A large number of real-world graphs or networks are inherently heterogeneous, involving a diversity of node types and relation types. Heterogeneous graph embedding is to embed rich structural and semantic information of a heterogeneous graph into low-dimensional node representations. Existing models usually define multiple metapaths in a heterogeneous graph to capture the composite relations and guide neighbor selection. However, these models either omit node content features, discard intermediate nodes along the metapath, or only consider one metapath. To address these three limitations, we propose a new model named Metapath Aggregated Graph Neural Network (MAGNN) to boost the final performance. Specifically, MAGNN employs three major components, i.e., the node content transformation to encapsulate input node attributes, the intra-metapath aggregation to incorporate intermediate semantic nodes, and the inter-metapath aggregation to combine messages from multiple metapaths. Extensive experiments on three real-world heterogeneous graph datasets for node classification, node clustering, and link prediction show that MAGNN achieves more accurate prediction results than state-of-the-art baselines.

There is an important need for clinically relevant animal models for human cancers. Toward this goal, histologically intact human colon-cancer specimens derived surgically from patients were implanted orthotopically to the colon or cecum of nude mice. We have observed extensive orthotopic growth in 13 of 20 cases of implanted patient colon tumors. These showed various growth patterns with subsequent regional, lymph-node, and liver metastasis, as well as general abdominal carcinomatosis. Thus, models for human colon cancer have been developed that show (i) local growth, (it) abdominal metastasis, (iiM) general abdominal carcinomatosis with extensive peritoneal seeding, (iv) lymph-node metastasis, (v) liver metastasis, and (vW) colonic obstruction. These models permit the passage of the tumors to form large cohorts. They will facilitate research into the biology of colon cancer metastatic capability and the development of new drugs active against metastatic cancer. These models may also predict the clinical course and the in vivo response to drugs of the cancer of individual patients.There is a need for the development of better animal models for human cancer. Models based on athymic nude mice have been used for this purpose. However, metastatic rates from subcutaneous or intramuscular xenografts have been low or nonexistent even from tumors that were highly metastatic in the patient from whom the tissue was derived (1-5).Recent work from a number of laboratories has indicated that implanting human tumor cells orthotopically in the corresponding organ of nude mice resulted in much higher metastatic rates. For example, a human renal-cell carcinoma obtained from a surgical specimen was dissociated by enzymatic treatment and subcutaneously injected into the renal capsule of nude mice as well as other sites. The injection of human renal-cell carcinoma cells into the kidney ofnude mice produced the highest incidence oftumor establishment and of metastasis to the lungs and other peritoneal organs. The nude-mouse renal capsule appears to be a most advantageous site for implantation of human renal-cell carcinoma (6-8). However, the subrenal capsule may be an advantageous implant site for other tumor types also (9). Human coloncancer cells were dissociated, grown in culture, and subsequently injected into the cecum of nude mice to produce tumors that eventually metastasized to the liver, demonstrating that orthotopic implantation can enhance the metastatic capability of human tumor cells in nude mice (5, 10-13). Similar results also have been achieved for orthotopic implantation of cell lines of human lung cancer (14), human pancreatic cancer (15), bladder cancer (16, 17), melanoma (18, 19), breast cancer (20-22), and head and neck cancer (23). It should be noted, however, that the effects of orthotopicity have not been fully evaluated in that, at least in some cases, metastasis may arise from nonorthotopic sites.Our approach is to avoid disruption of tumor integrity and to orthotopically implant histologically int...

Pancreatic cancer is one of the most intractable and least understood of all human cancers. Pancreatic cancer is the fourth-leading cause ofcancer-related mortality in the United States with <2% of the patients surviving for 5 yr.In an effort to help develop more effective treatment moities for pancreatic cancer and improve detection, we report an animal model for individual human pancreatic-cancer patients. The model involves orthotopic transplantation of histologically intact pancreatic-cancer specimens -to the nude-mouse pancreas, which can result in models that resemble the clinical picture including (i) extensive local tumor growth, (it) extension of the locally growing human pancreatic cancer to the nudemouse stomach and duodenum, (Wi) metastases of the human pancreatic tumor to the nude-mouse liver and regional lymph nodes, and (iv) distant metastases of the human pancreatic tumor to the nude-mouse adrenal gland, diaphragm, and mediastinal lymph nodes. In a series of five patient cases, a 100% take rate has been demonstrated, and of 17 mice transplanted, 15 supported tumor growth. Immunohistochemical analysis of the antigenic phenotype of the transplanted human pancreatic tumors showed a similar pattern of expression of two different human tumor-associated antigens, such as tumor-associated glycoprotein 72 and carcinoembryonic antigen in the transplanted tumors when compared with the original surgical biopsy, suggesting similarity between the two. This model should, therefore, prove valuable for treatment evaluation of individual cancer patients, as well as for evaluation of experimental treatment modalities for this disease. Our approach is to avoid disruption of tumor integrity and to orthotopically implant histologically intact patient tumor tissue directly after surgery or biopsy. Such a model should better resemble the original properties of the human cancer and could be ofgreat value in developing additional drugs and treatment strategies for cancer. Guided by this overall strategy, we have used nude mice to construct human coloncancer models that directly use surgical specimens and can exhibit the variety of clinical behaviors seen in human subjects (31). These behaviors include (i) local growth, (ii) abdominal metastasis, (iii) general abdominal carcinomatosis with extensive peritoneal seeding, (iv) lymph-node metastasis, (v) liver metastasis, and (vi) colonic obstruction: a tumor-establishment rate of 13 cases in 20 attempts was found (31). We have constructed a similar set of models for human bladder cancer (32).We report here the use of the orthotopic-transplant strategy of histologically intact patient specimens to develop a human pancreatic-cancer model with a 100% take rate and subsequent growth and metastatic behavior while retaining human tumor-associated antigens (TAAs), thereby resembling the clinical picture. Cancer of the pancreas is one of the most intractable cancers and is the fourth-leading cause of cancer death in the United States (1-3). At surgery, most patients are foun...

The P582S HIF-1alpha is a stable variant and HIF-1alpha mutation is a mechanism for enhancing HIF-1alpha activity in human cancer. The recent identification of the identical P582S HIF-1alpha as a polymorphism suggests that this variant may increase tumor susceptibility or cause more aggressive biological behavior.