Edge-aware Graph Attention Networks: Joint Reasoning On Text and Knowledge Graphs for Biomedical Question Answering

Much of the real-world dataset, including textual data, can be represented using graph structures. The use of graphs to represent textual data has many advantages, mainly related to maintaining a more significant amount of information, such as the relationships between words and their types. In recent years, many neural network architectures have been proposed to deal with tasks on graphs. Many of them consider only node features, ignoring or not giving the proper relevance to relationships between them. However, in many node classification tasks, they play a fundamental role. This thesis aims to analyze the main GNNs, evaluate their advantages and disadvantages, propose an innovative solution considered as an extension of GAT, and apply them to a case study in the biomedical field. We propose the reference GNNs, implemented with methodologies later analyzed, and then applied to a question answering system in the biomedical field as a replacement for the pre-existing GNN. We attempt to obtain better results by using models that can accept as input both node and edge features. As shown later, our proposed models can beat the original solution and define the state-of-the-art for the task under analysis.

Learning representations for graph-structured socio-technical systems

The recent widespread use of social media platforms and web services has led to a vast amount of behavioral data that can be used to model socio-technical systems. A significant part of this data can be represented as graphs or networks, which have become the prevalent mathematical framework for studying the structure and the dynamics of complex interacting systems. However, analyzing and understanding these data presents new challenges due to their increasing complexity and diversity. For instance, the characterization of real-world networks includes the need of accounting for their temporal dimension, together with incorporating higher-order interactions beyond the traditional pairwise formalism. The ongoing growth of AI has led to the integration of traditional graph mining techniques with representation learning and low-dimensional embeddings of networks to address current challenges. These methods capture the underlying similarities and geometry of graph-shaped data, generating latent representations that enable the resolution of various tasks, such as link prediction, node classification, and graph clustering. As these techniques gain popularity, there is even a growing concern about their responsible use. In particular, there has been an increased emphasis on addressing the limitations of interpretability in graph representation learning. This thesis contributes to the advancement of knowledge in the field of graph representation learning and has potential applications in a wide range of complex systems domains. We initially focus on forecasting problems related to face-to-face contact networks with time-varying graph embeddings. Then, we study hyperedge prediction and reconstruction with simplicial complex embeddings. Finally, we analyze the problem of interpreting latent dimensions in node embeddings for graphs. The proposed models are extensively evaluated in multiple experimental settings and the results demonstrate their effectiveness and reliability, achieving state-of-the-art performances and providing valuable insights into the properties of the learned representations.

Ontological representation of tumor-node-metastasis classification and an ontology-driven classifier: a study on colorectal cancer

Dissertação de mestrado integrado em Engenharia Biomédica (área de especialização em Informática Médica)

The interaction between N-category and N-ratio as a new tool to improve lymph node metastasis staging in gastric cancer: Results of a single cancer center in Brazil

Background: Depth of tumor invasion (T-category) and the number of metastatic lymph nodes (N-category) are the most important prognostic factors in patients with gastric cancer. Recently, the ratio between metastatic and dissected lymph nodes (N-ratio) has been established as one. The aim of this study is to evaluate the impact of N-ratio and its interaction with N-category as a prognostic factor in gastric cancer. Methods: This was a retrospective study in which we reviewed clinical and pathological data of 165 patients who had undergone curative surgery at our institution through a 9-year period. The exclusion criteria included metastases, gastric stump tumors and gastrectomy with less than 15 lymph nodes dissected. Results: The median age of the patients was 63 years and most of them were male. Total gastrectomy was the most common procedure and 92.1% of the patients had a D2-lymphadenectomy. Their 5-year overall survival was 57.7%. T-category, N-category, extended gastrectomy, and N-ratio were prognostic factors in overall and disease-free survival in accordance with univariate analysis. In accordance with TNM staging, N1 patients who have had NR1 had 5-year survival in 75.5% whereas in the NR2 group only 33% of the cases had 5-year survival. In the multivariate analysis, the interaction between N-category and N-ratio was an independent prognostic factor. Conclusion: Our findings confirmed the role of N-ratio as prognostic factor of survival in patients with gastric cancer surgically treated with at least 15 lymph nodes dissected. The relationship between N-category and N-ratio is a better predictor than lymph node metastasis staging. (C) 2010 Elsevier Ltd. All rights reserved.

Lymph Node Involvement and Not the Histophatologic Subtype Is Correlated with Outcome After Resection of Adenocarcinoma of the Ampulla of Vater

Background Intestinal and pancreaticobiliary types of Vater`s ampulla adenocarcinoma have been considered as having different biologic behavior and prognosis. The aim of the present study was to determine the best immunohistochemical panel for tumor classification and to analyze the survival of patients having these histological types of adenocarcinoma. Method Ninety-seven resected ampullary adenocarcinomas were histologically classified, and the prognosis factors were analyzed. The expression of MUC1, MUC2, MUC5AC, MUC6, CK7, CK17, CK20, CD10, and CDX2 was evaluated by using immunohistochemistry. Results Forty-three Vater`s ampulla carcinomas were histologically classified as intestinal type, 47 as pancreaticobiliary, and seven as other types. The intestinal type had a significantly higher expression of MUC2 (74.4% vs. 23.4%), CK20 (76.7% vs. 29.8%), CDX2 (86% vs. 21.3%), and CD10 (81.4% vs. 51.1%), while MUC1 (53.5% vs. 82.9%) and CK7 (79.1% vs. 95.7%) were higher in pancreatobiliary adenocarcinomas. The most accurate markers for immunohistochemical classification were CDX2, MUC1, and MUC2. Survival was significantly affected by pancreaticobiliary type (p=0.021), but only lymph node metastasis, lymphatic invasion, and stage were independent risk factors for survival in a multivariate analysis. Conclusion The immunohistochemical expression of CDX2, MUC1, and MUC2 allows a reproducible classification of ampullary carcinomas. Although carcinomas of the intestinal type showed better survival in the univariate analysis, neither histological classification nor immunohistochemistry were independent predictors of poor prognosis.

Axillary Lymph Node Dissection for Sentinel Lymph Node Micrometastases May Be Safely Omitted in Early-Stage Breast Cancer Patients: Long-Term Outcomes of a Prospective Study.

OBJECTIVES: To evaluate the long-term disease-free and overall survival of patients with sentinel lymph node (SLN) micrometastases, in whom a completion axillary lymph node dissection (ALND) was systematically omitted. BACKGROUND: The use of step sectioning and immunohistochemistry for SLN analysis results in a more accurate histopathologic examination and a higher detection rate of micrometastases. However, the clinical relevance and therapeutic implications of SLN micrometastases remain a matter of debate. METHODS: In this prospective study, 236 SLN biopsies were performed in 234 consecutive early-stage breast cancer patients (T1, T2 </= 3 cm, cN0 M0) between 1998 and 2002. The SLN were examined by step sectioning and stained with hematoxylin and eosin and immunohistochemistry. None of the patients with negative SLN or SLN micrometastases (International Union Against Cancer classification, >.2 mm to </=2 mm) underwent a completion ALND or radiation to the axilla. Long-term overall and disease-free survivals were compared between patients with negative SLN and those with SLN micrometastases by log rank tests. RESULTS: The SLN was negative in 55% of patients (123 of 224). SLN micrometastases were detected in 27 patients (27 of 224, 12%). After a median follow-up of 77 months (range, 24-106 months), neither locoregional recurrences nor distant metastases occurred in any of the 27 patients with SLN micrometastases. There were no statistically significant differences for overall (P = .656), locoregional (P = .174), and axillary and distant disease-free survival (P = .15) between patients with negative SLN and SLN micrometastases. CONCLUSIONS: This analysis of unselected patients provides evidence that a completion level I and II ALND may be safely omitted in early-stage breast cancer patients with SLN micrometastases.

Morbidity, survival, and site of recurrence after mediastinal lymph-node dissection versus systematic sampling after complete resection for non-small cell lung cancer

BACKGROUND: Mediastinal lymph-node dissection was compared to systematic mediastinal lymph-node sampling in patients undergoing complete resection for non-small cell lung cancer with respect to morbidity, duration of chest tube drainage and hospitalization, survival, disease-free survival, and site of recurrence. METHODS: A consecutive series of one hundred patients with non-small-cell lung cancer, clinical stage T1-3 N0-1 after standardized staging, was divided into two groups of 50 patients each, according to the technique of intraoperative mediastinal lymph-node assessment (dissection versus sampling). Mediastinal lymph-node dissection consisted of removal of all lymphatic tissues within defined anatomic landmarks of stations 2-4 and 7-9 on the right side, and stations 4-9 on the left side according to the classification of the American Thoracic Society. Systematic mediastinal lymph-node sampling consisted of harvesting of one or more representative lymph nodes from stations 2-4 and 7-9 on the right side, and stations 4-9 on the left side. RESULTS: All patients had complete resection. A mean follow-up time of 89 months was achieved in 92 patients. The two groups of patients were comparable with respect to age, gender, performance status, tumor stage, histology, extent of lung resection, and follow-up time. No significant difference was found between both groups regarding the duration of chest tube drainage, hospitalization, and morbidity. However, dissection required a longer operation time than sampling (179 +/- 38 min versus 149 +/- 37 min, p < 0.001). There was no significant difference in overall survival between the two groups; however, patients with stage I disease had a significantly longer disease-free survival after dissection than after sampling (60.2 +/- 7 versus 44.8 +/- 8 months, p < 0.03). Local recurrence was significantly higher after sampling than after dissection in patients with stage I tumor (12.5% versus 45%, p = 0.02) and in patients with nodal tumor negative mediastinum (N0/N1 disease) (46% versus 13%, p = 0.004). CONCLUSION: Our results suggest that mediastinal lymph-node dissection may provide a longer disease-free survival in stage I non-small cell lung cancer and, most importantly, a better local tumor control than mediastinal lymph-node sampling after complete resection for N0/N1 disease without leading to increased morbidity.

Combined radiohormonotherapy in node positive lymph node prostate cancer

Purpose: Pelvic radiation therapy (RT) represents a therapeutic option in the treatment of node-positive prostate cancer but it remains controversial, because of its high rate toxicities. New radiation technique such as IMRT may reduce these complications. In this study, we aimed to assess the rate of toxicities according to CTC-NCI.v3 in such patients treated with either 3DCRT or IMRT (Tomotherapy).Methods and Materials: From January 2008 to December 2010, data were analyzed from 30 consecutive patients including 29 node-positive prostate cancer undergoing definitive or adjuvant RT (IMRT and/or 3DCRT) after radical prostatectomy and lymphadenectomy combined to hormonal therapy. Median age was 66 years (range : 52-83). Median preoperative PSA value was 12 ng/ml (range: 2.72-165). According to the pT-classification, there were 4 pT2, 7 pT3a, 10 pT3b, and 1 pT4 patients. Pathologic positive lymph nodes were found in 23 patients. Radiologic positive lymph nodes were found in 5 patients. Two patients were node negative. Gleason score was ranging between 7 to 10. Twelve patients were treated by Tomotherapy including 4 with simultaneous integrated boost (SIB). Eighteen patients were treated by Tomotherapy including 2 with SIB to the whole pelvis and 3DCRT boost to the prostate. V50% for bladder and rectum were recorded. Acute and late toxicities were assessed according to CTC-NCI.v3 classification.Results: With a median follow-up of 17 months, only one patient presented nodal and metastatic failure. Urinary incontinence was graded 1 after surgery for 6 patients and grade 2 in two. Sexual impuissance was noted in 3 patients. Acute toxicities during RT were proctitis grade 0 in 23 patients (76.5%), grade 1 in 7 (23.5%). Nocturia grade 1 in 9 patients. Interruption of treatment was seen in only case because of grade 3 urinary incontinence. Late effects included erectile dysfunction in 5 patients (83%) and one patient had grade 3proctitis requiring colostomy 3 months after RT. Median Dose-Volume Histogram according to radiation techniques V50% bladder V50% rectum Tomotherapy (IMRT) 36.25 Gy 39 Gy Tomotherapy + 3DCRT 41.26 Gy 39.18 GyConclusion: Based on our above-mentioned findings, there is no a significant difference in morbidity in patients treated with Tomotherapy or Tomotherapy with 3DCRT boost.

Validation and clinical utility of a 70-gene prognostic signature for women with node-negative breast cancer.

BACKGROUND: A 70-gene signature was previously shown to have prognostic value in patients with node-negative breast cancer. Our goal was to validate the signature in an independent group of patients. METHODS: Patients (n = 307, with 137 events after a median follow-up of 13.6 years) from five European centers were divided into high- and low-risk groups based on the gene signature classification and on clinical risk classifications. Patients were assigned to the gene signature low-risk group if their 5-year distant metastasis-free survival probability as estimated by the gene signature was greater than 90%. Patients were assigned to the clinicopathologic low-risk group if their 10-year survival probability, as estimated by Adjuvant! software, was greater than 88% (for estrogen receptor [ER]-positive patients) or 92% (for ER-negative patients). Hazard ratios (HRs) were estimated to compare time to distant metastases, disease-free survival, and overall survival in high- versus low-risk groups. RESULTS: The 70-gene signature outperformed the clinicopathologic risk assessment in predicting all endpoints. For time to distant metastases, the gene signature yielded HR = 2.32 (95% confidence interval [CI] = 1.35 to 4.00) without adjustment for clinical risk and hazard ratios ranging from 2.13 to 2.15 after adjustment for various estimates of clinical risk; clinicopathologic risk using Adjuvant! software yielded an unadjusted HR = 1.68 (95% CI = 0.92 to 3.07). For overall survival, the gene signature yielded an unadjusted HR = 2.79 (95% CI = 1.60 to 4.87) and adjusted hazard ratios ranging from 2.63 to 2.89; clinicopathologic risk yielded an unadjusted HR = 1.67 (95% CI = 0.93 to 2.98). For patients in the gene signature high-risk group, 10-year overall survival was 0.69 for patients in both the low- and high-clinical risk groups; for patients in the gene signature low-risk group, the 10-year survival rates were 0.88 and 0.89, respectively. CONCLUSIONS: The 70-gene signature adds independent prognostic information to clinicopathologic risk assessment for patients with early breast cancer.

Concentric characterization and classification of complex network nodes: Application to an institutional collaboration network

Differently from theoretical scale-free networks, most real networks present multi-scale behavior, with nodes structured in different types of functional groups and communities. While the majority of approaches for classification of nodes in a complex network has relied on local measurements of the topology/connectivity around each node, valuable information about node functionality can be obtained by concentric (or hierarchical) measurements. This paper extends previous methodologies based on concentric measurements, by studying the possibility of using agglomerative clustering methods, in order to obtain a set of functional groups of nodes, considering particular institutional collaboration network nodes, including various known communities (departments of the University of Sao Paulo). Among the interesting obtained findings, we emphasize the scale-free nature of the network obtained, as well as identification of different patterns of authorship emerging from different areas (e.g. human and exact sciences). Another interesting result concerns the relatively uniform distribution of hubs along concentric levels, contrariwise to the non-uniform pattern found in theoretical scale-free networks such as the BA model. (C) 2008 Elsevier B.V. All rights reserved.

Two-way regionalized classification of multivariate datasets and its application to the assessment of hydrodynamic dispersion

Zones of mixing between shallow groundwaters of different composition were unravelled by two-way regionalized classification, a technique based on correspondence analysis (CA), cluster analysis (ClA) and discriminant analysis (DA), aided by gridding, map-overlay and contouring tools. The shallow groundwaters are from a granitoid plutonite in the Funda o region (central Portugal). Correspondence analysis detected three natural clusters in the working dataset: 1, weathering; 2, domestic effluents; 3, fertilizers. Cluster analysis set an alternative distribution of the samples by the three clusters. Group memberships obtained by correspondence analysis and by cluster analysis were optimized by discriminant analysis, gridded memberships as follows: codes 1, 2 or 3 were used when classification by correspondence analysis and cluster analysis produced the same results; code 0 when the grid node was first assigned to cluster 1 and then to cluster 2 or vice versa (mixing between weathering and effluents); code 4 in the other cases (mixing between agriculture and the other influences). Code-3 areas were systematically surrounded by code-4 areas, an observation attributed to hydrodynamic dispersion. Accordingly, the extent of code-4 areas in two orthogonal directions was assumed proportional to the longitudinal and transverse dispersivities of local soils. The results (0.7-16.8 and 0.4-4.3 m, respectively) are acceptable at the macroscopic scale. The ratios between longitudinal and transverse dispersivities (1.2-11.1) are also in agreement with results obtained by other studies.

Clinicopathological features influencing pelvic lymph node metastasis and vaginal and parametrial involvement in patients with carcinoma of the cervix

Purpose: This study was undertaken to evaluate clinical and pathologic findings that predicted pelvic lymph node metastasis and parametrial and vaginal involvement in patients with stage IB carcinoma of the cervix. Methods: 71 patients with diagnosis of stage IB (FIGO) cervical cancer were prospectively studied from December 1997 to August 2002. The patient's age, clinical stage (IB1 or IB2), histological classification, grade of differentiation, tumor volume, and lymphatic vascular space invasion (LVSI) were evaluated. Statistical methods included chi(2) test and Fisher's exact test to evaluate significant differences between the groups. The level of significance was set at p < 0.05. Results: the clinical stage was IB1 in 51 patients (71.8%) and IB2 in 20 patients (28.2%). The histological classification identified squamous cell carcinoma in 60 patients (84.5%) and adenocarcinoma in 11 patients (15.5%). The average tumoral volume was 22.8 &PLUSMN; 8 24.3 cm(3) (0.3-140.0 cm(3)). The tumor was well differentiated (G1) in 8 (11.3%), moderately differentiated (G2) in 40 (56.3%) and poorly differentiated in 23 (32.4%) of the cases. The presence of LVSI was detected in 14 patients (19.7%) and was associated with pelvic lymph node metastasis and vaginal and parametrial involvement (p = 0.002, p = 0.001 and p < 0.001; respectively). The average number of positive pelvic lymph nodes was significantly higher in the patients with LVSI compared with patients without LVSI (2.47 +/- 2.8 vs. 0.33 +/- 0.74; p = 0.001). There was no association of age, clinical stage, histological classification, grade of differentiation or tumor volume with pelvic lymph node metastasis and vaginal and parametrial involvement. Conclusion: the presence of LVSI is significantly associated with pelvic lymph node metastasis and vaginal and parametrial involvement in patients with stage IB cervical carcinoma. Copyright (C) 2005 S. Karger AG, Basel.

A new approach to soil classification mapping based on the spatial distribution of soil properties

Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES)

Topoisomerase expression in oral squamous cell carcinoma: relationship with cancer stem cells profiles and lymph node metastasis

BACKGROUND: The relationship between predictive proteins and tumors presenting cancer stem cells (CSCs) profiles in oral tumors is still poorly understood. This study aims to identify the relationship between topoisomerases I, II alpha, and III alpha and putative CSCs immunophenotype in oral squamous cell carcinoma (OSCC) and determine its influence on prognosis. METHODS: The following data were retrieved from 127 patients: age, gender, primary anatomic site, smoking and alcohol intake, recurrence, metastases, histologic classification, treatment, and survival. An immunohistochemical study for topoisomerases I, II alpha, and III alpha was performed in a tissue microarray containing 127 paraffin blocks of OSCCs. RESULTS: In univariate analysis, topoisomerases expression showed significant differences according to CSCs profiles and p53 immunoexpression, but not with survival. Topoisomerases II alpha and III alpha also showed significant relationship with lymph node metastasis. The multivariate test confirmed these associations. CONCLUSIONS: The results that all topoisomerases correlates with OSCC CSCs may indicate a role for topoisomerases in head and neck carcinogenesis. Notwithstanding, it is plausible that other members of topoisomerases family could represent novel therapeutical targets in oral squamous cell carcinoma. J Oral Pathol Med (2012) 41: 762-768