ABSTRACT
Link prediction in complex networks has recently attracted a great deal of attraction in diverse scientific domains, including social and biological sciences. Given a snapshot of a network, the goal is to predict links that are missing in the network or that are likely to occur in the near future. This problem has both theoretical and practical significance; it not only helps us to identify missing links in a network more efficiently by avoiding the expensive and time consuming experimental processes, but also allows us to study the evolution of a network with time. To address the problem of link prediction, numerous attempts have been made over the recent years that exploit the local and the global topological properties of the network to predict missing links in the network. In this paper, we use parametrised matrix forest index (PMFI) to predict missing links in a network. We show that, for small parameter values, this index is linked to a heat diffusion process on a graph and therefore encodes geometric properties of the network. We then develop a framework that combines the PMFI with a local similarity index to predict missing links in the network. The framework is applied to numerous networks obtained from diverse domains such as social network, biological network, and transport network. The results show that the proposed method can predict missing links with higher accuracy when compared to other state-of-the-art link prediction methods.
ABSTRACT
[This corrects the article DOI: 10.1371/journal.pone.0263390.].
ABSTRACT
Topic models extract latent concepts from texts in the form of topics. Lifelong topic models extend topic models by learning topics continuously based on accumulated knowledge from the past which is updated continuously as new information becomes available. Hierarchical topic modeling extends topic modeling by extracting topics and organizing them into a hierarchical structure. In this study, we combine the two and introduce hierarchical lifelong topic models. Hierarchical lifelong topic models not only allow to examine the topics at different levels of granularity but also allows to continuously adjust the granularity of the topics as more information becomes available. A fundamental issue in hierarchical lifelong topic modeling is the extraction of rules that are used to preserve the hierarchical structural information among the rules and will continuously update based on new information. To address this issue, we introduce a network communities based rule mining approach for hierarchical lifelong topic models (NHLTM). The proposed approach extracts hierarchical structural information among the rules by representing textual documents as graphs and analyzing the underlying communities in the graph. Experimental results indicate improvement of the hierarchical topic structures in terms of topic coherence that increases from general to specific topics.
ABSTRACT
BACKGROUND: Numerous approaches have been proposed for the detection of epistatic interactions within GWAS datasets in order to better understand the drivers of disease and genetics. METHODS: A selection of state-of-the-art approaches were assessed. These included the statistical tests, fast-epistasis, BOOST, logistic regression and wtest; swarm intelligence methods, namely AntEpiSeeker, epiACO and CINOEDV; and data mining approaches, including MDR, GSS, SNPRuler and MPI3SNP. Data were simulated to provide randomly generated models with no individual main effects at different heritabilities (pure epistasis) as well as models based on penetrance tables with some main effects (impure epistasis). Detection of both two and three locus interactions were assessed across a total of 1,560 simulated datasets. The different methods were also applied to a section of the UK biobank cohort for Atrial Fibrillation. RESULTS: For pure, two locus interactions, PLINK's implementation of BOOST recovered the highest number of correct interactions, with 53.9% and significantly better performing than the other methods (p = 4.52e - 36). For impure two locus interactions, MDR exhibited the best performance, recovering 62.2% of the most significant impure epistatic interactions (p = 6.31e - 90 for all but one test). The assessment of three locus interaction prediction revealed that wtest recovered the highest number (17.2%) of pure epistatic interactions(p = 8.49e - 14). wtest also recovered the highest number of three locus impure epistatic interactions (p = 6.76e - 48) while AntEpiSeeker ranked as the most significant the highest number of such interactions (40.5%). Finally, when applied to a real dataset for Atrial Fibrillation, most notably finding an interaction between SYNE2 and DTNB.
Subject(s)
Atrial Fibrillation/genetics , Epistasis, Genetic , Genetic Loci , Models, Genetic , Penetrance , Algorithms , Alleles , Data Mining/methods , Dystrophin-Associated Proteins/genetics , Gene Frequency , Genome-Wide Association Study/methods , Genotype , Humans , Linear Models , Microfilament Proteins/genetics , Multifactor Dimensionality Reduction , Nerve Tissue Proteins/genetics , Neuropeptides/genetics , Polymorphism, Single Nucleotide , ROC CurveABSTRACT
BACKGROUND: Mortality remains unacceptably high in patients with heart failure and reduced left ventricular ejection fraction (LVEF) despite advances in therapeutics. We hypothesised that a novel artificial intelligence approach could better assess multiple and higher-dimension interactions of comorbidities, and define clusters of ß-blocker efficacy in patients with sinus rhythm and atrial fibrillation. METHODS: Neural network-based variational autoencoders and hierarchical clustering were applied to pooled individual patient data from nine double-blind, randomised, placebo-controlled trials of ß blockers. All-cause mortality during median 1·3 years of follow-up was assessed by intention to treat, stratified by electrocardiographic heart rhythm. The number of clusters and dimensions was determined objectively, with results validated using a leave-one-trial-out approach. This study was prospectively registered with ClinicalTrials.gov (NCT00832442) and the PROSPERO database of systematic reviews (CRD42014010012). FINDINGS: 15â659 patients with heart failure and LVEF of less than 50% were included, with median age 65 years (IQR 56-72) and LVEF 27% (IQR 21-33). 3708 (24%) patients were women. In sinus rhythm (n=12â822), most clusters demonstrated a consistent overall mortality benefit from ß blockers, with odds ratios (ORs) ranging from 0·54 to 0·74. One cluster in sinus rhythm of older patients with less severe symptoms showed no significant efficacy (OR 0·86, 95% CI 0·67-1·10; p=0·22). In atrial fibrillation (n=2837), four of five clusters were consistent with the overall neutral effect of ß blockers versus placebo (OR 0·92, 0·77-1·10; p=0·37). One cluster of younger atrial fibrillation patients at lower mortality risk but similar LVEF to average had a statistically significant reduction in mortality with ß blockers (OR 0·57, 0·35-0·93; p=0·023). The robustness and consistency of clustering was confirmed for all models (p<0·0001 vs random), and cluster membership was externally validated across the nine independent trials. INTERPRETATION: An artificial intelligence-based clustering approach was able to distinguish prognostic response from ß blockers in patients with heart failure and reduced LVEF. This included patients in sinus rhythm with suboptimal efficacy, as well as a cluster of patients with atrial fibrillation where ß blockers did reduce mortality. FUNDING: Medical Research Council, UK, and EU/EFPIA Innovative Medicines Initiative BigData@Heart.
Subject(s)
Adrenergic beta-Antagonists/therapeutic use , Atrial Fibrillation/drug therapy , Cluster Analysis , Heart Failure/drug therapy , Machine Learning , Aged , Comorbidity , Double-Blind Method , Female , Heart Failure/mortality , Humans , Male , Middle Aged , Stroke Volume , Ventricular Function, LeftABSTRACT
Multimorbidity, frequently associated with aging, can be operationally defined as the presence of two or more chronic conditions. Predicting the likelihood of a patient with multimorbidity to develop a further particular disease in the future is one of the key challenges in multimorbidity research. In this paper we are using a network-based approach to analyze multimorbidity data and develop methods for predicting diseases that a patient is likely to develop. The multimorbidity data is represented using a temporal bipartite network whose nodes represent patients and diseases and a link between these nodes indicates that the patient has been diagnosed with the disease. Disease prediction then is reduced to a problem of predicting those missing links in the network that are likely to appear in the future. We develop a novel link prediction method for static bipartite network and validate the performance of the method on benchmark datasets. By using a probabilistic framework, we then report on the development of a method for predicting future links in the network, where links are labelled with a time-stamp. We apply the proposed method to three different multimorbidity datasets and report its performance measured by different performance metrics including AUC, Precision, Recall, and F-Score.
Subject(s)
Chronic Disease/trends , Multimorbidity/trends , Forecasting/methods , Humans , ProbabilityABSTRACT
Observational and experimental evidence has linked chronotype to both psychological and cardiometabolic traits. Recent Mendelian randomization (MR) studies have investigated direct links between chronotype and several of these traits, often in isolation of outside potential mediating or moderating traits. We mined the EpiGraphDB MR database for calculated chronotype-trait associations (p-value < 5 × 10-8). We then re-analyzed those relevant to metabolic or mental health and investigated for statistical evidence of horizontal pleiotropy. Analyses passing multiple testing correction were then investigated for confounders, colliders, intermediates, and reverse intermediates using the EpiGraphDB database, creating multiple chronotype-trait interactions among each of the the traits studied. We revealed 10 significant chronotype-exposure associations (false discovery rate < 0.05) exposed to 111 potential previously known confounders, 52 intermediates, 18 reverse intermediates, and 31 colliders. Chronotype-lipid causal associations collided with treatment and diabetes effects; chronotype-bipolar associations were mediated by breast cancer; and chronotype-alcohol intake associations were impacted by confounders and intermediate variables including known zeitgebers and molecular traits. We have reported the influence of chronotype on several cardiometabolic and behavioural traits, and identified potential confounding variables not reported on in studies while discovering new associations to drugs and disease.
Subject(s)
Bipolar Disorder/genetics , Circadian Rhythm/genetics , Phenotype , Alcohol Drinking , Alcohols , Databases, Genetic , Humans , Mendelian Randomization Analysis , WorkflowABSTRACT
Link prediction in a complex network is a problem of fundamental interest in network science and has attracted increasing attention in recent years. It aims to predict missing (or future) links between two entities in a complex system that are not already connected. Among existing methods, local similarity indices are most popular that take into account the information of common neighbours to estimate the likelihood of existence of a connection between two nodes. In this paper, we propose global and quasi-local extensions of some commonly used local similarity indices. We have performed extensive numerical simulations on publicly available datasets from diverse domains demonstrating that the proposed extensions not only give superior performance, when compared to their respective local indices, but also outperform some of the current, state-of-the-art, local and global link-prediction methods.
ABSTRACT
Inferring the topology of a gene regulatory network (GRN) from gene expression data is a challenging but important undertaking for gaining a better understanding of gene regulation. Key challenges include working with noisy data and dealing with a higher number of genes than samples. Although a number of different methods have been proposed to infer the structure of a GRN, there are large discrepancies among the different inference algorithms they adopt, rendering their meaningful comparison challenging. In this study, we used two methods, namely the MIDER (Mutual Information Distance and Entropy Reduction) and the PLSNET (Partial least square based feature selection) methods, to infer the structure of a GRN directly from data and computationally validated our results. Both methods were applied to different gene expression datasets resulting from inflammatory bowel disease (IBD), pancreatic ductal adenocarcinoma (PDAC), and acute myeloid leukaemia (AML) studies. For each case, gene regulators were successfully identified. For example, for the case of the IBD dataset, the UGT1A family genes were identified as key regulators while upon analysing the PDAC dataset, the SULF1 and THBS2 genes were depicted. We further demonstrate that an ensemble-based approach, that combines the output of the MIDER and PLSNET algorithms, can infer the structure of a GRN from data with higher accuracy. We have also estimated the number of the samples required for potential future validation studies. Here, we presented our proposed analysis framework that caters not only to candidate regulator genes prediction for potential validation experiments but also an estimation of the number of samples required for these experiments.
Subject(s)
Carcinoma, Pancreatic Ductal/genetics , Computational Biology/methods , Gene Regulatory Networks , Inflammatory Bowel Diseases/genetics , Leukemia, Myeloid, Acute/genetics , Pancreatic Neoplasms/genetics , Algorithms , Gene Expression Profiling/methods , Gene Expression Regulation , Genetic Markers , Glucuronosyltransferase/genetics , Humans , Least-Squares Analysis , Sulfotransferases/genetics , Thrombospondins/geneticsABSTRACT
Reversible Data Hiding (RDH) techniques have gained popularity over the last two decades, where data is embedded in an image in such a way that the original image can be restored. Earlier works on RDH was based on the Image Histogram Modification that uses the peak point to embed data in the image. More recent works focus on the Difference Image Histogram Modification that exploits the fact that the neighbouring pixels of an image are highly correlated and therefore the difference of image makes more space to embed large amount of data. In this paper we propose a framework to increase the embedding capacity of reversible data hiding techniques that use a difference of image to embed data. The main idea is that, instead of taking the difference of the neighboring pixels, we rearrange the columns (or rows) of the image in a way that enhances the smooth regions of an image. Any difference based technique to embed data can then be used in the transformed image. The proposed method is applied on different types of images including textures, patterns and publicly available images. Experimental results demonstrate that the proposed method not only increases the message embedding capacity of a given image by more than 50% but also the visual quality of the marked image containing the message is more than the visual quality obtained by existing state-of-the-art reversible data hiding technique. The proposed technique is also verified by Pixel Difference Histogram (PDH) Stegoanalysis and results demonstrate that marked images generated by proposed method is undetectable by PDH analysis.
Subject(s)
Algorithms , Computer Security/standards , Computer Simulation , Databases, Factual , Histological Techniques/methods , Image Processing, Computer-Assisted/methods , Humans , RecordsABSTRACT
Lifelong machine learning (LML) models learn with experience maintaining a knowledge-base, without user intervention. Unlike traditional single-domain models they can easily scale up to explore big data. The existing LML models have high data dependency, consume more resources, and do not support streaming data. This paper proposes online LML model (OAMC) to support streaming data with reduced data dependency. With engineering the knowledge-base and introducing new knowledge features the learning pattern of the model is improved for data arriving in pieces. OAMC improves accuracy as topic coherence by 7% for streaming data while reducing the processing cost to half.
Subject(s)
Databases as Topic , Information Storage and Retrieval , Knowledge , Machine Learning , Online Systems , Humans , Learning , Models, TheoreticalABSTRACT
The aim of this paper is to explore the use of backtrackless walks and prime cycles for characterizing both labeled and unlabeled graphs. The reason for using backtrackless walks and prime cycles is that they avoid tottering, and can increase the discriminative power of the resulting graph representation. However, the use of such methods is limited in practice because of their computational cost. In this paper, we present efficient methods for computing graph kernels, which are based on backtrackless walks in a labeled graph and whose worst case running time is the same as that of kernels based on random walks. For clustering unlabeled graphs, we construct feature vectors using Ihara coefficients, since these coefficients are related to the frequencies of prime cycles in the graph. To efficiently compute the low order coefficients, we present an O(|V|(3)) algorithm which is better than the O(|V|(6)) worst case running time of previously known algorithms. In the experimental evaluation, we apply the proposed method to clustering both labeled and unlabeled graphs. The results show that using backtrackless walks and prime cycles instead of random walks can increase the accuracy of recognition.