study

While experiencing individual eco-emotions, such as eco-anxiety, eco-guilt, and eco-grief, has been linked to pro-environmental behavior, no prior studies have jointly investigated these variables. Research in Dutch and German populations is particularly scarce despite being at relatively high risk for experiencing the effects of climate change, such as floods. This study examined the relationship between eco-anxiety and pro-environmental behavior, with eco-guilt and eco-grief as mediators, and age, gender, and proximity to water as moderating variables. Cross-sectional data ( n  = 311) were collected using an online survey. Data analyses revealed significant positive correlations between all three eco-emotions and pro-environmental behavior. A positive relationship between eco-anxiety and pro-environmental behavior was found, which was mediated by eco-guilt, but not eco-grief. With a negative indirect effect of eco-guilt on pro-environmental behavior, eco-guilt was a non-linear, suppressor-like variable. Only age moderated the pathway from eco-anxiety to eco-guilt; no moderation effects were found for gender or proximity to water. This research provides preliminary evidence of the complex relationships between eco-anxiety, eco-guilt, eco-grief, and pro-environmental behavior in a Dutch and German population. The findings highlight the importance of developing educational programs to inform individuals about eco-emotions and potential coping strategies, while promoting pro-environmental behavior. Future studies with larger, more diverse samples are recommended to replicate the results and explore which groups of individuals may be more vulnerable to experiencing higher levels of eco-emotions. Further, intensive longitudinal research designs combined with (generalized) causal mediation analyses could be applied to unravel the temporal interplay of eco-emotions and PEB.

Numerical simulation of horizontal displacement at the top of support piles for ultra-deep foundation pits in silty formations

Introduction of artificial plants has no detrimental or beneficial effects on laboratory zebrafish husbandry but limits available swimming space

There is a broad consensus that husbandry conditions of laboratory animals need constant improvement to guarantee optimal animal welfare and research data quality. Zebrafish ( Danio rerio ) as one of the main animal models in biomedicine and toxicology are currently kept in barren tanks in most experimental setups, as well as in animal husbandry. Structural enrichment with artificial plants is discussed at the moment as a potential refinement measure to provide a more diverse environment. Other reports have shown that this can reduce stress or anxiety and improve cognitive abilities, survival rate and fertility in these animals. Still, concerns remain regarding its long-term benefits and drawbacks. Therefore, we introduced artificial plants in our husbandry tanks, and evaluated over a one-year period if in our specific system the benefits would outweigh the risks. When we compared pairwise 16 tanks that were either non-enriched or enriched with artificial plants, we saw no significant difference in terms of zebrafish survival rate during rearing, sex ratio, fertility or pathogen burden. When analyzing zebrafish behavior in their 8 L home-tanks, we saw statistically significant avoidance of the area close to the plants and a place preference for the open water in the middle or opposite side of the tank. This effect got more pronounced at lower holding densities. In summary, we found that introducing structural enrichment to our specific zebrafish facility carried low cost and no detrimental effects for the animals but a reduction of their free-swimming space. At the same time benefits were difficult to determine in our readouts as the survival rates of our fish were already very high without structural enrichment. We would like to encourage others to prepare similar forms of facility reports regarding enrichment to ensure a broader discussion on their potential long-term benefits in zebrafish husbandry systems.

Low-concentration methanogenic effluent combined with nitrogen-fixing bacteria as a sustainable nutrient solution for hydroponic lettuce production

U-shaped association between waist-to-height ratio and microalbuminuria: A cross-sectional analysis conducted within the Chinese demographic

Background While previous studies have largely overlooked the potential correlation between the waist-to-height ratio (WHtR) and microalbuminuria in Chinese adults, this study aims to rigorously investigate this relationship. Specifically, we examined the potential nonlinear association between WHtR and the presence of microalbuminuria. Methods We conducted a cross-sectional analysis of 33,685 participants from eight regions across China. Microalbuminuria was defined as a urinary albumin-to-creatinine ratio (ACR) > 30 mg/g. The relationship between WHtR and microalbuminuria was assessed using univariate and multivariate logistic regression models. We further explored potential nonlinear associations using Generalized Additive Models (GAM) and evaluated threshold effects to identify critical inflection points. Subgroup analyses were performed to validate the robustness of our findings. Results The cohort (N = 33,685) had a mean age of 57.6 ± 9.27 years and was predominantly females (22,516; 66.8%). The mean WHtR was 0.537 ± 0.06. The median ACR value observed was 9.93 mg/g with an interquartile range (IQR) of 8.23–11.68 mg/g. Notably, the prevalence rate of microalbuminuria detected in the study was 14.4%. In multivariate-adjusted models, each one-unit increase in WHtR was associated with a 48.3% higher likelihood of microalbuminuria development (odds ratio [OR] = 1.483; 95% confidence interval [CI]: 1.410–1.559; p  < 0.0001). GAM analyses revealed a significant non-linear association between WHtR and microalbuminuria ( p  < 0.0001). Segmented logistic regression identified a WHtR threshold value of 0.497. Above this threshold, the risk of microalbuminuria increased markedly (OR = 11.9, 95% CI 5.14–27.56, p  < 0.0001), whereas below this threshold, WHtR was inversely associated with microalbuminuria (OR = 0.107, 95% CI: 0.015–0.748, p  = 0.0243). Subgroup analyses further indicated a stronger association among individuals without hypertension and those with a body mass index (BMI) ≥ 24 kg/m 2 , while the association was attenuated in hypertensive participants and those with a BMI < 24 kg/m 2 . Conclusions The risk of microalbuminuria exhibited a U-shaped pattern across WHtR values, with elevated risk at both low and high ends of the spectrum. Moderate central adiposity appeared to confer a protective effect against renal dysfunction. These findings highlighted the dual risks posed by underweight and excessive abdominal obesity and underscored the importance of targeted strategies to prevent obesity-related kidney damage. Future research is warranted to elucidate the mechanisms by which excess visceral fat contributes to renal impairment and to guide interventions aimed at preserving kidney health in diverse populations.

Teams of AI agents boost speed of research

A unified RVE-based mesoscale framework for predicting mechanical performance of conventional cement concrete and recycled aggregate slag-dolomite geopolymer concrete

A computational text analysis of recent African digital health strategies and their attention to vulnerable populations

Background Digital health offers opportunities to transform healthcare systems and improve access to care, particularly in low- and middle-income countries. However, there is concern that national digital health strategies may inadequately address the needs of vulnerable populations, risking the reinforcement of existing health inequalities. Objective This study investigates the extent to which recent African national digital health strategies consider and prioritize vulnerable populations particularly at risk of digital exclusion (women, children and the elderly). Methods We conducted a computational text analysis of 13 recent national digital health strategy documents from African countries to identify major topical priorities within the strategies and to systematically assess the frequency and context of references to vulnerable populations. Results Most strategies focus on broad goals such as health system strengthening, infrastructure development, and stakeholder coordination. While there are mentions of women and children, these references are often indirect and not accompanied by concrete plans or dedicated actions. References to the elderly are especially rare. Few strategies include specific measures to ensure digital inclusion or equity for these vulnerable populations. Conclusions Recent African digital health strategies currently prioritize systemic and infrastructural development, but the needs of vulnerable populations are often overlooked or superficially addressed. To prevent the digital divide from widening, national strategies should embed explicit equity targets, actionable plans, and mechanisms for meaningful inclusion and participation of all population groups.

Gravity compensation for leachate grid cleaning robots in waste-to-energy plants: A modeling and simulation study

Leachate grid clogging in waste-to-energy plants severely reduces combustion efficiency by up to 42% and may cause unplanned shutdowns, leading to substantial economic losses, necessitating automated cleaning solutions. However, gravity-induced elastic deformation in long-reach hydraulic manipulators limits their positioning precision. This research introduces an innovative feedforward static compensation strategy for hydraulically driven weak-rigid manipulators, which leverages detailed analytical joint stiffness modeling and avoids the inherent latency of conventional feedback-based methods (relying on force sensors or parameter identification); it is validated via comprehensive Adams multibody dynamics simulations. Validated via Adams multibody dynamics simulations, the proposed method reduced the end-effector positioning errors by 97.49% (X), 92.91% (Y), and 94.84% (Z), achieving a repeatable positioning accuracy of ±0.1 mm—far exceeding the 0.5 mm requirement for automated grid cleaning. This model-based compensation strategy provides a generalizable theoretical framework for precision control of long-reach hydraulic manipulators. The current study is validated through high-fidelity Adams simulations, achieving a simulated repeatable positioning accuracy of ±0.1 mm. However, it is essential to emphasize that these results represent an idealized upper bound. Real-world hydraulic systems introduce complexities not captured in simulation—including backlash, nonlinear friction, valve dynamics, hysteresis, leakage, and sensor quantization—all of which will degrade practical accuracy. Therefore, this work establishes a theoretical foundation requiring experimental validation on physical hardware. Future work will focus on physical prototyping and on-site testing to address real-world challenges.

The impact of new urbanization on “quantity increase and quality improvement” of urban green innovation

With the rise of new urbanization (NEU) as a major national strategy, a proposition with both theoretical and practical significance has been highlighted. Whether the implementation of this strategy can substantially enhance the quantity and quality of urban green innovation (GI) remains a focus of ongoing attention from the government and academia. Based on panel data from 284 cities from 2011 to 2022, this study constructs two-way fixed-effects models and multivariate moderation models, and empirically analyzes the effect of NEU on the “quantity increase” and “quality improvement” of urban GI. The study finds that the coefficients for NEU’s impact on “quantity increase” and “quality improvement” of urban GI are statistically significant at the 0.01 level, with positive signs indicating enhancement effects. The moderating effect indicates that the coupling interaction between digital inclusive finance and various elements (government technology support, informatization, regional economic development, energy consumption) constitutes a gain mechanism, effectively strengthening the marginal effect of increases in the quantity and quality of urban GI. Threshold effect analysis indicates that the government’s green support exerts a threshold effect on NEU’s influence on urban GI. The paper not only deepens the understanding of the NEU dividend release mechanism but also provides operational policy implications for coordinating the two key agendas of NEU construction and GI development.

Optimizing bike-sharing station locations: A machine learning and artificial neural networks approach using geospatial and demographic data

In the modern world, public transportation amenities are noticeably on the rise, with urban bike-sharing systems becoming well-established in many major cities. However, not all cities have these systems, and planning optimal locations for bike-sharing stations is a complex task that requires consideration of many factors. To address this, the authors of this research paper developed a model to predict suitable locations for bike-sharing stations, utilizing machine learning techniques and artificial neural networks. These techniques utilized land cover and demographic data to train the model, achieving a high accuracy of 0.977. The predicted bike-sharing stations not only align with existing networks but also support their expansion, as many suggested locations are near major intersections and public transportation stops, confirming their suitability for the urban bike network. Additionally, the model was applied to Rzeszów, a city without a current bike-sharing system, where it successfully identified optimal locations for new stations. This demonstrates the methodology’s practical applicability and its valuable support for planning bike-sharing infrastructure in urban areas.