Drug repositioning has become a transformative paradigm in pharmaceutical research, with accurate prediction of drug-disease associations serving as its computational cornerstone. While graph convolutional networks have shown promising results in this domain, their architecture designed for homogeneous graphs fundamentally limits the effectiveness in modeling heterogeneous biological networks with diverse entity types and interactions, especially the propagation of collaborative signals across different entities. In this study, we propose HeteroNGCF, a network representation learning framework that effectively models high-order connectivity patterns in the drug-gene-proteindisease network while preserving the semantic heterogeneity of different entities. Specifically, HeteroNGCF firstly projects node features in homogeneous graphs into a unified latent space, enabling the alignment of heterogeneous entities while preserving their distinct semantic properties. To capture collaborative signals latent in heterogeneous interactions, we employ the neural graph collaborative filtering technique that integrates bipartite graph structures into the embedding process by recursively aggregating neighborhood information through layer-wise propagation. Afterwards, we aggregate embeddings learned from diverse interaction networks to obtain expressive representations of drugs and diseases, thereby enhancing prediction performance. Extensive experiments demonstrate that HeteroNGCF outperforms baseline methods across all evaluation metrics (AUC, AUPR, F1-score, recall and precision).

Benefit from the remarkable development of cloud computing, massive Web APIs have been published and numerous innovative mashups have emerged on the Internet. The rapid increase of available Web APIs brings a significant challenge on how to discover suitable APIs for mashup creation. Various approaches have been proposed to recommend Web APIs, which mainly focus on employing collaborative filtering techniques such as matrix factorization and deep learning techniques such as graph neural network. However, existing studies suffer from a few limitations: collaborative filtering based methods are not able to capture the high-order connectivity information between APIs and mashups, and deep learning based methods are not expressive enough to capture the diversity of connectivity patterns due to the simple graph structure. To tackle the limitations, we proposes a biased random walk based Web API recommendation approach. We first construct a heterogeneous network to model different types of connectivity. And then we design a strategy in terms of neighbor nodes’ types to guide the biased random walk and to capture diversity of high-order connectivity patterns. Finally we generate candidate Web APIs for mashups based on their embedding vectors. Experiments conducted on a real-world dataset show that our approach improves in precision by 13.89%, in recall by 17.16%, in F1 by 15.86%, and in NDCG by 2.85%.

Android has recently become one of the leading operating systems for mobile app development. The permission- based mechanism in Android forces app developers to determine permissions required by apps besides implementing the functionality, which increases the burden on developers. App permission recommendation becomes necessary and meaningful to assist developers determine appropriate needed permissions. Existing approaches for app permission recommendation have various limitations, such as suffering from the cold-start problem, needing to learn both of the app and permission embedding matrices. To address these issues, we define a sparse matrix factorization model, in which API categories are utilized as latent factors, app-API calls are applied for app representation, and only one sparse matrix is to be learned for permission representation. We further present an efficient approach by utilizing the Alternating Direction Method of Multipliers to solve the optimization problem. We conduct a comprehensive set of experiments on a real-world dataset, which show that our approach outperforms the state-of-the-art approaches in terms of four well-known metrics.

The increasing number of studies have shown that lncRNAs are involved in various biological processes and play crucial roles in many complex human diseases. The small-size of validated lncRNA-disease associations creates a pressing demand to develop effective computational methods for inferring potential associations. However, most of existing methods suffer from the insufficiency of embedding representations, namely they merely extract features of lncRNAs and diseases from themselves, while leave out the rich information contained in their neighbors. In this paper, we propose a network based representation learning method that extracts high-level features from a node as well as its neighborhood to predict associations between lncRNAs and diseases. Firstly, we construct the lncRNA-disease network by taking lncRNA/disease similarities and lncRNA-disease associations into account. And then, we build the initial embedding of a lncRNA (a disease) with the combination of one-hot vector and probability vector. Subsequently, we utilize message passing on the network and augment the embedding of a lncRNA (a disease) by capturing messages propagated from its neighbor nodes. Finally, we predict lncRNA-disease associations based on representations that are not only from themselves but also propagated from neighbors. To generate a balanced training dataset, we select reliable negative examples from unlabeled samples. Experimental results show that our approach can achieve better performance than other existing methods.

More and more evidences have shown that lncRNAs are involved with various complex human diseases. The small minority of experimentally validated lncRNA-disease associations have created an urgent demand for computational prediction models. Although many related approaches have been proposed, there still have much room for improvement. To address the cold-start problem and accurately represent associations, this paper considers the prediction of lncRNA-disease associations as a recommendation problem and proposes a matrix factorization and neural network based method. Firstly, to better represent lncRNAs and diseases, their embeddings are learned based on matrix factorization. And then, features of associations are represented by integrating embeddings of lncRNAs and diseases. Finally, neural network is applied for predicting potential associations. Experimental results show that our method can achieve better performance than the state-of-the-art approaches from several perspectives.

The permission mechanism in Android imposes additional requirements on app developers, since developers have to learn not only the APIs to be used, but also the permissions to be declared. Recommending permissions for apps becomes necessary and meaningful to help developers determine suitable permissions to be declared in apps. Previous studies suffer from the cold-start problem and do not consider the fact that categories of APIs invoked by apps may influence permissions required by apps, since APIs with similar usage may request same permissions. To address these issues, this paper proposes a Category aware Non-negative Matrix Factorization (CNMF) framework to recommend app permissions. The framework firstly calculates semantic similarities among APIs based on word embeddings and clusters similar APIs into the same category, and then computes the probabilities of apps using APIs in each category and integrates the app-category information into the non-negative matrix factorization. Experimental results on a real-world dataset show that our framework can achieve better performance than the state-of-the-art approaches.

With the advent of Web services, service interoperability has always been an active research issue. In recent years, many approaches have been proposed. However, how to achieve fast composition and guarantee correct and executable composite service remains an open issue. For this problem, this paper presents a three-phase framework for accurate and efficient service interoperability. Since service collaborations should follow certain constraints for success invocation, the goal of the first phase is to automatically clarify constraints on Web services. And then the second phase utilizes constraints acquired in previous phase to check Web services' constraint compatibility for accurate collaborations. In order to reduce the time on exhaustive analysis of service matching, the concept of expanded parameters is proposed, thus the problem of semantic matching is transformed into set operation. Subsequently, the third phase achieves interoperability among Web services by constructing initial composite services on the basis of collaborations, optimizing initial compositions to generate minimal composition alternatives with no redundant Web services, and executing final composition services. Experimental results show that our framework can dramatically reduce the time spending on service matching and effectively generate minimal composition alternatives in a rather short time.

Service relations facilitate the automation of service reuse. Most of studies on the service relations focus on the inputs and outputs. However, different Web Services tend to utilize the same parameters without formally specifying their constraints. Due to this, the semantics, introduced by semantic annotation, is still not rich enough for accurate descriptions, thus generating a large number of inappropriate service relations. To address this, we propose an approach for augmenting semantics of Web Services based on constraints, which can be regarded as a complement to semantic annotation. The semantics is augmented via a hybrid analysis of heterogeneous constraints, including the server constraint and object constraint.