Unobserved Interactions Research Articles

RevGNN: Negative Sampling Enhanced Contrastive Graph Learning for Academic Reviewer Recommendation

Acquiring reviewers for academic submissions is a challenging recommendation scenario. Recent graph learning-driven models have made remarkable progress in the field of recommendation, but their performance in the academic reviewer recommendation task may suffer from a significant false negative issue. This arises from the assumption that unobserved edges represent negative samples. In fact, the mechanism of anonymous review results in inadequate exposure of interactions between reviewers and submissions, leading to a higher number of unobserved interactions compared to those caused by reviewers declining to participate. Therefore, investigating how to better comprehend the negative labeling of unobserved interactions in academic reviewer recommendations is a significant challenge. This study aims to tackle the ambiguous nature of unobserved interactions in academic reviewer recommendations. Specifically, we propose an unsupervised Pseudo Neg-Label strategy to enhance graph contrastive learning (GCL) for recommending reviewers for academic submissions, which we call RevGNN. RevGNN utilizes a two-stage encoder structure that encodes both scientific knowledge and behavior using Pseudo Neg-Label to approximate review preference. Extensive experiments on three real-world datasets demonstrate that RevGNN outperforms all baselines across four metrics. Additionally, detailed further analyses confirm the effectiveness of each component in RevGNN.

A VAE-Based User Preference Learning and Transfer Framework for Cross-Domain Recommendation

The core idea of cross-domain recommendation is to alleviate the problem of data scarcity. Previous methods have made brilliant successes. However, many of them mainly focus on learning an ideal mapping function across-domains, ignoring the user preferences within a specific domain, which leads to suboptimal results. In this paper, we propose a Cross-Domain Recommendation Variational AutoEncoder framework (CDRVAE), a novel extension of a variational autoencoder on cross-domain recommendations for user behaviour distribution modeling. It applies a new hybrid architecture of VAE as the backbone and simultaneously constructs two information flows, within-domain and cross-domain modeling. For the former, an asymmetric codec structure is designed to reconstruct preference distribution from domain-specific latent factors. To relieve the posterior collapse dilemma, a combined prior is employed to increase the distribution complexity. The equivalent transition by a transformation matrix and the unobserved interaction generation by cross-domain reconstruction contribute to the latter. We combine all the above components for the more accurate and reliable user features. Extensive experiments are conducted on three public benchmark datasets to validate the effectiveness of the proposed CDRVAE. Experimental results demonstrate that CDRVAE is consistently superior to other state-of-the-art alternative baseline models.

Structure and Molecular Mechanism of Signaling for the Glucagon-like Peptide-1 Receptor Bound to Gs Protein and Exendin-P5 Biased Agonist.

The glucagon-like peptide-1 receptor (GLP-1R) is a key regulator of blood glucose and a prime target for the treatment of type II diabetes and obesity with multiple public drugs. Here we present a comprehensive computational analysis of the interactions of the activated GLP-1R-Gs signaling complex with a G protein biased agonist, Exendin P5 (ExP5), which possesses a unique N-terminal sequence responsible for the signal bias. Using a refined all-atom model of the ExP5-GLP-1R-Gs complex in molecular dynamics (MD) simulations, we propose a novel mechanism of conformation transduction in which the unique interaction network of ExP5 N-terminus propagates the binding signal across an array of conserved residues at the transmembrane domain to enhance Gs protein coupling at the cytoplasmic end of the receptor. Our simulations reveal previously unobserved interactions important for activation by ExP5 toward GDP-GTP signaling, providing new insights into the mechanism of class B G protein-coupled receptor (GPCR) signaling. These findings offer a framework for the structure-based design of more effective therapeutics.

Predicting nonlinear dynamics of short-lived penaeid shrimp species in the Gulf of Mexico

Predicting the dynamics of harvested species is essential for assessing stock status and establishing index-based management strategies. However, conventional approaches for short-lived species predict dynamics poorly, possibly because unobserved interactions with other species and abiotic factors are often treated as noise. Alternatively, the empirical dynamic modeling (EDM) approach, which uses the time delays of the observed states to compensate for unobserved interactions, may improve the predictions for short-lived species. We test this idea using time series data of two federally managed, short-lived penaeid shrimp species, whose abundances were surveyed over 30 years (1987–2018) across the US Gulf of Mexico. We show that ( i) abundance dynamics of these annual shrimp stocks are well-predicted by EDM, ( ii) the dynamics are spatially similar across most of the gulf, and ( iii) the stock dynamics are characterized by nonlinear density-dependent interaction and vary with temperature. Our findings suggest that EDM may be more responsive than single-species, catch-at-age models in assessing the stock dynamics for short-lived penaeid shrimp species.

GraphTGI: an attention-based graph embedding model for predicting TF-target gene interactions

Interaction between transcription factor (TF) and its target genes establishes the knowledge foundation for biological researches in transcriptional regulation, the number of which is, however, still limited by biological techniques. Existing computational methods relevant to the prediction of TF-target interactions are mostly proposed for predicting binding sites, rather than directly predicting the interactions. To this end, we propose here a graph attention-based autoencoder model to predict TF-target gene interactions using the information of the known TF-target gene interaction network combined with two sequential and chemical gene characters, considering that the unobserved interactions between transcription factors and target genes can be predicted by learning the pattern of the known ones. To the best of our knowledge, the proposed model is the first attempt to solve this problem by learning patterns from the known TF-target gene interaction network. In this paper, we formulate the prediction task of TF-target gene interactions as a link prediction problem on a complex knowledge graph and propose a deep learning model called GraphTGI, which is composed of a graph attention-based encoder and a bilinear decoder. We evaluated the prediction performance of the proposed method on a real dataset, and the experimental results show that the proposed model yields outstanding performance with an average AUC value of 0.8864 +/- 0.0057 in the 5-fold cross-validation. It is anticipated that the GraphTGI model can effectively and efficiently predict TF-target gene interactions on a large scale. Python code and the datasets used in our studies are made available at https://github.com/YanghanWu/GraphTGI.

Variational Autoencoder-Based Hybrid Recommendation With Poisson Factorization for Modeling Implicit Feedback

Hybrid recommendation, which is based on collaborative filtering and supplemented with auxiliary content information, is being actively researched due to its ability to overcome the cold-start problem. Many proposed hybrid methods make recommendations using Gaussian distribution-based collaborative filtering even though they handle variables that tend to be non-Gaussian, such as the number of interactions. We present a method that uses a hybrid recommendation framework based on collaborative filtering that models the number of interactions as a Poisson-distributed and variational autoencoder-based content information generation process that shares latent variables with collaborative filtering. As a prior for the shared latent variables, we use a gamma distribution, which is a conjugate prior of a Poisson distribution. An implicit-derivative-based reparameterization trick enables the use of a gamma distribution in a variational autoencoder. The latent variables in the generative model are inferred using the stochastic gradient variational Bayes approach, taking the number of interactions corresponding to users and items and content information as input. In accordance with the inference, unobserved interactions between users and items are predicted for recommendation. The use of a neural-network-based generative model for content information enables the framework to handle various types of content information. Experimental results show that the proposed method utilizes content information effectively for predicting the number of interactions and that it should aid in overcoming the cold-start problem.

A New Econometric Approach for Modeling Several Count Variables: A Case Study of Crash Frequency Analysis by Crash Type and Severity

There is limited adoption of research modeling crash severity frequency considering different crash types due to the challenge associated with analyzing large number of dependent variables. The proposed research contributes to burgeoning econometric and safety literature by developing a joint modeling approach that can accommodate for several dependent variables within a parsimonious structure. By recasting the analysis levels for dependent variables, the proposed approach allows for flexible consideration of crashes by type and severity within a single framework. Specifically, we employ a Panel Mixed Negative Binomial- Generalized Ordered Probit Fractional Spilt (PMNB-GOPFS) model where the first component (NB) accommodates for crash frequency by crash type and the later component (GOPFS) studies the fraction of severity outcome for different crash types. The proposed model system increases interaction between dependent variables through observed variables thus reducing the dependency on unobserved interactions across dependent variables. Thus, the proposed approach allows for the estimation of parsimonious specifications reducing the need for computationally intensive simulation based estimation. The proposed system is also flexible to accommodate for common unobserved effects including: 1) common unobserved factors simultaneously affecting crash counts of different crash types; 2) common unobserved factors simultaneously affecting crash severity proportions of different crash types; and 3) common unobserved factors that simultaneously impact crash counts and severity proportions by different crash types. The model system performance is illustrated using a simulation study. The empirical analysis was conducted using zonal level crash count data for the year 2016 from Central Florida while considering a comprehensive set of exogenous variables including roadway, built environment, land-use, traffic and sociodemographic characteristics. To illustrate the applicability of our proposed system, we carried out a comparison exercise between our proposed joint PMNB-GOPFS and the traditional multivariate system for predicting crash counts across different crash severities. The resulting goodness of fit measures clearly highlight the superior/equivalent performance of the proposed PMNB-GOPFS model over the traditional RPMNB model with less than half the number of parameters. The proposed framework can predict several dimensions including total crash counts, total crash counts by crash types, crash counts for each severity level and finally, proportions and counts of crashes for each crash type by severity.

Improving Business Pages Recommendation in Social Network Using Link Prediction Methods

Recently Social Network has become one of the favorite means for a modern society to perform social interaction and exchange information via the internet. Link prediction is a common problem that has broad application in such social networks, ranging from predicting unobserved interaction to recommending related items. In this paper, we investigate link recommendations over business pages on Facebook Social Network. More specifically, given a company in thenetwork, we want to recommend potential companies to connect with. We start by introducing existing work in link recommendations and some link prediction models as our baseline. We then talk about the Graph Neural Network model SEAL to make a link recommendations in the network. Our results show that SEAL outperformed the compared baseline model while reaching above 94% Area Under Curve accuracy in link recommendations.

Adoption of agricultural technologies among rice farmers in Benin

AbstractAdoption of agricultural technologies in Benin remains at a very low level. This paper analyzes the factors that determine the adoption of agricultural technologies by rice producers in Benin. It employs a simple probit and Poisson regression models, as well as a multivariate probit model to account for the unobserved interaction between technology adoption decisions. Results reveal that variables such as education, access to extension services, membership of a farmers‐based organization, access to credit, media and use of a mobile phone are important in the process of increasing adoption of agricultural technologies. These variables must be taken into consideration in the elaboration of the agricultural modernization policy.

An Application of Social Network Analysis to Focus Group Discussions: Unobserved Interaction Between Participants and Discussion Topics

While focus group discussions have become popular across the communication research field, some researchers have pointed out the importance of studying group interactions within focus groups. However, most researchers used a qualitative approach and mostly explored the interaction between participants or between participants and the moderator. This study claimed there might be a third and unobservable interaction that occurs between participants and structured discussion topics. An affiliation network in social network analysis was applied to investigate this phenomenon among eight groups of existing focus group conversations. By affiliation network data and network visualization, this study demonstrated diverse structures between groups. The interaction between participants and discussion topics was significantly different between at least two groups. The results showed that social network analysis could be a tool to help examine focus group interactions from qualitative data, and the interaction between participants and discussion topics could be a potential third interaction in focus group studies. This study is limited by comparing the mean difference between groups. Future research may examine short- and long-term causal effects of group interaction on participation or attitude changes after focus group discussions.

The Relationship Between School Engagement and Conscious Self-regulation of Learning Activity: The Current State of the Problem and Research Perspectives in Russia and Abroad

The article presents a theoretical and methodological analysis of the problem of school engagement, the research of which is highly relevant nowadays in foreign educational psychology. School engagement is a reliable marker of efficient learning organisation and also an important academic result, the pursuit of which indicates the efficiency of the school educational environment. The purpose of the article was to analyse the current state of the problem of school engagement and the specifics of the relationship between school engagement and self-regulation of learning activity in Russian and foreign psychology. The authors formulated a working definition of engagement, according to which school engagement is understood as a stable, directed, active participation of students both in learning activities and in school life in general, including observable and unobservable interactions with the academic social environment; assessed various models of school engagement used for research and practice; substantiated the efficiency of the approach, according to which school engagement is viewed as a multidimensional construct, including behavioural, emotional and cognitive components; examined the differences between school engagement and academic motivation; and analysed the specific features of the main components of school engagement and their relationship with self-processes, in particular, with self-regulation. It was also noted that, in Russian psychology, issues related to the study of school engagement and the specifics of its relationship with various psychological phenomena have not yet received widespread consideration. In conclusion, the authors outlined the perspectives for verifying foreign models of school engagement and diagnostic tools on Russian samples as well as studying conscious self-regulation of learning activity as a resource for student engagement in different learning periods.

Simultaneous Control in Belief Space for Circular Insertion in Precision Assembly

Simultaneously inserting multiple objects is an essential topic in precision assembly to compose complicated shapes. This task involves the acquisition difficulty of unobservable interaction states, which makes it hard to plan the insertion. To solve it, this article investigates the circular assembly of multiple objects and proposes a strategy to control the simultaneous insertion in belief space. We first present the insertion state transition and observation models, in which the stochastic parts are modeled as Gaussian noise, and then estimate the belief state using an extended Kalman filter. An optimization approach is discussed, for the compensational movement planning, to decrease the estimated radial interaction forces and the toward-center movement is thus determined, considering the optimized compensational movement and the belief state. Experiments are carried out to demonstrate the validation of the proposed method.

Best Buys and Own Brands: Investment Platforms’ Recommendations of Mutual Funds

Abstract Individuals increasingly buy mutual funds via online platforms, whose “best-buy” recommendations heavily influence flows. As intermediaries of mutual funds, platforms provide none of the unobservable interaction or intangible benefits of brokers, and so allow clean tests of the determinants, influence, and value of their fund recommendations. Using unique U.K. data, we find that platforms favor “own-brand” funds and those paying them a higher commission share. Investors discount own-brand recommendations, but not those paying high commission shares (which were not observable in the United Kingdom). A regulatory ban on commission sharing lowered costs and improved the informativeness of platform recommendations.

Dynamics of uS19 C-Terminal Tail during the Translation Elongation Cycle in Human Ribosomes

Ribosomes undergo multiple conformational transitions during translation elongation. Here, we report the high-resolution cryoelectron microscopy (cryo-EM) structure of the human 80S ribosome in the post-decoding pre-translocation state (classical-PRE) at 3.3-Å resolution along with the rotated (hybrid-PRE) and the post-translocation states (POST). The classical-PRE state ribosome structure reveals a previously unobserved interaction between the C-terminal region of the conserved ribosomal protein uS19 and the A- and P-site tRNAs and the mRNA in the decoding site. In addition to changes in the inter-subunit bridges, analysis of different ribosomal conformations reveals the dynamic nature of this domain and suggests a role in tRNA accommodation and translocation during elongation. Furthermore, we show that disease-associated mutations in uS19 result in increased frameshifting. Together, this structure-function analysis provides mechanistic insights into the role of the uS19 C-terminal tail in the context of mammalian ribosomes.

Economic Loss Analysis of Fishing Boat Collisions Considering Spatial-Temporal Interaction Effects

Considering unobserved spatial-temporal interaction effects, this study proposes a Bayesian spatial-temporal interaction model for predicting economic loss from fishing boat collisions using 10-year (2004–2014) collision records from six different areas in the waters of Fujian, China. Results show strong spatial heterogeneity and correlation effects in fishing boat collisions, while the economic loss from boat collisions gradually decreases with the time trend. Collision time, collision location, visibility and the involvement of LNG/LPG/chemical-carrying ships show similar marginal effects on the economic loss for the two collision types: fishing boat collisions and collisions involving no fishing boats. Navigational status and the involvement of cargo ships exhibit much bigger effects in fishing boat collisions compared with collisions involving no fishing boat. Unlike collisions involving no fishing boat, fishing boat collisions are associated with reduced economic loss in poor weather conditions characterised by strong wind/waves because in Fujian waters additional safety measures are adopted for fishing boats in such conditions. The proposed model is useful for policymakers in adopting safety enhancement strategies to decrease the economic loss resulting from fishing boat collisions.

A hierarchical Bayesian model for predicting ecological interactions using scaled evolutionary relationships

Identifying undocumented or potential future interactions among species is a challenge facing modern ecologists. Recent link prediction methods rely on trait data; however, large species interaction databases are typically sparse and covariates are limited to only a fraction of species. On the other hand, evolutionary relationships, encoded as phylogenetic trees, can act as proxies for underlying traits and historical patterns of parasite sharing among hosts. We show that, using a network-based conditional model, phylogenetic information provides strong predictive power in a recently published global database of host-parasite interactions. By scaling the phylogeny using an evolutionary model, our method allows for biological interpretation often missing from latent variable models. To further improve on the phylogeny-only model, we combine a hierarchical Bayesian latent score framework for bipartite graphs that accounts for the number of interactions per species with host dependence informed by phylogeny. Combining the two information sources yields significant improvement in predictive accuracy over each of the submodels alone. As many interaction networks are constructed from presence-only data, we extend the model by integrating a correction mechanism for missing interactions which proves valuable in reducing uncertainty in unobserved interactions.

Old drug repositioning and new drug discovery through similarity learning from drug-target joint feature spaces

BackgroundDetection of new drug-target interactions by computational algorithms is of crucial value to both old drug repositioning and new drug discovery. Existing machine-learning methods rely only on experimentally validated drug-target interactions (i.e., positive samples) for the predictions. Their performance is severely impeded by the lack of reliable negative samples.ResultsWe propose a method to construct highly-reliable negative samples for drug target prediction by a pairwise drug-target similarity measurement and OCSVM with a high-recall constraint. On one hand, we measure the pairwise similarity between every two drug-target interactions by combining the chemical similarity between their drugs and the Gene Ontology-based similarity between their targets. Then we calculate the accumulative similarity with all known drug-target interactions for each unobserved drug-target interaction. On the other hand, we obtain the signed distance from OCSVM learned from the known interactions with high recall (≥0.95) for each unobserved drug-target interaction. After normalizing all accumulative similarities and signed distances to the range [0,1], we compute the score for each unobserved drug-target interaction via averaging its accumulative similarity and signed distance. Unobserved interactions with lower scores are preferentially served as reliable negative samples for the classification algorithms. The performance of the proposed method is evaluated on the interaction data between 1094 drugs and 1556 target proteins. Extensive comparison experiments using four classical classifiers and one domain predictive method demonstrate the superior performance of the proposed method. A better decision boundary has been learned from the constructed reliable negative samples.ConclusionsProper construction of highly-reliable negative samples can help the classification models learn a clear decision boundary which contributes to the performance improvement.

Variation at an adhesin locus suggests sociality in natural populations of the yeast Saccharomyces cerevisiae.

Microbes engage in numerous social behaviours that are critical for survival and reproduction, and that require individuals to act as a collective. Various mechanisms ensure that collectives are composed of related, cooperating cells, thus allowing for the evolution and stability of these traits, and for selection to favour traits beneficial to the collective. Since microbes are difficult to observe directly, sociality in natural populations can instead be investigated using evolutionary genetic signatures, as social loci can be evolutionary hotspots. The budding yeast has been studied for over a century, yet little is known about its social behaviour in nature. Flo11 is a highly regulated cell adhesin required for most laboratory social phenotypes; studies suggest it may function in cell recognition and its heterogeneous expression may be adaptive for collectives such as biofilms. We investigated this locus and found positive selection in the areas implicated in cell–cell interaction, suggesting selection for kin discrimination. We also found balancing selection at an upstream activation site, suggesting selection on the level of variegated gene expression. Our results suggest this model yeast is surprisingly social in natural environments and is probably engaging in various forms of sociality. By using genomic data, this research provides a glimpse of otherwise unobservable interactions.

Factoring in the spatial effects of symbolic number representation

Spatial constituents of adult symbolic number representation produce effects of size-value congruity, Spatial Numerical Association of Response Codes (SNARC), and numerical distance. According to behavioral experiments, these effects belong to distinct processing stages. Yet, these effects evoke overlapping responses in both early and late Event Related Potentials (ERPs). To probe whether these overlaps indicate sharing of resources, all relevant stimulus and response conditions were factorially combined in a numerical value comparison task. To secure ERP validity, same numbers were compared against variable reference values. This design resulted in previously unobserved interactions in behavior but inhibited late ERP effects. All effects arose early in the P1 component (around 100 ms) and most showed hemispheric specificity. Independency of congruity and SNARC effects was observed, whereas SNARC and numerical distance were closely intertwined. Differences in hemispheric specificity, rather than stage-wise separation, were key to independence.

SFLLN: A sparse feature learning ensemble method with linear neighborhood regularization for predicting drug–drug interactions

Drug–drug interactions are one of the major concerns of drug discovery, and the accurate prediction of drug–drug interactions is important for drug safety surveillance. However, most existing methods are suitable for predicting unobserved interactions between drugs with known interactions, but not suitable for those without known interactions. In this paper, we propose a sparse feature learning ensemble method with linear neighborhood regularization, abbreviated as SFLLN, to predict drug–drug interactions. First, we integrate four drug features, chemical substructures, targets, enzymes and pathways, by mapping drugs in different feature spaces into the common interaction space through sparse feature learning. Then, we introduce the linear neighborhood regularization to describe drug–drug relations in the interaction space by using known drug–drug interactions. In computational experiments, SFLLN produces high accuracy and outperforms benchmark methods. We present an iterative algorithm to solve the objective function of SFLLN, and analyze the convergence of the algorithm. The experiment shows that SFLLN has high efficiency and costs a reasonable amount of running time. Moreover, the case studies demonstrate that the proposed method can find novel interactions, which are validated by public evidence. In conclusion, SFLLN has great potential for predicting drug–drug interactions.