Both explicit and implicit feedbacks can reflect user opinions on items, which are essential for learning user preferences in recommendation. However, most current recommendation algorithms merely focus on implicit positive feedbacks (e.g., click), ignoring other informative user behaviors. In this paper, we aim to jointly consider explicit/implicit and positive/negative feedbacks to learn user unbiased preferences for recommendation. Specifically, we propose a novel Deep feedback network (DFN) modeling click, unclick and dislike behaviors. DFN has an internal feedback interaction component that captures fine-grained interactions between individual behaviors, and an external feedback interaction component that uses precise but relatively rare feedbacks (click/dislike) to extract useful information from rich but noisy feedbacks (unclick). In experiments, we conduct both offline and online evaluations on a real-world recommendation system WeChat Top Stories used by millions of users. The significant improvements verify the effectiveness and robustness of DFN. The source code is in https://github.com/qqxiaochongqq/DFN.

The increase in biological data and the formation of various biomolecule interaction databases enable us to obtain diverse biological networks. These biological networks provide a wealth of raw materials for further understanding of biological systems, the discovery of complex diseases and the search for therapeutic drugs. However, the increase in data also increases the difficulty of biological networks analysis. Therefore, algorithms that can handle large, heterogeneous and complex data are needed to better analyze the data of these network structures and mine their useful information. Deep learning is a branch of machine learning that extracts more abstract features from a larger set of training data. Through the establishment of an artificial neural network with a network hierarchy structure, deep learning can extract and screen the input information layer by layer and has representation learning ability. The improved deep learning algorithm can be used to process complex and heterogeneous graph data structures and is increasingly being applied to the mining of network data information. In this paper, we first introduce the used network data deep learning models. After words, we summarize the application of deep learning on biological networks. Finally, we discuss the future development prospects of this field.

The utility of cell-free nucleic acids in monitoring cancer has been recognized by both scientists and clinicians. In addition to human transcripts, a fraction of cell-free nucleic acids in human plasma were proven to be derived from microbes and reported to have relevance to cancer. To obtain a better understanding of plasma cell-free RNAs (cfRNAs) in cancer patients, we profiled cfRNAs in ~300 plasma samples of 5 cancer types (colorectal cancer, stomach cancer, liver cancer, lung cancer, and esophageal cancer) and healthy donors (HDs) with RNA-seq. Microbe-derived cfRNAs were consistently detected by different computational methods when potential contaminations were carefully filtered. Clinically relevant signals were identified from human and microbial reads, and enriched Kyoto Encyclopedia of Genes and Genomes pathways of downregulated human genes and higher prevalence torque teno viruses both suggest that a fraction of cancer patients were immunosuppressed. Our data support the diagnostic value of human and microbe-derived plasma cfRNAs for cancer detection, as an area under the ROC curve of approximately 0.9 for distinguishing cancer patients from HDs was achieved. Moreover, human and microbial cfRNAs both have cancer type specificity, and combining two types of features could distinguish tumors of five different primary locations with an average recall of 60.4%. Compared to using human features alone, adding microbial features improved the average recall by approximately 8%. In summary, this work provides evidence for the clinical relevance of human and microbe-derived plasma cfRNAs and their potential utilities in cancer detection as well as the determination of tumor sites.

Chronic stress affects brain function, so assessing its hazards is important for mental health. To overcome the limitations of behavioral data, we combined behavioral and event-related potentials (ERPs) in an attention network task. This task allowed us to differentiate between three specific aspects of attention: alerting, orienting, and execution. Forty-one participants under chronic stress and 31 non-stressed participants were enrolled. On the performance level, the chronically stressed group showed a significantly slower task response and lower accuracy. Concerning ERP measures, smaller cue-N1, cue-N2, and larger cue-P3 amplitudes were found in the stressed group, indicating that this group was less able to assign attention to effective information, i.e., they made inefficient use of cues and had difficulty in maintaining alerting. In addition, the stressed group showed larger target-N2 amplitudes, indicating that this group needed to allocate more cognitive resources to deal with the conflict targets task. Subgroup analysis revealed lower target-P3 amplitudes in the stressed than in the non-stressed group. Group differences associated with the attention networks were found at the ERP level. In the stressed group, excessive depletion of resources led to changes in attention control. In this study, we examined the effects of chronic stress on individual executive function from a neurological perspective. The results may benefit the development of interventions to improve executive function in chronically stressed individuals.