Predicting biomolecular interactions is significant for understanding biological systems. Most existing methods for link prediction are based on graph convolution. Although graph convolution methods are advantageous in extracting structure information of biomolecular interactions, two key challenges still remain. One is how to consider both the immediate and high-order neighbors. Another is how to reduce noise when aggregating high-order neighbors. To address these challenges, we propose a novel method, called mixed high-order graph convolution with filter network via LSTM and channel attention (HGLA), to predict biomolecular interactions. Firstly, the basic and high-order features are extracted respectively through the traditional graph convolutional network (GCN) and the two-layer Higher-Order Graph Convolutional Architectures via Sparsified Neighborhood Mixing (MixHop). Secondly, these features are mixed and input into the filter network composed of LayerNorm, SENet and LSTM to generate filtered features, which are concatenated and used for link prediction. The advantages of HGLA are: 1) HGLA processes high-order features separately, rather than simply concatenating them; 2) HGLA better balances the basic features and high-order features; 3) HGLA effectively filters the noise from high-order neighbors. It outperforms state-of-the-art networks on four benchmark datasets.

Heterogeneous feature transfer (HeFT) learning can leverage the semantically related source domain from a different feature space for modeling the target domain with insufficient information. Although HeFT learning has made significant progress, it still faces two major challenges: weak interpretability of the transfer process and underutilization of the hidden information of the heterogeneous source and target domains. To address these two challenges, a framework called heterogeneous feature transfer using fuzzy inference rules (HeFT-FIR) is proposed. The HeFT-FIR framework has two parts: First, design of Takagi–Sugeno–Kang fuzzy systems (TSK-FSs) for the source and target domains, respectively, to achieve HeFT and enhance the interpretability of the transfer process; and second, integration of the HeFT learning mechanism with fuzzy inference rules to optimize the parameters of TSK-FSs and mine the hidden information of the two domains. Based on the framework, a TSK-FS-based heterogeneous feature transfer learning method is then developed with three fuzzy feature space-based learning mechanisms for joint distribution adaptation, local geometric property preservation, and heterogeneous discriminant information extraction, respectively. The mechanisms reduce the difference in distribution between the heterogeneous source and target domains in a common feature subspace, preserve the local geometric properties of two domains, and extract the global discriminant information of them. Extensive analyses are conducted to verify the superiority of the proposed framework and method.

Protein is the most important component in organisms and plays an indispensable role in life activities. In recent years, a large number of intelligent methods have been proposed to predict protein function. These methods obtain different types of protein information, including sequence, structure and interaction network. Among them, protein sequences have gained significant attention where methods are investigated to extract the information from different views of features. However, how to fully exploit the views for effective protein sequence analysis remains a challenge. In this regard, we propose a multi-view, multi-scale and multi-attention deep neural model (MMSMA) for protein function prediction. First, MMSMA extracts multi-view features from protein sequences, including one-hot encoding features, evolutionary information features, deep semantic features and overlapping property features based on physiochemistry. Second, a specific multi-scale multi-attention deep network model (MSMA) is built for each view to realize the deep feature learning and preliminary classification. In MSMA, both multi-scale local patterns and long-range dependence from protein sequences can be captured. Third, a multi-view adaptive decision mechanism is developed to make a comprehensive decision based on the classification results of all the views. To further improve the prediction performance, an extended version of MMSMA, MMSMAPlus, is proposed to integrate homology-based protein prediction under the framework of multi-view deep neural model. Experimental results show that the MMSMAPlus has promising performance and is significantly superior to the state-of-the-art methods. The source code can be found at https://github.com/wzy-2020/MMSMAPlus.

Multiview clustering is an important topic in multiview learning, where the cooperation of different views is used to improve clustering performance. Although multiview clustering has made considerable progress, most existing methods only utilize the information of the original visible views, or only consider some hidden space information shared by different views. Two of the challenges are: 1) insufficient exploitation of cooperative learning between visible and hidden information despite some preliminary attempts, and 2) inadequate consideration of topological information for improving multiview clustering. To meet the challenges, we propose the cooperation enhanced multiview fuzzy clustering method (CE-MVFC) in this article. First, we characterize multiview data with two hidden views, which are obtained by adaptive multiview non-negative matrix factorization (NMF) and fuzzy partition information of each sample in different clusters. Then, we integrated the hidden views and the original visible views to realize visible-hidden cooperation learning. Furthermore, we establish a similarity matrix for each visible view and the hidden view obtained through NMF to describe the data topology in these views. Based on the spatial topological relationship of the samples and the representation of hidden view obtained by fuzzy partition, the network least absolute shrinkage and selection operator is constructed to constrain multiview learning. Finally, we develop the multiview clustering method by exploiting the visible-hidden information cooperation and the spatial topological information constraints. Experiments on benchmark multiview datasets are conducted to demonstrate the highly competitive performance of the proposed CE-MVFC against the state-of-the-art methods.

Multilabel (ML) classification can effectively identify the relevant labels of an instance from a given set of labels. However, the modeling of the relationship between the features and the labels is critical to classification performance. To this end, in this article, we propose a new ML classification method, called ML Takagi-Sugeno-Kang fuzzy system (ML-TSK FS), to improve the classification performance. The structure of ML-TSK FS is designed using fuzzy rules to model the relationship between features and labels. The FS is trained by integrating fuzzy inference-based ML correlation learning with ML regression loss. The proposed ML-TSK FS is evaluated experimentally on 12 benchmark ML datasets. The results show that the performance of ML-TSK FS is competitive with existing methods in terms of various evaluation metrics, indicating that it is able to model the feature-label relationship effectively using fuzzy inference rules and enhances the classification performance.