Limits...
Gene network inference by fusing data from diverse distributions.

Žitnik M, Zupan B - Bioinformatics (2015)

Bottom Line: In a simulation study, we demonstrate good predictive performance of FuseNet in comparison to several popular graphical models.Fusion of datasets offers substantial gains relative to inference of separate networks for each dataset.Our results demonstrate that network inference methods for non-Gaussian data can help in accurate modeling of the data generated by emergent high-throughput technologies.

View Article: PubMed Central - PubMed

Affiliation: Faculty of Computer and Information Science, University of Ljubljana, Ljubljana, Slovenia and Department of Molecular and Human Genetics, Baylor College of Medicine, Houston, TX, USA.

Show MeSH

Related in: MedlinePlus

An overview of FuseNet in a toy application to network inference. FuseNet’s input is a collection of datasets that can follow different exponential family distributions. The example from the figure uses two datasets: (a) gene expressions from next-generation sequencing follow the Poisson distribution, and (b) somatic mutation data follow the multinomial distribution. (c) FuseNet infers a network by collectively modeling dependencies between any two genes conditioned on the rest of the genes. The absence of an edge between s2 and s3 (dotted line in grey) implies that s2 acts independently of s3 given s1 and s4, the neighbors of s2. The  symbol stands for conditional independence. Genes s1 and s2 are linked because data profiles of s2 in (a, b) are still predictive of the profile values of s1 given s4, the neighbor of s2. (d) Shown are FuseNet-inferred coefficients that relate s2 to all other genes. Nonzero values indicate gene dependency. In the resulting network, gene s2 has two neighbors, s1 and s4
© Copyright Policy - creative-commons
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4542780&req=5

btv258-F1: An overview of FuseNet in a toy application to network inference. FuseNet’s input is a collection of datasets that can follow different exponential family distributions. The example from the figure uses two datasets: (a) gene expressions from next-generation sequencing follow the Poisson distribution, and (b) somatic mutation data follow the multinomial distribution. (c) FuseNet infers a network by collectively modeling dependencies between any two genes conditioned on the rest of the genes. The absence of an edge between s2 and s3 (dotted line in grey) implies that s2 acts independently of s3 given s1 and s4, the neighbors of s2. The symbol stands for conditional independence. Genes s1 and s2 are linked because data profiles of s2 in (a, b) are still predictive of the profile values of s1 given s4, the neighbor of s2. (d) Shown are FuseNet-inferred coefficients that relate s2 to all other genes. Nonzero values indicate gene dependency. In the resulting network, gene s2 has two neighbors, s1 and s4

Mentions: We have developed a novel approach, called FuseNet, for inference of undirected networks from a number of high-dimensional datasets (Fig. 1). Our approach builds upon recent theoretical results about Markov networks (Yang et al., 2012, 2013) and, unlike the previous works in Markov modeling, can be applied to settings where data arise from multiple related but otherwise nonidentical distributions. To achieve this level of modeling flexibility, we represent model parameters with latent factors. FuseNet implements data fusion through sharing of latent factors that are common to all datasets and distributions, and handles data diversity through inference of factors specific to a particular dataset.Fig. 1.


Gene network inference by fusing data from diverse distributions.

Žitnik M, Zupan B - Bioinformatics (2015)

An overview of FuseNet in a toy application to network inference. FuseNet’s input is a collection of datasets that can follow different exponential family distributions. The example from the figure uses two datasets: (a) gene expressions from next-generation sequencing follow the Poisson distribution, and (b) somatic mutation data follow the multinomial distribution. (c) FuseNet infers a network by collectively modeling dependencies between any two genes conditioned on the rest of the genes. The absence of an edge between s2 and s3 (dotted line in grey) implies that s2 acts independently of s3 given s1 and s4, the neighbors of s2. The  symbol stands for conditional independence. Genes s1 and s2 are linked because data profiles of s2 in (a, b) are still predictive of the profile values of s1 given s4, the neighbor of s2. (d) Shown are FuseNet-inferred coefficients that relate s2 to all other genes. Nonzero values indicate gene dependency. In the resulting network, gene s2 has two neighbors, s1 and s4
© Copyright Policy - creative-commons
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4542780&req=5

btv258-F1: An overview of FuseNet in a toy application to network inference. FuseNet’s input is a collection of datasets that can follow different exponential family distributions. The example from the figure uses two datasets: (a) gene expressions from next-generation sequencing follow the Poisson distribution, and (b) somatic mutation data follow the multinomial distribution. (c) FuseNet infers a network by collectively modeling dependencies between any two genes conditioned on the rest of the genes. The absence of an edge between s2 and s3 (dotted line in grey) implies that s2 acts independently of s3 given s1 and s4, the neighbors of s2. The symbol stands for conditional independence. Genes s1 and s2 are linked because data profiles of s2 in (a, b) are still predictive of the profile values of s1 given s4, the neighbor of s2. (d) Shown are FuseNet-inferred coefficients that relate s2 to all other genes. Nonzero values indicate gene dependency. In the resulting network, gene s2 has two neighbors, s1 and s4
Mentions: We have developed a novel approach, called FuseNet, for inference of undirected networks from a number of high-dimensional datasets (Fig. 1). Our approach builds upon recent theoretical results about Markov networks (Yang et al., 2012, 2013) and, unlike the previous works in Markov modeling, can be applied to settings where data arise from multiple related but otherwise nonidentical distributions. To achieve this level of modeling flexibility, we represent model parameters with latent factors. FuseNet implements data fusion through sharing of latent factors that are common to all datasets and distributions, and handles data diversity through inference of factors specific to a particular dataset.Fig. 1.

Bottom Line: In a simulation study, we demonstrate good predictive performance of FuseNet in comparison to several popular graphical models.Fusion of datasets offers substantial gains relative to inference of separate networks for each dataset.Our results demonstrate that network inference methods for non-Gaussian data can help in accurate modeling of the data generated by emergent high-throughput technologies.

View Article: PubMed Central - PubMed

Affiliation: Faculty of Computer and Information Science, University of Ljubljana, Ljubljana, Slovenia and Department of Molecular and Human Genetics, Baylor College of Medicine, Houston, TX, USA.

Show MeSH
Related in: MedlinePlus