Huawei Technologies (China)

Dormant cancer cells are a significant source of cancer recurrence and metastasis and exhibit robust resistance to conventional therapies. Therefore, the exploration of novel therapeutic strategies to eliminate these cells has become a hot topic in cancer research. Ferroptosis, a newly identified form of regulated cell death, has garnered considerable attention in the field of cancer therapy in recent years. As a novel form of regulated cell death, the core mechanism of ferroptosis lies in the accumulation of intracellular iron and the induction of lipid peroxidation. Oxidative stress, the transforming growth factor-β (TGF-β) signaling pathway, autophagy, and lipid metabolism play dual roles in the survival of dormant cancer cells and the process of ferroptosis, influencing the response of dormant cancer cells to ferroptosis. These complex molecular mechanisms form a regulatory network between ferroptosis and dormant cancer cells, which holds significant implications for the development of future anti-tumor therapeutic strategies. This review synthesizes current evidence on targeting ferroptosis to eliminate dormant cancer cells, positions ferroptosis as a precision modality against dormant cancer cells, and discusses its therapeutic promise as a conceptual framework for developing next-generation anti-tumor strategies.

Metal-organic frameworks (MOFs) show great promise as carbonic anhydrase (CA) mimics due to their high surface area, tunable porosity, and abundant active sites. This study enhances a reported 2-aminoimidazole-based Zn-MOF by incorporating Co2+, Ni2+, or Cu2+ via in situ doping, systematically investigating how metal doping regulates morphology, crystal structure, and catalytic performance. The results reveal that the type and ratio of doped metal ions critically influence the framework: Co2+ demonstrated the highest compatibility, while Ni2+ and Cu2+ induced structural distortion at higher levels. All doped samples exhibited reversible catalytic kinetics akin to natural CA, with improved maximum reaction rates (Vmax) and substrate affinity. Specifically, Co75%/Zn-MOF, Ni50%/Zn-MOF, and Cu25%/Zn-MOF achieved optimal performance, showing esterase activities of 0.47 ± 0.02, 0.56 ± 0.02, and 0.39 ± 0.02 U/mg─increases of 67.9%, 100.0%, and 40.0% over pristine Zn-MOF, respectively. These materials also displayed exceptional high-temperature (80 °C) activity, hydrothermal stability, pH tolerance, and recyclability, retaining 67-79% activity after six cycles. This work provides a theoretical and experimental basis for designing efficient and stable MOF-based CA mimics, highlighting their potential for carbon capture, utilization, and storage (CCUS) applications.

Retrieval-Augmented Generation (RAG) has proven effective in integrating external knowledge into large language models (LLMs) for solving question-answer (QA) tasks. The state-of-the-art RAG approaches often use the graph data as the external data since they capture the rich semantic information and link relationships between entities. However, existing graph-based RAG approaches cannot accurately identify the relevant information from the graph and also consume large numbers of tokens in the online retrieval process. To address these issues, we introduce a novel graph-based RAG approach, called Attributed Community-based Hierarchical RAG (ArchRAG), by augmenting the question using attributed communities, and also introducing a novel LLM-based hierarchical clustering method. To retrieve the most relevant information from the graph for the question, we build a novel hierarchical index structure for the attributed communities and develop an effective online retrieval method. Experimental results demonstrate that ArchRAG outperforms existing methods in both accuracy and token cost.

While adapting pretrained vision models to downstream dense prediction tasks is widely used, current methods often overlook adaptation efficiency, especially in the context of multi-task learning (MTL). Although parameter-efficient fine-tuning (PEFT) methods can enhance parameter efficiency, broader aspects such as GPU memory and training time efficiency remain underexplored. In this paper, we propose a new paradigm that simultaneously achieves efficiency in Parameters, GPU Memory, and Training Time for Multi-Task Dense Vision Adaptation. Specifically, we propose a dual-branch framework, in which a frozen pretrained backbone serves as the generic main branch, and the proposed Bi-Directional Task Adaptation (BDTA) modules are integrated in parallel to form a task bypass branch that extracts adaptation features required by multiple specific tasks. This adaptation module is lightweight, efficient, and does not require backpropagation through the large pre-trained backbone, thus avoiding resource-intensive gradient computations. Moreover, a Mixture of Task Experts mechanism (MoTE) is further proposed to integrate adaptation features across tasks and scales, thereby obtaining more robust representations tailored for dense prediction tasks. On the PASCAL-Context benchmark, our method achieves over 2× relative performance improvement compared to the best prior multi-task PEFT method, while using only ~30% of the parameters, ~50% of the memory, and ~60% of the training time, demonstrating superior overall adaptation efficiency.

ABSTRACT Aims The tolerance of patients with compensated hepatitis B cirrhosis to interferon (IFN) therapy remains controversial. Therefore, this study aimed to evaluate the safety of pegylated interferon‐alpha in chronic hepatitis B (CHB) patients with compensatory cirrhosis. Methods Data from two prospective cohorts (the OASIS Project and CHESS 2306) from January 2018 to January 2024 were synthesized and analyzed. Patients with Child‐Pugh Class A hepatitis B cirrhosis who received IFN‐based therapy ( n = 920) were included. The control groups included patients with compensated hepatitis B receiving nucleos(t)ide analog (Nuc) monotherapy ( n = 714) and patients without cirrhosis receiving IFN‐based therapy ( n = 4111), respectively. Propensity score matching was used to control for confounding factors; 566 versus 566 cases were analyzed among patients with cirrhosis treated with IFN‐based therapy or Nuc monotherapy, and 785 versus 785 cases were analyzed among patients with and without cirrhosis treated with IFN‐based therapy. Primary outcomes included decompensation events, and secondary outcomes included severe adverse events, overall adverse events, and treatment‐related hospitalizations or deaths. Results In patients with hepatitis B virus‐related cirrhosis, the incidence of decompensation events was similar between IFN‐based therapy and Nuc monotherapy (6/566 [1.1%] vs. 3/566 [0.5%], p = 0.506). No hospitalizations or deaths were associated with adverse events during the observation period (48 weeks). The incidences of severe adverse events were similar in patients with cirrhosis under IFN‐based therapy or Nuc monotherapy (severe neutropenia: 1/450 [0.2%] vs. 0/378 [0], p = 0.999; severe thrombocytopenia: 4/435 [0.9%] vs. 0/296 [0], p = 0.153; severe alanine aminotransferase level elevation: 1/523 [0.2%] vs. 1/458 [0.2%], p = 0.999; and severe total bilirubin [TBIL] level elevation: 5/419 [1.2%] vs. 3/385 [0.8%], p = 0.727). The incidences of severe adverse events were similar between patients with and without cirrhosis receiving IFN‐based therapy, except that severe TBIL level elevation was more frequent in patients with cirrhosis who already had mildly to moderately elevated TBIL levels at baseline than in those without (6/153 [3.9%] vs. 0/118 [0], p = 0.040). Conclusion IFN‐based therapy demonstrates favorable safety in Child‐Pugh A compensated cirrhosis. It did not increase decompensation events or severe adverse events compared to Nuc monotherapy, and adverse event profiles were largely similar between cirrhotic and non‐cirrhotic patients, except for a higher risk of severe hyperbilirubinemia in those with pre‐existing TBIL elevation. ClinicalTrials.gov identifier: NCT04896255. Chinese Clinical Trial Registry number: ChiCTR2500107592.

To address the high-dimensional design optimization of an adaptive cycle fan (ACF), this paper proposes a new multi-objective optimization (MOO) method based on explainable artificial intelligence (XAI)-driven feature selection. The proposed method integrates a neural network surrogate model, Shapley additive explanation (SHAP) analysis, and a genetic algorithm. By considering Pareto front quality, surrogate model accuracy, and optimization preference, a composite evaluation metric, Q, is defined to guide a bidirectional feature selection process based on SHAP analysis, thereby establishing a dynamic, closed-loop process of simultaneous feature selection and MOO. The results indicate that the proposed method significantly enhances global search capability, accurately identifying 66 optimal features from 119 initial features. A further comparison with results without forward selection confirms the necessity of dynamically adjusting the feature space during optimization. Under the same condition, the optimal design increases the core pressure ratio from 2.71 to 2.81 and core efficiency from 80.80% to 82.92%. The flow mechanism analysis reveals that the performance gains mainly result from the reconstruction of shock structures and the suppression of shock–boundary layer interactions and secondary flows. The XAI-enhanced surrogate-assisted evolutionary algorithm (SAEA) proposed in this paper provides a promising methodology for high-dimensional MOO of aeroengines and other complex systems.