Unsupervised Graph Embedding via Adaptive Graph Learning.

Graph autoencoders (GAEs) are powerful tools in representation learning for graph embedding. However, the performance of GAEs is very dependent on the quality of the graph structure, i.e., of the adjacency matrix. In other words, GAEs would perform poorly when the adjacency matrix is incomplete or be disturbed. In this paper, two novel unsupervised graph embedding methods, unsupervised graph embedding via adaptive graph learning (BAGE) and unsupervised graph embedding via variational adaptive graph learning (VBAGE) are proposed. The proposed methods expand the application range of GAEs on graph embedding, i.e, on the general datasets without graph structure. Meanwhile, the adaptive learning mechanism can initialize the adjacency matrix without being affected by the parameter. Besides that, the latent representations are embedded with the Laplacian graph structure to preserve the topology structure of the graph in the vector space. Moreover, the adjacency matrix can be self-learned for better embedding performance when the original graph structure is incomplete. With adaptive learning, the proposed method is much more robust to the graph structure. Experimental studies on several datasets validate our design and demonstrate that our methods outperform baselines by a wide margin in node clustering, node classification, link prediction, and graph visualization tasks.

Robust kernel principal component analysis with optimal mean.

The kernel principal component analysis (KPCA) serves as an efficient approach for dimensionality reduction. However, the KPCA method is sensitive to the outliers since the large square errors tend to dominate the loss of KPCA. To strengthen the robustness of KPCA method, we propose a novel robust kernel principal component analysis with optimal mean (RKPCA-OM) method. RKPCA-OM not only possesses stronger robustness for outliers than the conventional KPCA method, but also can eliminate the optimal mean automatically. What is more, the theoretical proof proves the convergence of the algorithm to guarantee that the optimal subspaces and means are obtained. Lastly, exhaustive experimental results verify the superiority of our method.

Differentiated super-enhancers in lung cancer cells.

Super-enhancers (SEs) are regulatory elements with enriched accumulation of key transcription factors. Few studies were done investigating SEs in lung cancers. Here we analyzed epigenetic profiling data to identify SEs in lung cancer cell lines. Enhancers were classified as SEs and typical enhancers (TEs). Most of the TEs were overlapped between normal cell and cancer cells. A great portion of SEs were differentiated comparing these cells. Analysis of GO terms associated with SEs revealed SE remodeling (lost on some sites while gain on others) between normal and lung cancer cells. By comparing the average number of SEs in each GO term in cancer cells with the number in control cells, surprisingly, no GO terms with significantly increased SE number in cancer condition were observed. On the contrary, in aspects such as "cell-cell adhesion", "receptor activity" and "negative regulation of canonical Wnt signaling pathway", the related SEs were significantly reduced in cancer cells. These findings suggest that in lung cancer, cells may not gain decisive gene expression in the related aspect, instead, they may have lost control of the fateful genes. Taken together, our work with the usability of omics data identified SEs in lung cancer cells and further showed cancer-specific features of SE-related terms.

Serum sRANKL and sRANKL/OPG ratio: Novel biomarkers in non-small cell lung cancer.

Osteoprotegerin (OPG) and soluble receptor activator of nuclear factor-κB ligand (sRANKL) are bone-regulating molecules. The two molecules have each been indicated to be involved in carcinogenesis. However, the diagnostic significance in non-small cell lung cancer (NSCLC) remains to be investigated. Thus, the objective of the present study was to investigate the serum levels of OPG and sRANKL in NSCLC patients, and to analyze their clinical significance. Serum OPG and sRANKL levels were determined in 50 patients with NSCLC, matched with 25 patients with benign lung nodule and 25 healthy controls by enzyme-linked immunosorbent assay. The level of serum sRANKL and the sRANKL/OPG ratio were significant elevated in the patients with NSCLC compared with the benign lung nodule patients and the healthy controls. Receiver operating characteristic curves were used to evaluate the performance of sRANKL and sRANKL/OPG. When the cut-off values for the sRANKL level and the sRANKL/OPG ratio were set at 4.20 pmol/l and 0.60, respectively, the sensitivity, specificity and accuracy of sRANKL were 74.0, 84.0 and 77.3%, respectively. Moreover, the sensitivity, specificity and accuracy of the sRANKL/OPG ratio were 84.0, 88.0 and 85.3%, respectively. On the other hand, when the cut-off values of the serum sRANKL level and the sRANKL/OPG ratio were set at 5.24 pmol/l and 0.63, respectively, the sensitivity, specificity and accuracy of sRANKL were 60.0, 84.0 and 68.0%, respectively. The sensitivity, specificity and accuracy of the ratio were 78.0, 64.0 and 73.3%, respectively. The OPG/RANKL system may be involved in the pathogenesis of NSCLC. More importantly, the serum sRANKL level and the sRANKL/OPG ratio may have the potential to be novel biomarkers for the diagnosis of NSCLC.

Krüppel-like factor 9 was down-regulated in esophageal squamous cell carcinoma and negatively regulated beta-catenin/TCF signaling.

Krüppel-like factor 9 (KLF9) has been found to play suppressive roles in several types of tumor. However, the expression pattern and biological functions of KLF9 in esophageal squamous cell carcinoma (ESCC) are still unknown. In this study, it was found that the expression of KLF9 was significantly down-regulated in ESCC compared to their adjacent normal esophageal tissues. Meanwhile, the expression of KLF9 was inversely correlated with the clinical features of ESCC patients. Moreover, in the biological function study, KLF9 was further validated to inhibit the growth, migration, and metastasis of ESCC cells in vitro and in vivo. Mechanistically, KLF9 bind with TCF4 and suppressed the beta-catenin/TCF signaling as well as the expression of its target gene Cyr61. Collectively, our study clarified the function of KLF9 in both ESCC progression and the regulation of beta-catenin/TCF signaling.

Neuroendocrine Neoplasms of the Major Duodenal Papilla With Focus on Histopathological Features and Prognosis.

Neuroendocrine neoplasms (NENs) are generally indolent and progress slowly. However, NENs of major duodenal papilla are uncommon. We retrospectively assessed relevant clinicopathological findings in 9 consecutive patients treated for major duodenal papilla NENs by pancreaticoduodenectomy in our hospital from 2009 to 2013. Eight of the 9 patients (89%) presented with painless obstructive jaundice and one with intermittent fever, attributable to pancreatitis, without jaundice. The diagnostic accuracy was 75% (6/8) for biopsies obtained under duodenoscope guidance. Enhanced multidetector computed tomography detected 89% (8/9) of tumors. Patients with uncertain preoperative diagnoses all underwent intraoperative frozen section pathological diagnosis. Tumor cells extended to at least the muscularis propria in all patients. There were 5 neuroendocrine tumors, 2 neuroendocrine carcinomas, and 2 mixed adenoneuroendocrine carcinomas. Two, 4, and 3 cases were grades 1, 2, and 3, respectively. Grade 3 tumor patients had poor prognoses with tumor recurrence or metastasis within 2 months and all died within 1 year. The overall survival rate (1 year) of grade 3 was lower than in grades 1 and 2 (P < .05). Patients with grade 1 tumors had a similar prognosis to grade 2 (P > .05). To date, only 4 cases of this tumor have been reported in the Chinese literature. The prognosis can be predicted accurately by histopathological features accordingly to the World Health Organization 2010 classification. Multiple imaging techniques and pathological examination should be carried out appropriately to diagnose the disease early.

Serum metabolomic signatures of lymph node metastasis of esophageal squamous cell carcinoma.

Lymph node metastasis was recently proven to be the single most important prognostic factor for esophageal cancer, an important malignant tumor with poor prognosis. A global metabolomics approach was applied to study lymph node metastasis of esophageal squamous cell carcinoma (ESCC). Metabolomics analyses were performed using gas chromatography/mass spectrometry together with univariate and multivariate statistical analyses. There were clear metabolic distinctions between ESCC patients and healthy subjects. ESCC patients could be well-classified according to lymph node metastasis. We further identified a series of differential serum metabolites for ESCC and lymph node metastatic ESCC patients, suggesting metabolic dysfunction in proliferation (aerobic glycolysis, glutaminolysis, fatty acid metabolism, and branched-chain amino acid consumption), apoptosis, migration, immune escape, and oxidative stress of cancer cells in metastatic ESCC patients. In total, three serum metabolites (valine, γ-aminobutyric acid, and pyrrole-2-carboxylic acid) were selected by binary logistic regression analysis, and their combined use resulted in high diagnostic capacity for ESCC metastasis by receiver operating characteristic analysis. The present metabolomics study staged ESCC patients by lymph node metastasis, and the results suggest promising applications of this approach in prognostic prediction, tailored therapeutics, and understanding the pathological mechanisms of poor prognosis of ESCC patients.