Extreme multi-label text classification(XMC) is a task for finding the most relevant labels from a large label set. Nowadays deep learning-based methods have shown significant success in XMC. However, the existing methods (e.g., AttentionXML and X-Transformer etc) still suffer from 1) combining several models to train and predict for one dataset, and 2) sampling negative labels statically during the process of training label ranking model, which will harm the performance and accuracy of model. To address the above problems, we propose LightXML, which adopts end-to-end training and dynamical negative labels sampling. In LightXML, we use GAN like networks to recall and rank labels. The label recalling part will generate negative and positive labels, and the label ranking part will distinguish positive labels from these labels. Based on these networks, negative labels are sampled dynamically during label ranking part training. With feeding both label recalling and ranking parts with the same text representation, LightXML can reach high performance. Extensive experiments show that LightXML outperforms state-of-the-art methods in five extreme multi-label datasets with much smaller model size and lower computational complexity. In particular, on the Amazon dataset with 670K labels, LightXML can reduce the model size up to 72% compared to AttentionXML. Our code is available at http://github.com/kongds/LightXML.
Extreme Multi-label text Classification (XMC) is a task of finding the most relevant labels from a large label set. Nowadays deep learning-based methods have shown significant success in XMC. However, the existing methods (e.g., Atten-tionXML and X-Transformer etc) still suffer from 1) combining several models to train and predict for one dataset, and 2) sampling negative labels statically during the process of training label ranking model, which reduces both the efficiency and accuracy of the model. To address the above problems, we proposed LightXML, which adopts endto-end training and dynamic negative labels sampling. In LightXML, we use generative cooperative networks to recall and rank labels, in which label recalling part generates negative and positive labels, and label ranking part distinguishes positive labels from these labels. Through these networks, negative labels are sampled dynamically during label ranking part training by feeding with the same text representation. Extensive experiments show that LightXML outperforms state-of-the-art methods in five extreme multi-label datasets with much smaller model size and lower computational complexity. In particular, on the Amazon dataset with 670K labels, LightXML can reduce the model size up to 72% compared to AttentionXML. Our code is available at http://github.com/kongds/LightXML.
Background: Acute lung injury (ALI) is a life-threatening disease without effective chemotherapy at present. Liang-Ge-San (LGS) is a famous traditional Chinese medicine formula, which is used to treat ALI in China. However, only a few studies have addressed the mechanisms of LGS in ALI.
Purpose: To evaluate the anti-inflammatory effects of LGS on lipopolysaccharide (LPS)-induced ALI, and to explore its underlying molecular mechanism.
Methods: Murine RAW264.7 cells were treated with LGS and LPS (1 μg/ml). The generation of IL-6, TNF-α, IL-1β was detected by ELISA. The protein expressions of STAT3 and P-STAT3 (Tyr705) were determined by Western blotting and fluorescence confocal microscopy. STAT3 transcriptional activity was investigated by luciferase reporter gene assay. qPCR was used to detect the expressions of microRNA-21 (miR-21), STAT3, and IL-6. DSS cross-linking assay was used to assess the change of STAT3 dimer. In vivo anti-inflammatory effects of LGS were evaluated in an ALI mouse model induced by tracheal instillation of LPS (3 mg/kg). The anti-ALI effects were evaluated by ELISA, qPCR, Western blotting, BCA, and H&E assays.
Results: LGS suppressed LPS-stimulated IL-6, TNF-α, and IL-1β generation in murine macrophages RAW264.7. Moreover, LGS down-regulated protein levels of P-STAT3 (Tyr705) and STAT3, inhibited STAT3 transcriptional activity, and up-regulated miR-21. Furthermore, blockage of miR-21 antagonized the inhibitory effects of LGS on the production of IL-6 and the expressions of P-STAT3 (Tyr705) and STAT3 as well as the formation of STAT3 dimer. Critically, LGS up-regulated the expression of miR-21 and inhibited the protein expressions of STAT3 and P-STAT3 (Tyr705) to reduce the release of IL-6 and inflammatory cell infiltration as well as the degree of edema in LPS-induced ALI mice.
Conclusion: LGS inhibited LPS-induced ALI through up-regulating miR-21 and subsequently inhibiting the STAT3 signaling pathway, thereby decreasing the release of IL-6.
Acute lung injury (ALI) is a life-threatening disease without effective pharmacotherapies, so far. Forsythia suspensa is frequently used in the treatment of lung infection in traditional Chinese medicine. In searching for natural anti-inflammatory components, the activity and the underlying mechanism of Forsythoside A (FA) from Forsythia suspensa were explored. In this paper, BALB/c mice and murine RAW 264.7 cells were stimulated by LPS to establish inflammation models. Data showed that FA inhibited the production of TNF-α and IL-6 and the activation of STAT3 in LPS-stimulated RAW 264.7 cells. Additionally, FA increased the expression level of miroRNA-124 (miR-124). Furthermore, the inhibitory effect of FA on STAT3 was counteracted by the treatment of miR-124 inhibitor. Critically, FA ameliorated LPS-induced ALI pathological damage, the increase of lung water content and inflammatory cytokine, cells infiltration and activation of the STAT3 signaling pathway in BALB/c mice. Meanwhile, FA up-regulated the expression of miR-124 in lungs, while administration with miR-124 inhibitor attenuated the protective effects of FA. Our results indicated that FA alleviates LPS-induced inflammation through up-regulating miR-124 in vitro and in vivo. These findings indicate the potential of FA and miR-124 in the treatment of ALI.
Cyclins are a family of proteins that regulate the cell cycle by activating cyclin-dependent kinases or a group of enzymes required in the cell cycle. Constructing a model to classify Cyclins is of importance to understand their function. It is urgent to construct a machine learning based model to identify Cyclins because of low similarity between the sequence of Cyclins. In this study, a method based on support vector machine (SVM) is developed to recognize Cyclins only using amino acid sequence information. 18 feature descriptors with a total of 13151-dimension features were extracted, and the feature dimension were reduced to 8 through feature selection technique. The reserved features show some of feature descriptors such as Autocorrelation, AAC and CTDC are important in the identification of Cyclins. Jackknife cross-validation results indicate our model would classify Cyclins with an accuracy of 91.9%, which is superior to a recent study using the same data set. Our work provides an important tool for discriminating Cyclins.
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.