Tweet
Bioinformatics. 2013 Oct 31. [Epub ahead of print]
Prediction and early diagnosis of complex diseases by edge-network.
Yu X, Li G, Chen L.
Source
Key Laboratory of Systems Biology, SIBS-Novo Nordisk Translational Research Centre for PreDiabetes, Shanghai Institutes for Biological Sciences, Chinese Academy of Sciences, Shanghai 200031, China.
Abstract
MOTIVATION:
In this paper, we develop a novel edge-based network i.e., edge-network, to detect early-signals of diseases by identifying the corresponding edge-biomarkers with their DNB score from dynamical network biomarkers. Specifically, we derive an edge-network based on the second-order statistics representation of gene expression profiles, which is able to accurately represent the stochastic dynamics of the original biological system (with Gaussian distribution assumption) by combining with the traditional node-network which is based only on the first-order statistics representation of the noisy data. In other words, we show that the stochastic network of a biological system can be described by the integration of its node-network and its edge-network in an accurate manner.
RESULTS:
By applying edge-network analysis to gene expressions of healthy adults within live influenza experiment sampling at time points before the appearance of infection symptoms, we identified the edge-biomarkers (80 edges with 22 densely connected genes) discovered in edge-networks corresponding to symptomatic adults, which were used to predict the subsequent outcomes of influenza infection. In particular, we not only correctly predict the final infection outcome of each individual at an early time point before his/her clinic symptom but also reveal the key molecules during the disease progression. The prediction accuracy achieves about 90% under the leave-one-out cross-validation. Furthermore, we demonstrate the superiority of our method on disease classification and predication by comparing with the conventional node-biomarkers. Our edge-network analysis not only opens a new way to understand pathogenesis at a network level due to the new representation for a stochastic network, but also provides a powerful tool to make the early diagnosis of diseases.
CONTACT:
lnchen@sibs.ac.cn.
PMID:
24177717
[PubMed - as supplied by publisher]
Prediction and early diagnosis of complex diseases by edge-network.
Yu X, Li G, Chen L.
Source
Key Laboratory of Systems Biology, SIBS-Novo Nordisk Translational Research Centre for PreDiabetes, Shanghai Institutes for Biological Sciences, Chinese Academy of Sciences, Shanghai 200031, China.
Abstract
MOTIVATION:
In this paper, we develop a novel edge-based network i.e., edge-network, to detect early-signals of diseases by identifying the corresponding edge-biomarkers with their DNB score from dynamical network biomarkers. Specifically, we derive an edge-network based on the second-order statistics representation of gene expression profiles, which is able to accurately represent the stochastic dynamics of the original biological system (with Gaussian distribution assumption) by combining with the traditional node-network which is based only on the first-order statistics representation of the noisy data. In other words, we show that the stochastic network of a biological system can be described by the integration of its node-network and its edge-network in an accurate manner.
RESULTS:
By applying edge-network analysis to gene expressions of healthy adults within live influenza experiment sampling at time points before the appearance of infection symptoms, we identified the edge-biomarkers (80 edges with 22 densely connected genes) discovered in edge-networks corresponding to symptomatic adults, which were used to predict the subsequent outcomes of influenza infection. In particular, we not only correctly predict the final infection outcome of each individual at an early time point before his/her clinic symptom but also reveal the key molecules during the disease progression. The prediction accuracy achieves about 90% under the leave-one-out cross-validation. Furthermore, we demonstrate the superiority of our method on disease classification and predication by comparing with the conventional node-biomarkers. Our edge-network analysis not only opens a new way to understand pathogenesis at a network level due to the new representation for a stochastic network, but also provides a powerful tool to make the early diagnosis of diseases.
CONTACT:
lnchen@sibs.ac.cn.
PMID:
24177717
[PubMed - as supplied by publisher]
