Find the papers that actually matter

Question: Does atomic fact-checking, which decomposes AI treatment recommendations into individually verifiable claims linked to source guideline documents, increase clinician trust compared to traditional explainability approaches? Findings: In this randomized trial of 356 clinicians generating 7,476 trust ratings, atomic fact-checking produced a large effect on trust (Cohen's d = 0.94), increasing the proportion of clinicians expressing trust from 26.9% to 66.5%. Traditional transparency mechanisms showed a dose-response gradient of improvement over baseline (d = 0.25 to 0.50). Meaning: Decomposing AI recommendations into individually verifiable claims linked to source guidelines produces substantially higher clinician trust than traditional explainability approaches in high-stakes clinical decisions.

The scope of AI safety and alignment work in generative artificial intelligence (GenAI) has so far mostly been limited to harms related to: (a) discrimination and hate speech, (b) harmful/inappropriate (violent, sexual, illegal) content, (c) information hazards, and (d) use cases related to malicious actors, such as cybersecurity, child abuse, and chemical, biological, radiological, and nuclear threats. The public conversation around AI, on the other hand, has also been focusing on threats to our cognition, mental health, and welfare at large, related to over-relying on new technologies, most recently, those related to GenAI. Examples include deskilling associated with cognitive offloading and the atrophy of critical thinking as a result of over-reliance on GenAI systems, and addiction associated with attachment and dependence on GenAI systems. Such risks are rarely addressed, if at all, in the AI safety and alignment literature. In this paper, we highlight and quantify this discrepancy and discuss some initial thoughts on how safety and alignment work could address cognitive and mental health concerns. Finally, we discuss how information campaigns and regulation can be used to mitigate such prominent risks.

The clinical integration of deep learning models for brain tumor diagnosis in neuro-oncology is severely constrained by limited expert-annotated MRI data and substantial inter-institutional domain shift arising from variations in scanners, imaging protocols, and contrast settings. These challenges significantly impair model generalization in real-world settings. To address this, we propose a novel orientation-aware unsupervised domain-adaptive framework for automated brain tumor classification using mixed 2D MRI slices. Initially, a CNN with large receptive field first categorizes input slices into axial, sagittal, and coronal views. For each orientation, a CNN architecture with ResNet50 backbone augmented with four fully connected layers is trained to extract discriminative features for tumor classification. To mitigate annotation scarcity and domain discrepancies, we introduce a slice-wise unsupervised domain adaptation strategy that transfers knowledge from the multi-modal such as T1, T2, and FLAIR source domain to the post-contrast T1 target domain. Feature-level alignment is enforced using maximum mean discrepancy loss, complemented by pseudo-label guided adaptation to preserve class discriminability. Extensive experiments demonstrate improved target-domain performance over prior approaches, highlighting the benefits of orientation-specific learning, multi-modal knowledge transfer, pseudo-label-guided adaptation, and unsupervised domain adaptation.

Radiotherapy can lead to late-onset toxicity, to varying extents between individuals due to differences in radiosensitivity. Predicting which patients are most at risk is key to augmenting the therapeutic window. However, the underlying biological mechanisms remain poorly understood, and current experimental methods often lack clinical applicability. This study employs Raman spectroscopy to analyse biochemical profiles in peripheral lymphocytes and plasma, aiming to monitor radiotherapeutic response and predict intrinsic radiosensitivity in high-risk localised prostate cancer patients treated with stereotactic radiotherapy. Partial-least squares discriminant analysis classification of Raman spectra at baseline (n = 20) from post-hormone therapy (n = 19), mid-treatment (pre-4th fraction; n = 21) and 3-months after treatment (n = 18) returned mean area under the curve values ranging from 0.88 to 0.93. Ensemble classifiers applied to imbalanced late toxicity datasets (grade 0-1, n = 16; grade 2+, n = 4) yielded mean F1 scores of 0.74 (random forest, lymphocytes) and 0.69 (AdaBoost, plasma); metrics based on best performing model for minority-class. Classical least squares lymphocyte and plasma toxicity models identified major concentration differences in amino acids, proteins, lipids, DNA and related biomolecules (p < 0.05). These findings demonstrate the potential of Raman spectroscopy as a minimally invasive, objective tool for classifying blood-based biochemical profiles across radiotherapy treatment time points and distinguishing patients with late grade 0-1 and grade 2+ toxicity.

Raman spectroscopy enabled accurate discrimination of posterior fossa ependymomas and medulloblastomas in both frozen and formalin-fixed, paraffin-embedded (FFPE) specimens in this retrospective study. We acquired Raman spectra (532 nm excitation) from frozen and FFPE tissues to evaluate a principal component analysis-support vector machine classifier by using fivefold cross-validation. We also performed a pilot spatial transcriptomics analysis on three FFPE sections by using the 10× Genomics Xenium In Situ v2 FFPE workflow. In total, 34 specimens (21 frozen and 13 FFPE) were analyzed, and the classification models achieved >90% accuracy in distinguishing ependymomas from medulloblastomas under spectrum-level fivefold cross-validation, suggesting discriminative biochemical differences, whereas patient-level performance requires further validation in larger cohorts. Ependymomas had higher lipid-associated Raman bands (1084, 1128, and 1654 cm-1), whereas medulloblastomas exhibited higher deoxyhemoglobin-related Raman bands (1356, 1548, and 1604 cm-1). Compared with normal controls, tumor tissues had increased carotenoid-related Raman bands and reduced lipid-associated Raman bands. The observed spectral differences are consistent with differences in lipid-associated composition and the heme/oxygenation-related tissue context, and pilot transcriptomic profiling provided qualitative biological context related to lipid metabolism and angiogenesis. These findings support Raman spectroscopy as a label-free spectroscopic technique that may complement conventional diagnostics and aid surgical and therapeutic decision-making. Larger, prospective studies are warranted to further evaluate the clinical generalizability and intraoperative translation of our results.

Traf2-and Nck-interacting kinase (TNIK), a downstream effector of the Wnt/β-catenin pathway and a key regulatory component of the TCF4/β-catenin transcriptional complex, has emerged as a potential therapeutic target for colorectal cancer. In this study, based on compound 1, a previously reported TNIK inhibitor, we developed a series of optimized inhibitors featuring a 1H-pyrrolo[2,3-b]pyridine scaffold. Among these, compound N15 exhibited the most potent activity, with exceptional TNIK inhibition in an in vitro enzymatic assay (IC50 = 0.49 nM) and favorable metabolic stability in human liver microsomes (T1/2 = 241 min). In HCT116 cells, N15 exhibited strong antitumor activity by suppressing proliferation, inducing apoptosis, and causing cell cycle arrest, while also downregulating Wnt pathway target gene expression. Furthermore, N15 significantly inhibited tumor growth in an HCT116 xenograft mouse model without inducing notable adverse effects. Collectively, these results identify N15 as a promising lead compound for further development of TNIK inhibitors.

Prostate cancer (PCa) is a frequently observed male cancer characterized by high morbidity and mortality. STAT3 is closely related to the occurrence and development of cancer, suggesting that it may be an antitumor therapeutic target. In this study, we prepared various napabucasins bearing sulfonylpiperazine scaffolds as STAT3 inhibitors to treat PCa. Among these compounds, YN11 was the most potent, with an IC50 value of 23 nM in DU145 cells, which is 8.8 times greater than the IC50 value of napabucasin. Mechanistic studies revealed that YN11 directly binds to the STAT3 SH2 domain, inhibiting the phosphorylation of STAT3 while reducing the expression of downstream target proteins. Moreover, YN11 triggered cell cycle arrest, promoted apoptosis, and effectively suppressed PCa cell invasion and migration. In vivo studies revealed that YN11 significantly inhibited tumor growth without inducing considerable weight loss or apparent histopathological alterations in major organs. Our findings indicate that YN11 is a potent STAT3 inhibitor for treating PCa.

Selective inhibition of hexokinase 2 (HK2) represents a promising therapeutic strategy due to the pivotal role of HK2 in the Warburg effect, enhancement of glycolysis and anti-apoptosis via HK2-Voltage-Dependent Anion Channel 1 (VDAC1) protein-protein interaction. Moreover, HK2 initiates glycolysis to generate lactate, hence this central enzyme can be pharmacologically targeted to enhance therapy outcomes. Currently, no HK2 inhibitors (HK2is) exist in the clinic. Herein, we employed GCVec, an artificial intelligence (AI)-based compound-protein interaction (CPI) prediction tool, along with molecular docking, to identify the HK2i, 106. This compound exhibited an IC50 of 0.79 ± 0.07 μM and a consistent Kd of 0.41 ± 0.03 μM against HK2 enzyme. It also apparently blocked HK2-VDAC1 interaction as indicated by the disrupted colocalization of HK2-GFP and VDAC1-mCherry. Furthermore, 106 demonstrated enhanced anticancer efficacy under hypoxia in tumor cells with elevated HIF-1α/HK2 and VDAC1 levels. Compound 106 selectively targeted SW480 colorectal cancer cells with high HK2 expression, achieving a growth inhibition IC50 value of 5.00 ± 0.94 μM. Consistently, knockout of HK2 in these tumor cells significantly rescued the IC50 values and eliminated the glycolytic inhibition induced by 106. We further showed that 106 reduced lactate and ATP levels and induced markers of apoptosis, including increased p-AMPK/AMPK ratio and increased Bax levels, as well as decreased Bcl2 levels. Collectively, our findings highlight the potential of GCVec in identifying 106, a first in class dual-function HK2i which emerges as a promising lead compound for further development into a possible anticancer therapeutic agent.

Mitochondria are essential for cancer cell survival, with the thioredoxin/thioredoxin reductase 2 (Trx/TrxR2) system acting as a key redox regulator. Steviol, an abundant natural ent-kaurane diterpenoid, exhibits negligible cytotoxicity, while most active ent-kaurane analogs depend on a reactive exo-methylene cyclopentanone moiety, raising selectivity and safety concerns. To address these limitations, 28 triphenylphosphonium (TPP)-conjugated steviol derivatives were synthesized to enhance mitochondrial accumulation and modulate mitochondrial signaling. SAR analysis revealed that dual functionalization at C-13 (TPP) and C-19 (esterification) markedly improved potency and selectivity. Conjugate 23d (C-13 TPP, C-19 benzyl ester) was the most potent (IC50 = 0.19 μM, SI = 15.42) and significantly suppressed Huh7 xenografts growth with favorable safety. Mechanistic studies demonstrated mitochondrial accumulation, TrxR2 inhibition, ROS elevation, and ASK1-mediated apoptosis. To our knowledge, 23d is the first non-electrophilic ent-kaurane derivative to combine mitochondrial targeting with TrxR2 inhibition and in vivo antitumor efficacy, highlighting dual modification as a promising strategy integrating biodistribution engineering with activity optimization for anticancer drug development.

The hinge region in the ATP binding site of kinase has become the promising target to design potent inhibitors for cancer therapy. Among the ongoing development of PIM inhibitors based on N-pyridinyl amide scaffold for acute myeloid leukemia (AML), the structural-activity relationship (SAR) associated with the fragment towards hinge region still remains an open question. Herein, we systematically optimized hinge region-binding heterocycle of PIM kinase inhibitors based on N-pyridinyl amide scaffold. SAR studies revealed that a 2-position nitrogen configuration capable of forming intramolecular hydrogen bond is optimal to stabilize bioactive conformation. The introduction of 6-position amino group on the heterocycle engaged with upper hinge region through hydrogen bond formation with Glu121, achieving sub-nanomolar PIM kinase inhibition. And it was found that the electronegativity of substituents on the ring exerts minimal modulation effects on this key hydrogen bond with Glu121. Whereas 6-aminopyrazine scaffold could strengthen this hydrogen bond interaction by the electron-withdrawing nature of the additional nitrogen atom adjacent to the amino group. These findings finally screened out compound FD2024 (compound 27), which demonstrated potent pan-PIM inhibition and anti-AML efficacy both in vitro and in vivo. This work highlighted the pivotal role of hinge region-binding fragment, specifically the 2-position nitrogen for bioactive conformation and 6-amino group for engaging Glu121 in improving PIM kinase inhibitor potency.