Browse Articles

Discover research articles across all indexed journals

Generative AI without guardrails can harm learning: Evidence from high school mathematics

Proceedings of the National Academy of Sciences Hamsa Bastani, Osbert Bastani, Alp Sungu et al. Jul 01, 2025 DOI: 10.1073/pnas.2422633122

Generative AI is poised to revolutionize how humans work, and has already demonstrated promise in significantly improving human productivity. A key question is how generative AI affects learning—namely, how humans acquire new skills as they perform tasks. Learning is critical to long-term productivity, especially since generative AI is fallible and users must check its outputs. We study this question via a field experiment where we provide nearly a thousand high school math students with access to generative AI tutors. To understand the differential impact of tool design on learning, we deploy two generative AI tutors: one that mimics a standard ChatGPT interface (“GPT Base”) and one with prompts designed to safeguard learning (“GPT Tutor”). Consistent with prior work, our results show that having GPT-4 access while solving problems significantly improves performance (48% improvement in grades for GPT Base and 127% for GPT Tutor). However, we additionally find that when access is subsequently taken away, students actually perform worse than those who never had access (17% reduction in grades for GPT Base)—i.e., unfettered access to GPT-4 can harm educational outcomes. These negative learning effects are largely mitigated by the safeguards in GPT Tutor. Without guardrails, students attempt to use GPT-4 as a “crutch” during practice problem sessions, and subsequently perform worse on their own. Thus, decision-makers must be cautious about design choices underlying generative AI deployments to preserve skill learning and long-term productivity.

Microglial cyclooxygenase-1 modulates cerebral capillary basal tone in vivo in mice

Nature Communications William A. Mills, Niesha A. Savory, Aida Oryza Lopez-Ortiz et al. Jul 01, 2025 DOI: 10.1038/s41467-025-60753-x

Abstract Microglia and border associated macrophages have been implicated in hypercapnia, but it is unknown which myeloid cell modulates which vessel type. Previously, we documented in mice myeloid cell association with the brain vasculature but did not distinguish their localization along the vascular tree. Using molecular approaches to distinguish microglia and perivascular macrophages, we show that microglia are the only myeloid cells associating with capillaries. To determine if loss of microglia is sufficient to reduce capillary tone, we employ global and focal ablations and find significant reductions in capillary diameter and red blood cell flux, suggesting vasodilatory regulation by microglia. Cyclooxygenase-1 (COX1), an enzyme with known vasodilatory action, is predominantly expressed by microglia. To determine the necessity of microglial COX1 in regulating cerebral basal capillary tone in vivo, we perform genetic ablation and find a significant reduction in capillary flux and diameter. Together, this study using male mouse models reveals a role for microglial COX1 in maintaining basal capillary tone in vivo.

Highly ordered clustering of TNFα and BAFF ligand-receptor-intracellular adaptor complexes on a lipid membrane

Nature Communications Chan Seok Lim, Jisun Lee, Ji Won Kim et al. Jul 01, 2025 DOI: 10.1038/s41467-025-61271-6

Structure of a transcribing Pol II-DSIF-SPT6-U1 snRNP complex

Nature Communications Luojia Zhang, Christopher Batters, Shintaro Aibara et al. Jul 01, 2025 DOI: 10.1038/s41467-025-60979-9

Abstract In eukaryotic cells, splicing occurs predominantly co-transcriptionally, enhancing splicing efficiency and fidelity while introducing an additional layer of regulation over gene expression. RNA polymerase II (Pol II) facilitates co-transcriptional splicing by recruiting the U1 small nuclear ribonucleoprotein particle (U1 snRNP) to the nascent transcripts. Here, we report the cryo-electron microscopy structure of a transcribing Pol II-U1 snRNP complex with elongation factors DSIF and SPT6. In addition, our biochemical analysis reveals that the phosphorylated Pol II carboxyl-terminal domain and SPT6 interact directly with U1 snRNP proteins, facilitating its recruitment to the elongation complex. This multivalent interaction between U1 snRNP and the transcription elongation complex may both allow efficient spliceosome assembly and ensure transcription processivity.

Nonbonding Electron Inversion‐Driven Structural Engineering: Synergistic Enhancement of Linear and Nonlinear Optical Properties

Angewandte Chemie International Edition Jia‐Xiang Zhang, Sheng‐Hua Zhou, Xin‐Tao Wu et al. Jul 01, 2025 DOI: 10.1002/anie.202506658

Abstract The simultaneous optimization of large birefringence (Δn, a linear optical property) and strong second‐harmonic generation (SHG, a nonlinear optical (NLO) property) in a single crystal remains a significant challenge due to the inherently distinct structural requirements for these properties. Although nonbonding electrons have been extensively studied in oxides and chalcogenides, research has predominantly focused on their role along polar axes, leaving their influence along axes in tetrahedral stacking largely unexplored. Herein, we propose a nonbonding electron‐inversion strategy to overcome phase‐matching limitations in defect diamond‐like structures. By incorporating T2‐[Ga 4 S 10 ] supertetrahedral motifs, we successfully synthesized [Ba 4 Cl 2 ][CdGa 4 S 10 ] (space group: I ), which exhibits a 219% enhancement in Δn compared to the nonphase‐matching parent structure Cd 2 GaS 4 . The weakly bound nonbonding electrons, governed by atomic potentials, demonstrate strong SHG responses under optical fields. The compound [Ba 4 Cl 2 ][CdGa 4 S 10 ] not only achieves a broad transmission range (0.28–18.6 µm) and a high laser‐induced damage threshold (40.1 × AgGaS 2 ) but also optimally balances a wide bandgap ( E g  = 3.58 eV) and a large SHG response (1.4 × AgGaS 2 ), representing one of the best‐performing Cd‐based materials to date. This work introduces the first phase‐matching design strategy based on nonbonding electron‐driven structure–property relationships, providing critical insights for the rational design of high‐performance NLO materials.

Fuzzy cooperative game analysis within the international railway freight coalition framework

Scientific Reports Jiao Wu, Shansong Wu Jul 01, 2025 DOI: 10.1038/s41598-025-05977-z

Botnet detection in internet of things using stacked ensemble learning model

Scientific Reports Mudasir Ali, Muhammad Faheem Mushtaq, Urooj Akram et al. Jul 01, 2025 DOI: 10.1038/s41598-025-02008-9

Population dynamics of fruit flies (Diptera: Tephritidae) in a semirural area under subtropical monsoon climate of Bangladesh

Scientific Reports Mahfuza Momen, Md. Aftab Hossain, Kajla Seheli et al. Jul 01, 2025 DOI: 10.1038/s41598-025-03749-3

The effect of changing heat use patterns on residential energy efficiency in a Japanese smart community

Scientific Reports Le Na Tran, Qian Wu, Huong Thanh Hoang Jul 01, 2025 DOI: 10.1038/s41598-025-04760-4

Knowledge, attitude, and practice of orthodontic treatment among student patients preparing for or undergoing treatment

Scientific Reports Yaping Zhang, Liangmei Gu, Bingxue Du et al. Jul 01, 2025 DOI: 10.1038/s41598-025-97801-x

Functional immune reconstitution after allogeneic hematopoietic stem cell transplantation in myeloablative and non-myeloablative conditioned patients

Scientific Reports Patrick Terrence Brooks, Lia Minculescu, Rebecca Svanberg Teglgaard et al. Jul 01, 2025 DOI: 10.1038/s41598-025-06718-y

Associations between oxidative balance score and chronic kidney disease events in type 2 diabetes mellitus patients: a cross‑sectional study

Scientific Reports Yunhe Ding, Zhen Feng, Xuehua Liu et al. Jul 01, 2025 DOI: 10.1038/s41598-025-05281-w

Lessons learned from RadiologyNET foundation models for transfer learning in medical radiology

Scientific Reports Mateja Napravnik, Franko Hržić, Martin Urschler et al. Jul 01, 2025 DOI: 10.1038/s41598-025-05009-w

Mechanistic insights into gasdermin-mediated pyroptosis

Nature Reviews Molecular Cell Biology Yang Bai, Youdong Pan, Xing Liu Jul 01, 2025 DOI: 10.1038/s41580-025-00837-0

Fine structural design of 3βHSD1 inhibitors for prostate cancer therapy

Proceedings of the National Academy of Sciences Dongyin He, Luyao Zhang, Leiye Yu et al. Jul 01, 2025 DOI: 10.1073/pnas.2422267122

Prostate cancer is a global health challenge, particularly for patients resistant to the second-generation anti-androgen receptor pathway inhibitors. The steroidogenic enzyme 3β-hydroxysteroid dehydrogenase type 1 (3βHSD1) has emerged as a promising therapeutic target and the corresponding inhibitors, biochanin-A (BCA) and its derivatives, suppress tumor growth in preclinical models and patients. However, the poor oral bioavailability of BCA hinders its clinical application. Here, we employed a sophisticated computational approach to refine the structural design of 3βHSD1 inhibitors. AlphaFold2 was utilized to construct detailed models of 3βHSD1 binding to various substrates. These models, in conjunction with the elucidated enzymatic mechanism of 3βHSD1, guided the optimization of a series of BCA-related compounds. Our structure–activity relationship studies identified HEAL-116 as a potent 3βHSD1 inhibitor. HEAL-116 exhibited enhanced binding specificity to the substrate-binding pocket of 3βHSD1 and effectively neutralized the local charge environment. The incorporation of hydrophilic groups in its structure also markedly enhanced its oral bioavailability. HEAL-116 robustly inhibited 3βHSD1 activity and exerted pronounced antitumor effect in biochemical, cellular, and mouse models. Our findings lay the foundation for the clinical translation of 3βHSD1 inhibitors, offering a promising therapeutic strategy for the management of prostate cancer and potentially other diseases.

Evolution of linear triterpenoid biosynthesis within the Euphorbia genus

Nature Communications Tomasz Czechowski, Yi Li, Alison D. Gilday et al. Jul 01, 2025 DOI: 10.1038/s41467-025-60708-2

Abstract Terpenoids are among the largest classes of plant natural products. Squalene, a high value commodity in the cosmetic, food and pharmaceutical industries, is a common linear precursor for the biosynthesis of C30 triterpenes and sterols across plant, animal and fungal kingdoms. The anti-fungal compound peplusol is another linear C30 triterpene, but has only been reported in the genus Euphorbia. Here, we show that peplusol production has evolved due to duplication of a sterol synthase gene with one copy acquiring peplusol synthase activity and the original gene retaining the ancestral function. We identify a number of key amino acid residues that can convert the squalene synthase enzyme into peplusol synthase and vice versa. The PEPLUSOL SYNTHASE gene from E. peplus is able to drive significant levels of peplusol production in alternate host production platforms including Nicotiana benthamiana (over 2.5% leaf dry weight) and Saccharomyces cerevisiae (30 mg/L culture).

Genetic risk predicts adolescent mood pathology via sexual differentiation of brain function and physiological aging

Nature Communications Raluca Petrican, Alex Fornito, Christopher Murgatroyd et al. Jul 01, 2025 DOI: 10.1038/s41467-025-60686-5

Abstract Recent evidence challenged the traditional, categorical approach to sex differences, indicating that each human brain comprises a mosaic of features, some of which are more common among males, others, among females, whereas the remaining are equally common between sexes. Thus, a focus on regional sexual differentiation of brain function, instead of holistic sex-based categorization, could be more useful for understanding psychiatric conditions, such as mood and behavioural disorders, to which males and females are differentially vulnerable. To probe this untested hypothesis, we estimate sexual differentiation within each brain in a longitudinal (N = 199) and cross-sectional (N = 277) sample of male and female adolescents. Greater feminization of association networks, involved in higher-order cognition, compared to sensory networks, at ages 9-10 correlates with earlier puberty and greater immune/metabolic dysregulation at ages 11-12, particularly among girls. Greater masculinization of association networks relates to later puberty and reduced immune/metabolic dysregulation, especially among boys. The brain and physiological profiles sequentially mediate the relationship between genetic risk and rising mood/behavioural symptoms. These links are replicated in the cross-sectional sample and shown to hold across sexes. Our study emphasizes the importance of integrating assessments of regional sexual differentiation and physiology in personalizing psychiatric intervention in adolescence.

Cardiomyocyte-derived USP13 protects hearts from hypertrophy via deubiquitinating and stabilizing STAT1 in male mice

Nature Communications Jibo Han, Liming Lin, Zimin Fang et al. Jul 01, 2025 DOI: 10.1038/s41467-025-61028-1

High GOLT1A expression induces poor prognosis in papillary thyroid carcinoma

Scientific Reports Tianjiao Ding, Yiyu Zhu, Hanjia Zhou et al. Jul 01, 2025 DOI: 10.1038/s41598-025-05747-x

Effect of ultrasonic atomization parameters on AlSi12 aluminum powder characteristics for additive manufacturing

Scientific Reports Andrei Yankin, Hussain Ali Murtaza, Boris Golman et al. Jul 01, 2025 DOI: 10.1038/s41598-025-06086-7