Ranking models lie at the heart of research on information retrieval (IR). During the past decades, different techniques have been proposed for constructing ranking models, from traditional heuristic methods, probabilistic methods, to modern machine learning methods. Recently, with the advance of deep learning technology, we have witnessed a growing body of work in applying shallow or deep neural networks to the ranking problem in IR, referred to as neural ranking models in this paper. The power of neural ranking models lies in the ability to learn from the raw text inputs for the ranking problem to avoid many limitations of hand-crafted features. Neural networks have sufficient capacity to model complicated tasks, which is needed to handle the complexity of relevance estimation in ranking. Since there have been a large variety of neural ranking models proposed, we believe it is the right time to summarize the current status, learn from existing methodologies, and gain some insights for future development. In contrast to existing reviews, in this survey, we will take a deep look into the neural ranking models from different dimensions to analyze their underlying assumptions, major design principles, and learning strategies. We compare these models through benchmark tasks to obtain a comprehensive empirical understanding of the existing techniques. We will also discuss what is missing in the current literature and what are the promising and desired future directions.
In multi-turn dialogue generation, response is usually related with only a few contexts. Therefore, an ideal model should be able to detect these relevant contexts and produce a suitable response accordingly. However, the widely used hierarchical recurrent encoderdecoder models just treat all the contexts indiscriminately, which may hurt the following response generation process. Some researchers try to use the cosine similarity or the traditional attention mechanism to find the relevant contexts, but they suffer from either insufficient relevance assumption or position bias problem. In this paper, we propose a new model, named ReCoSa, to tackle this problem. Firstly, a word level LSTM encoder is conducted to obtain the initial representation of each context. Then, the self-attention mechanism is utilized to update both the context and masked response representation. Finally, the attention weights between each context and response representations are computed and used in the further decoding process. Experimental results on both Chinese customer services dataset and English Ubuntu dialogue dataset show that ReCoSa significantly outperforms baseline models, in terms of both metric-based and human evaluations. Further analysis on attention shows that the detected relevant contexts by ReCoSa are highly coherent with human's understanding, validating the correctness and interpretability of ReCoSa.
Chemotherapy outcomes for the treatment of glioma remain unsatisfied due to the inefficient drug transport across BBB/BBTB and poor drug accumulation in the tumor site. Nanocarriers functionalized with different targeting ligands are considered as one of the most promising alternatives. However, few studies were reported to compare the targeting efficiency of the ligands and develop nanoparticles to realize BBB/BBTB crossing and brain tumor targeting simultaneously. In this study, six peptide-based ligands (Angiopep-2, T7, Peptide-22, c(RGDfK), D-SP5 and Pep-1), widely used for brain delivery, were selected to decorate liposomes, respectively, so as to compare their targeting ability to BBB or BBTB. Based on the in vitro cellular uptake results on BCECs and HUVECs, Peptide-22 and c(RGDfK) were picked to construct a BBB/BBTB dual-crossing, glioma-targeting liposomal drug delivery system c(RGDfK)/Pep-22-DOX-LP. In vitro cellular uptake demonstrated that the synergetic effect of c(RGDfK) and Peptide-22 could significantly increase the internalization of liposomes on U87 cells. In vivo imaging further verified that c(RGDfK)/Pep-22-LP exhibited higher brain tumor distribution than single ligand modified liposomes. The median survival time of glioma-bearing mice treated with c(RGDfK)/Pep-22-DOX-LP (39.5 days) was significantly prolonged than those treated with free doxorubicin or other controls. In conclusion, the c(RGDfK) and Peptide-22 dual-modified liposome was constructed based on the targeting ability screening of various ligands. The system could effectively overcome BBB/BBTB barriers, target to tumor cells and inhibit the growth of glioma, which proved its potential for improving the efficacy of chemotherapeutics for glioma therapy.
Delivery of macromolecular drugs to the brain is impeded by the blood brain barrier. The recruitment of leukocytes to lesions in the brain, a typical feature of neuroinflammation response which occurs in cerebral ischemia, offers a unique opportunity to deliver drugs to inflammation sites in the brain. In the present study, cross-linked dendrigraft poly-L-lysine (DGL) nanoparticles containing cis-aconitic anhydride-modified catalase and modified with PGP, an endogenous tripeptide that acts as a ligand with high affinity to neutrophils, were developed to form the cl PGP-PEG-DGL/CAT-Aco system. Significant binding efficiency to neutrophils, efficient protection of catalase enzymatic activity from degradation and effective transport to receiver cells were revealed in the delivery system. Delivery of catalase to ischemic subregions and cerebral neurocytes in MCAO mice was significantly enhanced, which obviously reducing infarct volume in MCAO mice. Thus, the therapeutic outcome of cerebral ischemia was greatly improved. The underlying mechanism was found to be related to the inhibition of ROS-mediated apoptosis. Considering that neuroinflammation occurs in many neurological disorders, the strategy developed here is not only promising for treatment of cerebral ischemia but also an effective approach for various CNS diseases related to inflammation.
The restriction of anti-cancer drugs entry to tumor sites in the brain is a major impediment to the development of new strategies for the treatment of glioma. Based on the finding that macrophages possess an intrinsic homing property enabling them to migrate to tumor sites across the endothelial barriers in response to the excretion of cytokines/chemokines in the diseased tissues, we exploited macrophages as ‘Trojan horses’ to carry drug-loading nanoparticles (NPs), pass through barriers, and offload them into brain tumor sites. Anticancer drugs were encapsulated in nanoparticles to avoid their damage to the cells. Drug loading NPs was then incubated with RAW264.7 cells in vitro to prepare macrophage-NPs (M-NPs). The release of NPs from M-NPs was very slow in medium of DMEM and 10% FBS and significantly accelerated when LPS and IFN-γ were added to mimic tumor inflammation microenvironment. The viability of macrophages was not affected when the concentration of doxorubicin lower than 25 μg/ml. The improvement of cellular uptake and penetration into the core of glioma spheroids of M-NPs compared with NPs was verified in in vitro studies. The tumor-targeting efficiency of NPs was also significantly enhanced after loading into macrophages in nude mice bearing intracranial U87 glioma. Our results provided great potential of macrophages as an active biocarrier to deliver anticancer drugs to the tumor sites in the brain and improve therapeutic effects of glioma.
Glioma remains difficult to treat because of the infiltrative growth of tumor cells and their resistance to standard therapy. Despite rapid development of targeted drug delivery system, the current therapeutic efficacy is still challenging. Based on our previous studies, macrophages have been proved to be promising drug carrier for active glioma delivery. To make full use of macrophage carrier, primary M1 macrophages were proposed to replace regular macrophage to deliver nanodrugs into glioma, because M1 macrophages not only have the natural ability to home into tumor tissues, but they also have stronger phagocytic capability than other types of macrophage, which can enable them to uptake enough drug-loaded nanoparticles for therapy. In addition, M1 macrophages are not easily affected by harsh tumor microenvironment and inhibit tumor growth themselves. In this study, M1 macrophage-loaded nanoparticles (M1-NPs) were prepared by incubating poly(lactide-co-glycolide) (PLGA) nanoparticles with primary M1 macrophages. In vitro cell assays demonstrated M1 macrophage still maintained good tumor tropism capability after particle loading, and could efficiently carry particles across endothelial barrier into tumor tissues. In vivo imaging verified that M1-NPs exhibited higher brain tumor distribution than free nanoparticles. DOX@M1-NPs (doxorubicin-loaded M1-NPs) presented significantly enhanced anti-glioma effect with prolonged survival median and increased cell apoptosis. In conclusion, the results provided a new strategy exploiting M1 macrophage as carrier for drug delivery, which improved targeting efficiency and therapeutic efficacy of chemodrugs for glioma therapy.
The HuPrime® human gastric neuroendocrine carcinoma derived xenograft model GA0087 was established in this study. GA0087 PDX model showed high gene expression of vascular endothelial growth factors (VEGF)-A and B, and high potential of lung metastasis. Circulating tumor cells (CTCs) with either large or small size, circulating tumor microemboli (CTM) and lung metastatic lesions were detected in GA0087 PDX mice. The number of CTC correlated to the number of metastatic nodules in lung. Both primary tumor growth and metastasis in terms of the number of dynamically monitored CTCs and metastatic nodules were effectively suppressed by Cisplatin. Diverse subtypes of CTCs in the context of sensitivity to Cisplatin were specifically identified by subtraction enrichment (SE) integrated with in situ Phenotyping of cytokeratin 18 (CK18) and Karyotyping of chromosome 8 (in situ PK CTC by CK-iFISH). All the CK18-/diploid and majority of CK18+/diploid CTC subtypes were chemosensitive, whereas a higher percentage of CK18+/multiploid subtype of CTC were Cisplatin-insensitive. Combined histopathological examination of metastatic lesion and in situ PK CTC in a metastatic PDX (mPDX) tumor model are of particular significance, and may provide an unique and robust platform for cancer research as well as pre-clinical evaluation of therapeutic efficacy of new anti-cancer drugs.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
334 Leonard St
Brooklyn, NY 11211
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.