In the area of large-scale graph data representation and semi-supervised learning, deep graph-based convolutional neural networks have been widely applied. However, typical graph convolutional network (GCN) aggregates information of neighbor nodes based on binary neighborhood similarity (adjacency matrix). It treats all neighbor nodes of one node equally, which does not suppress the influence of dissimilar neighbor nodes. In this paper, we investigate GCN based on similarity matrix instead of adjacency matrix of graph nodes. Gaussian heat kernel similarity in Euclidean space is first adopted, which is named EGCN. Then biologically inspired manifold similarity is trained in reproducing kernel Hilbert space (RKHS), based on which a manifold GCN (named MGCN) is proposed for graph data representation and semi-supervised learning with four different kernel types. The proposed method is evaluated with extensive experiments on four benchmark document citation network datasets. The objective function of manifold similarity learning converges very quickly on different datasets using various kernel functions. Compared with state-of-the-art methods, our method is very competitive in terms of graph node recognition accuracy. In particular, the recognition rates of MGCN (Gaussian kernel) and MGCN (Polynomial Kernel) outperform that of typical GCN about 3.8% on Cora dataset, 3.5% on Citeseer dataset, 1.3% on Pubmed dataset and 4% on Cora_ML dataset, respectively. Although the proposed MGCN is relatively simple and easy to implement, it can discover local manifold structure by manifold similarity learning and suppress the influence of dissimilar neighbor nodes, which shows the effectiveness of the proposed MGCN.
目的:基于网络药理学方法探讨生脉散治疗心房纤颤的作用靶点和相关信号通路并探讨其作用机制。方法:运用中药系统药理学成分分析平台(bioinformatics analysis tool for molecular mechanism of TCM,BATMAN-TCM)数据库获取生脉散的化学成分及作用靶标基因,通过GeneCards,OMIM,DisGeNET数据库收集心房纤颤的靶标基因。将两者取交集后得到生脉散-心房纤颤靶基因交集,运用STRING构建蛋白质间相互作用网络,并将结果进行网络可视化展示。将药物-疾病交集基因导入DAVID6.8数据库,进行基因本体(gene ontology,GO)分析和基于京都基因与基因组百科全书(Kyoto Encyclopedia of Genes and Geomes,KEGG)通路富集分析。结果:生脉散干预房颤的有效活性成分159个,药物靶点与疾病靶点交集后获得206个共有靶点,PPI蛋白互作网络分析发现AKT1,TP53,PRKACA,IL-1B,TNF,INS,PPAR,RXR,F2,CACAN1C,PKC等是生脉散治疗房颤的核心靶点。GO富集分析确定了175个条目(P0.05),其中生物过程主要心脏传导调节心率、动作电位时膜去极化等;分子功能主要包括电压门控钙通道、类固醇激素受体活性、肾上腺素结合等,在细胞组成方面,主要包括钠、钾、钙通道复合物等。KEGG通路富集分析确定了100条相关信号通路,主要有cGMP/PKG信号通路,cAMP信号通路,血清素能突触,肾素分泌,钙信号通路等。结论:生脉散治疗心房纤颤具有多途径、多靶点作用的特点。该研究初步探讨了其作用的关键靶点及涉及的生物学过程和信号通路,为生脉散治疗心房纤颤后续的实验研究提供一定的参考。 相似文献
OBJECTIVES: To assess the effectiveness and safety of additional bedtime H2‐receptor antagonists (H2RAs) in suppressing nocturnal gastric acid breakthrough (NAB) via a systematic review. METHODS: Eligible trials were identified by searching the Cochrane Central Register of Controlled Trials (CENTRAL) (Cochrane Library Issue 2, 2004), MEDLINE (January 1966–June 2004), EMBASE (January 1980–June 2004) and CINAHL (January 1982–June 2004). Additional hand‐searching was conducted on the proceedings of correlated conferences, eight important Chinese journals and references of all included trials. All randomized controlled trials evaluating H2RAs for the control of NAB were eligible for inclusion. The systematic review was conducted using methods recommended by The Cochrane Collaboration. RESULTS: Only two randomized crossover studies, comprising 32 participants, met the inclusion criteria. Because the design, dosage and duration of the treatments were different between the studies, it was not possible to conduct meta‐analysis. There were no consistent conclusions found between the two included studies in evaluating H2RAs for the control of NAB. CONCLUSIONS: No implications for practice at this stage can be concluded. Appropriately designed large‐scale randomized controlled trials with long‐term follow up are needed to determine the effects of additional bedtime H2RAs in suppressing NAB. 相似文献
Objective To investigate the relationship between degree of endometrioma adhesions and clinical feature, surgical treatment and postoperative recurrence. Methods From Jan 2003 to Mar 2008, 662 patients with endometrioma undergoing laparoscopic ovarian endometrioma excision in Peking Union Medical College Hospital were studied retrospectively. All patients were classified into four groups according to the extent of adhesions: 31 cases in none adhesions group, 123 cases in mild adhesions group (filmy thickness, avascular, easily separated adhesions), 310 cases in moderate adhesions group (less than a half of ovary was adjacent to dense thickness adhesions which was difficult to separate, or above a half of ovary were adjacent to filmy thickness adhesions) and 198 cases in severe adhesions group (above a half of ovary was adjacent to dense thickness, well vascularized adhesions which was difficult to separate, and always involved the other pelvic organs, observed angiogenesis). The comparison of degree, characteristics, period of pain, lab test, surgical management and postoperative recurrence was performed among those above groups. In the mean time, risk factors and multinomial logistic regression were analyzed. Results (1)Clinical characteristics: The incidence of patients with dysmenorrhea, dyspareunia, straining feeling in anus, chronic pelvic pain and the level of CA125 (>35 kU/L) was remarkably higher in moderate-to-severe adhesion groups than in none-to-mild adhesions groups (P=0.000, 0.000, 0.001, 0.006 and 0.000, respectively). Infertility rate were significantly higher in severe adhesions group(15.7%,31/198) than none adhesions group(3.2%,1/31), mild adhesions group(11.4%,14/123) and moderate adhesions group(9.7%,30/310, OR=1.728, P<0.05).(2)Operating time and blood loss: Operating time of each groups was as followed: (37±15) min in none adhesions group, (42±19) min in mild adhesions group, (50±20) min in moderate adhesions group and (63±22) min in severe adhesion group. Blood loss was (23±12) ml in none adhesion group, (31±27) ml in mild adhesion group, (40±32) ml in moderate adhesion group and (70±67) ml in severe adhesions group. Thicker adhesions result in longer operation time and more blood loss. (3)Combined with other disease: The ratio of patients who combined with adenomyosis or deeply infiltrating endometriosis in moderate-to-severer adhesion groups was higher than patients in none-to-mild adhesions groups (OR=3.466, P=0.000). (4) Postoperative recurrence: It was categorized into recurrence of pain and cyst. Moderate-to-severe adhesions was related to higher recurrence rate of pain (OR=1.685,P=0.046), but was irrelevant to recurrence of cyst. Conclusion The more extent of endometrioma adhesions was related to severer pelvic pain symptoms, longer operating time and more blood loss. Postoperative pain recurrence rate was observed in moderate-to-severe adhesion group. Extent of adhesions was irrelevant to cyst recurrence. 相似文献