Limits...
NEMix: single-cell nested effects models for probabilistic pathway stimulation.

Siebourg-Polster J, Mudrak D, Emmenlauer M, Rämö P, Dehio C, Greber U, Fröhlich H, Beerenwinkel N - PLoS Comput. Biol. (2015)

Bottom Line: Nested effects models have been used successfully for learning subcellular networks from high-dimensional perturbation effects that result from RNA interference (RNAi) experiments.As a consequence of this cellular heterogeneity, knock-downs result in variable effects among cells and lead to weak average phenotypes on the cell population level.Using a subset of genes with known interactions, we show that the inferred NEMix network has high accuracy and outperforms the classical nested effects model without hidden pathway activity.

View Article: PubMed Central - PubMed

Affiliation: Department of Biosystems Science and Engineering, ETH Zurich, Basel, Switzerland; SIB Swiss Institute of Bioinformatics, Basel, Switzerland.

ABSTRACT
Nested effects models have been used successfully for learning subcellular networks from high-dimensional perturbation effects that result from RNA interference (RNAi) experiments. Here, we further develop the basic nested effects model using high-content single-cell imaging data from RNAi screens of cultured cells infected with human rhinovirus. RNAi screens with single-cell readouts are becoming increasingly common, and they often reveal high cell-to-cell variation. As a consequence of this cellular heterogeneity, knock-downs result in variable effects among cells and lead to weak average phenotypes on the cell population level. To address this confounding factor in network inference, we explicitly model the stimulation status of a signaling pathway in individual cells. We extend the framework of nested effects models to probabilistic combinatorial knock-downs and propose NEMix, a nested effects mixture model that accounts for unobserved pathway activation. We analyzed the identifiability of NEMix and developed a parameter inference scheme based on the Expectation Maximization algorithm. In an extensive simulation study, we show that NEMix improves learning of pathway structures over classical NEMs significantly in the presence of hidden pathway stimulation. We applied our model to single-cell imaging data from RNAi screens monitoring human rhinovirus infection, where limited infection efficiency of the assay results in uncertain pathway stimulation. Using a subset of genes with known interactions, we show that the inferred NEMix network has high accuracy and outperforms the classical nested effects model without hidden pathway activity. NEMix is implemented as part of the R/Bioconductor package 'nem' and available at www.cbg.ethz.ch/software/NEMix.

No MeSH data available.


Related in: MedlinePlus

Inferred MAPK networks on HRV infection data.Best networks of the 5 top scoring siRNAs from the MAPK pathway for HRV infection for the different compared methods are displayed. (A) shows the known KEGG pathway. (B) is the inferred NEM and (C) the sc-NEM. (D) left shows the known network with the most likely attachment of the hidden variable Z (blue) and (E) is the inferred NEMix. For all networks their performance is summarized in Table 1. Subfigure (F) summarizes robustness of the MAPK network inference. For the inferred MAPK signaling networks on the HRV infection data, we assessed robustness of the accuracy for edge recovery. Box-plots display the result of 50 bootstrap samples for the three compared methods, on the 5 gene (n = 5) and 8 gene (n = 8) network.
© Copyright Policy
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4400057&req=5

pcbi.1004078.g003: Inferred MAPK networks on HRV infection data.Best networks of the 5 top scoring siRNAs from the MAPK pathway for HRV infection for the different compared methods are displayed. (A) shows the known KEGG pathway. (B) is the inferred NEM and (C) the sc-NEM. (D) left shows the known network with the most likely attachment of the hidden variable Z (blue) and (E) is the inferred NEMix. For all networks their performance is summarized in Table 1. Subfigure (F) summarizes robustness of the MAPK network inference. For the inferred MAPK signaling networks on the HRV infection data, we assessed robustness of the accuracy for edge recovery. Box-plots display the result of 50 bootstrap samples for the three compared methods, on the 5 gene (n = 5) and 8 gene (n = 8) network.

Mentions: The known KEGG network and the inferred results for the top 5 signaling genes are displayed in Fig. 3A-E. Results for the top-8 gene network are given in S10 Fig. To assess robustness of the learned networks, we repeated the inference on 50 bootstrap samples of the original data set. Both networks show high AUC values and even better accuracy (see Table 1). As can be seen from Fig. 3F, network inference was very robust for the top-5 gene network. For the top-8 gene network, performance had a slightly higher variation. Individual plots for sensitivity and specificity are given in supplementary S11 Fig. A, B. Also the estimate of p0 shows only little variation (S11 Fig. C). In all cases, the likelihood score of the known KEGG network is much lower than for the best inferred networks, indicating that under the assumptions of our model, the data and the KEGG database do not perfectly agree. Possible reasons for this observation include our model missing to explain part of the data correctly, the KEGG database being incomplete, and inaccuracies in the data generating process. Nevertheless, the accuracy value of 0.85 for the learned NEMix outperforms all other methods. All edges contained in the learned NEMix models are of high robustness (> 80% for 5 genes, and > 70% for 8 genes). Consensus networks of the bootstrap results are shown in supplementary S12 Fig.


NEMix: single-cell nested effects models for probabilistic pathway stimulation.

Siebourg-Polster J, Mudrak D, Emmenlauer M, Rämö P, Dehio C, Greber U, Fröhlich H, Beerenwinkel N - PLoS Comput. Biol. (2015)

Inferred MAPK networks on HRV infection data.Best networks of the 5 top scoring siRNAs from the MAPK pathway for HRV infection for the different compared methods are displayed. (A) shows the known KEGG pathway. (B) is the inferred NEM and (C) the sc-NEM. (D) left shows the known network with the most likely attachment of the hidden variable Z (blue) and (E) is the inferred NEMix. For all networks their performance is summarized in Table 1. Subfigure (F) summarizes robustness of the MAPK network inference. For the inferred MAPK signaling networks on the HRV infection data, we assessed robustness of the accuracy for edge recovery. Box-plots display the result of 50 bootstrap samples for the three compared methods, on the 5 gene (n = 5) and 8 gene (n = 8) network.
© Copyright Policy
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4400057&req=5

pcbi.1004078.g003: Inferred MAPK networks on HRV infection data.Best networks of the 5 top scoring siRNAs from the MAPK pathway for HRV infection for the different compared methods are displayed. (A) shows the known KEGG pathway. (B) is the inferred NEM and (C) the sc-NEM. (D) left shows the known network with the most likely attachment of the hidden variable Z (blue) and (E) is the inferred NEMix. For all networks their performance is summarized in Table 1. Subfigure (F) summarizes robustness of the MAPK network inference. For the inferred MAPK signaling networks on the HRV infection data, we assessed robustness of the accuracy for edge recovery. Box-plots display the result of 50 bootstrap samples for the three compared methods, on the 5 gene (n = 5) and 8 gene (n = 8) network.
Mentions: The known KEGG network and the inferred results for the top 5 signaling genes are displayed in Fig. 3A-E. Results for the top-8 gene network are given in S10 Fig. To assess robustness of the learned networks, we repeated the inference on 50 bootstrap samples of the original data set. Both networks show high AUC values and even better accuracy (see Table 1). As can be seen from Fig. 3F, network inference was very robust for the top-5 gene network. For the top-8 gene network, performance had a slightly higher variation. Individual plots for sensitivity and specificity are given in supplementary S11 Fig. A, B. Also the estimate of p0 shows only little variation (S11 Fig. C). In all cases, the likelihood score of the known KEGG network is much lower than for the best inferred networks, indicating that under the assumptions of our model, the data and the KEGG database do not perfectly agree. Possible reasons for this observation include our model missing to explain part of the data correctly, the KEGG database being incomplete, and inaccuracies in the data generating process. Nevertheless, the accuracy value of 0.85 for the learned NEMix outperforms all other methods. All edges contained in the learned NEMix models are of high robustness (> 80% for 5 genes, and > 70% for 8 genes). Consensus networks of the bootstrap results are shown in supplementary S12 Fig.

Bottom Line: Nested effects models have been used successfully for learning subcellular networks from high-dimensional perturbation effects that result from RNA interference (RNAi) experiments.As a consequence of this cellular heterogeneity, knock-downs result in variable effects among cells and lead to weak average phenotypes on the cell population level.Using a subset of genes with known interactions, we show that the inferred NEMix network has high accuracy and outperforms the classical nested effects model without hidden pathway activity.

View Article: PubMed Central - PubMed

Affiliation: Department of Biosystems Science and Engineering, ETH Zurich, Basel, Switzerland; SIB Swiss Institute of Bioinformatics, Basel, Switzerland.

ABSTRACT
Nested effects models have been used successfully for learning subcellular networks from high-dimensional perturbation effects that result from RNA interference (RNAi) experiments. Here, we further develop the basic nested effects model using high-content single-cell imaging data from RNAi screens of cultured cells infected with human rhinovirus. RNAi screens with single-cell readouts are becoming increasingly common, and they often reveal high cell-to-cell variation. As a consequence of this cellular heterogeneity, knock-downs result in variable effects among cells and lead to weak average phenotypes on the cell population level. To address this confounding factor in network inference, we explicitly model the stimulation status of a signaling pathway in individual cells. We extend the framework of nested effects models to probabilistic combinatorial knock-downs and propose NEMix, a nested effects mixture model that accounts for unobserved pathway activation. We analyzed the identifiability of NEMix and developed a parameter inference scheme based on the Expectation Maximization algorithm. In an extensive simulation study, we show that NEMix improves learning of pathway structures over classical NEMs significantly in the presence of hidden pathway stimulation. We applied our model to single-cell imaging data from RNAi screens monitoring human rhinovirus infection, where limited infection efficiency of the assay results in uncertain pathway stimulation. Using a subset of genes with known interactions, we show that the inferred NEMix network has high accuracy and outperforms the classical nested effects model without hidden pathway activity. NEMix is implemented as part of the R/Bioconductor package 'nem' and available at www.cbg.ethz.ch/software/NEMix.

No MeSH data available.


Related in: MedlinePlus