p53 mutations are etiologically associated with the development of oral squamous cell carcinomas (OSCCs) or are associated with exposure to specific carcinogens. In this study, we used PCR-single strand conformation polymorphism and DNA sequencing to analyze the conserved regions of the p53 gene (exons 5-9) in OSCC tumor specimens from 187 patients with varied histories of betel quid, tobacco and alcohol use. Ninety-one of the 187 OSCCs (48.66%) showed p53 gene mutations at exons 5-9. The incidence of p53 mutations was not associated with age, sex, TNM stage, status of cigarette smoking or betel quid chewing. However, alcohol drinkers exhibited a significantly higher incidence (57/101, 56.44%) of p53 mutations than non-users (39.53%, 34/86) (P = 0.02). The effect of alcohol on the incidence of p53 mutations was still statistically significant (RR = 2.24; 95% CI, 1.21-4.15) after adjustment for cigarette smoking and betel quid (BQ) chewing. G:C to A:T transitions were the predominant mutations observed and associated with BQ and tobacco use. Alcohol drinking could enhance these transitions. After adjustment for cigarette smoking and BQ chewing, alcohol drinking still showed an independent effect on G:C to A:T transitions (RR = 2.41; 95% CI, 1.01-5.74). These findings strongly suggest an important contributive role of tobacco carcinogens to p53 mutation in this series of Taiwanese OSCCs and alcohol might enhance these mutagenic effects. As safrole-DNA adducts have been detected in 77% (23/30) of the OSCC tissues from Taiwanese oral cancer patients with a BQ chewing history, we cannot rule out the possibility that safrole or other carcinogens present in the BQ may cause a similar pattern of mutagenesis. Determination of the role of safrole and other carcinogens present in BQ on the pattern of p53 gene mutation in OSCC will require further study.

Knowledge graph embedding aims at modeling entities and relations with low-dimensional vectors. Most previous methods require that all entities should be seen during training, which is unpractical for real-world knowledge graphs with new entities emerging on a daily basis. Recent efforts on this issue suggest training a neighborhood aggregator in conjunction with the conventional entity and relation embeddings, which may help embed new entities inductively via their existing neighbors. However, their neighborhood aggregators neglect the unordered and unequal natures of an entity's neighbors. To this end, we summarize the desired properties that may lead to effective neighborhood aggregators. We also introduce a novel aggregator, namely, Logic Attention Network (LAN), which addresses the properties by aggregating neighbors with both rules-and network-based attention weights. By comparing with conventional aggregators on two knowledge graph completion tasks, we experimentally validate LAN's superiority in terms of the desired properties.

Hepatocellular carcinoma (HCC) is a common and highly malignant tumor that is prone to recurrence and metastasis and has no effective treatment. Unsurprisingly, its prognosis is quite poor; early detection methods and effective low-toxicity treatments are urgently needed. To achieve these goals, we designed a multifunctional, U.S. Food and Drug Administration-approved Prussian blue (PB) nanoparticle (NP) with a porous metal organic frame loaded with sorafenib (SF), conjugated with HCC-specific targeting peptide SP94 and the near-infrared dye cyanine (Cy)5.5. These NPs are amenable to multimodal imaging for dynamic monitoring of their biodistribution and tumor-targeting effects. The SP94-PB-SF-Cy5.5 NPs achieved targeted delivery and controlled SF release and exhibited good photothermal effects. In this strategy, photothermal therapy and SF treatment complement each other, reducing the side effects of SF and achieving a therapeutic effect without local tumor recurrence. In addition, the catalase-like ability of the NPs alleviates tumor hypoxia, and their photothermal effects induce immunogenic cell death, leading to the release of tumor-associated antigens. These effects combine to trigger an antitumor immune response; the NPs also displayed promising inhibitory effects on tumor metastasis and recurrence and produced an abscopal effect and long-term immunological memory when combined with antiprogrammed death-ligand 1 (PD-L1) immunotherapy. These safe, multifunctional NPs represent a valuable treatment option for HCC. In addition, this next-generation treatment model may provide some ideas for the management of HCC and other cancers.

Existing work that augment question answering (QA) models with external knowledge (e.g., knowledge graphs) either struggle to model multi-hop relations efficiently, or lack transparency into the model's prediction rationale. In this paper, we propose a novel knowledge-aware approach that equips pretrained language models (PTLMs) with a multi-hop relational reasoning module, named multi-hop graph relation network (MHGRN). It performs multi-hop, multi-relational reasoning over subgraphs extracted from external knowledge graphs. The proposed reasoning module unifies path-based reasoning methods and graph neural networks and results in better interpretability and scalability. We also empirically show its effectiveness and scalability on CommonsenseQA and OpenbookQA datasets, and interpret its behaviors with case studies, with the code for experiments released 1 .

No abstract

Commonsense question answering (QA) requires background knowledge which is not explicitly stated in a given context. Prior works use commonsense knowledge graphs (KGs) to obtain this knowledge for reasoning. However, relying entirely on these KGs may not suffice, considering their limited coverage and the contextual dependence of their knowledge. In this paper, we augment a general commonsense QA framework with a knowledgeable path generator. By extrapolating over existing paths in a KG with a state-of-the-art language model, our generator learns to connect a pair of entities in text with a dynamic, and potentially novel, multi-hop relational path. Such paths can provide structured evidence for solving commonsense questions without finetuning the path generator. Experiments on two datasets show the superiority of our method over previous works which fully rely on knowledge from KGs (with up to 6% improvement in accuracy), across various amounts of training data. Further evaluation suggests that the generated paths are typically interpretable, novel, and relevant to the task.

No abstract

No abstract