Philip Morris International Research & Development, Quai Jeanrenaud 5, 2000 Neuchâtel, Switzerland T: +41 58 242 21 11, F: +41 58 242 28 11, W: www.philipmorrisinternational.com ISMB/ECCB 2011 Vienna, Austria 17-19 July 1. Network: Three fully curated TNF and NF-κB signaling related network models were constructed from the Selventa Knowledge Assembly based on BEL terms 3 . The nodes of the network correspond to causally related biological processes that directly regulate the expression of specific sets of genes. The networks under consideration are: • NF- B.direct : 155 genes curated out of 247 references; created to be a specific measure of NF- B activation • NF- B : 992 genes curated from 414 references; based on genes that are known to be modulated by perturbation of proteins in a model of signaling from the IkB kinase (IKK) proteins of NF-κB activation (see Figure 1) • TNF : 1741 genes curated from 589 references; based on genes that are known to be modulated by treatment of cells with TNFα Quantifying the Response of a Biological System using Network Perturbation Amplitudes Florian Martin 1 , Alain Sewer 1 , Ty Thomson 3 , Carole Mathis 1 , Vincenzo Belcastro 1 , Sam Ansari 2 , Dirk Weisensee 2 , Dexter Pratt 3 , David Drubin 3 , Julia Hoeng 1 , Manuel C. Peitsch 1 1 Philip Morris International R&D, Philip Morris Products S.A., Neuchâtel, Switzerland 2 Philip Morris International R&D, Philip Morris Research Laboratories GmbH, Cologne, Germany 3 Selventa™, One Alewife Center, Cambridge, MA, USA Introduction Materials and Methods There are various solutions to support the selection of relevant processes from integral systems biology approaches, e.g., gene expression profiling, such as measurement enrichment 1 and with gene set enrichment analysis 2 , which is more quantitative. Most of these solutions interpret data from a consequence point of view, assessing the functional impact of the changes themselves rather than from a causal point of view, identifying upstream pathways. The Selventa Knowledge Assembly 3 is a directed network of experimentally observed causal relationships between biological entities or processes (e.g., mRNAs, proteins, activities). We developed Network Perturbation Amplitude (NPA) scoring as an approach to causally assess the activity amplitude. The scoring method can be used to score any differential expression data that results from a comparison between treated and control conditions. In this poster we address three relevant performance assessments for the NPA scoring: 1. How does NPA scoring accord with individual experimental conditions, here, time-response and dose-response? 2. How does NPA scoring compare with phenotypic readouts? 3. How does NPA scoring compare across independent datasets? Results An approach to quantify network perturbation amplitude (NPA) was established based on the Selventa Knowledge Base. Two gene expression data sets of TNFα-treated NHBE cells were scored against assembled NF-κB causal network models. The results show good concordance with dose- and time-dependent responses to TNFα. When NPA scores were compared against phenotypic readouts, here NF-κB translocation, a clear monotonic relationship was seen. However, discrepancies were observed for two time points across all networks and both data sets, which were based on different behaviors in the differential gene expression. This finding underlines the importance of controlling experimental, biological, as well as using identical computational parameters when comparing NPA scores across different experiments. The NPA method provides an objective metric to quantify the global impact of external perturbations on a biological system by combining the knowledge contained in causal network models and systems response profiles such as gene expression. References Discussion and Conclusion Exp 1 TNFα concentration Control 0.1ng /ml 1ng /ml 10ng /ml 100ng /ml 1000ng /ml Duration in hours 0.5 3 3 3 3 3 3 2 3 3 3 3 3 3 4 3 3 3 3 3 3 24 3 3 3 3 3 3 Exp 2 TNFα concentration Control 10ng/ml 30ng/ml 100ng/ml Duration in hours 0.5 3 3 3 3 2 3 3 3 3 4 3 3 3 3 24 3 3 3 3 Figure 2: Treatment design for experiments 1 and 2 indicating number of biological replicates for each treatment. Green cells show comparable treatments across both experiments. Figure 5: NPA scoring results. Bars indicate NPA score for each gene expression profile. Error bars represent the 95% confidence interval. Green triangle shows increasing dose for each time point, red triangle shows increasing time. NF- B NF- B.direct TNF NF- B NF- B.direct TNF Dataset 1 Dataset 2 Comparison NF- B Comparison NF- B.direct Comparison TNF Dataset 1 Dataset 2 Dataset 1 Dataset 2 Dataset 1 Dataset 2 Figure 7: Error bars represent the 95% confidence interval. Bars indicate NPA score for each gene expression profile. Green triangle shows increasing dose for each time point, red triangle shows increasing time. Figure 1: The full NF- B network model is given (top), along with a schematic of the basic model architecture (middle). CHUK, IKBKB, and IKBKG act as inhibitors of NFKBIA, NFKBIB, and NFKBIE, which are in turn inhibitors of NFKB1, NFKB2, and RELA. The nodes used in the model are listed under each section. The nodes in bold represent nodes that have downstream gene expression measureables in the Selventa Knowledge Base, and the number of measureables is given in square brackets. The notations used in the Knowledge Base are: “CHUK P@S” for CHUK phosphorylated at serine, “CHUK P@ST” for CHUK phosphorylated at serine or threonine, “kaof(CHUK)” for the kinase activity of CHUK, “CHUK:IKBKB” for the complex of CHUK and IKBKB proteins, “IkappaB kinase complex Hs” for an aggregate of the various I B kinases in Homo sapiens (Hs), “degradationof(NFKBIA)” for the process of NFKBIA degradation, and “taof(NFKB1)” for the transcriptional activity of NFKB1. Figure 4: (A) TNFα dose-dependent induction of NF- B nuclear translocation. (B) Fluorescence intensity (FI) of NF- B in the nucleus acquired by standard image processing and statistical analysis tools per dose of TNFα. 2. Perturbation applied: Two independent experiments were performed where mRNA from NHBE cells with different doses of TNFα were collected for microarray measurement at different time points after treatment (see Figure 2) and gene expression profiles were retrieved. Activation of the stress- and immune-response transcription factor NF- B has been well-defined as a major mediator of TNFα induced signaling in a variety of systems 4,5 . Microarray experiments were performed on Affymetrix HG-U133Plus2 array. 3. Method: To score the network amplitude perturbed by TNFα, we employed the aggregated network using the individual gene response and the overall directionality of the network regulation (see Figure 3). Methods have been developed to score the aggregated network effect based on the individual gene response β i (specific contrast for dose versus control using limma 6 ), a weight factor ω i (here, ω i =(1 – FDR i ), where FDR i is the false discovery rate for β i ), and the model-based directionality of the network regulation d i : , where i is the measureable in the pooled network model nodes and N is the total number. 4. Phenotypic readout: For comparing NPA scores against phenotypic readouts, scores were computed based on the NF- B.direct model and compared to a different measure of NF- B activation, NF- B nuclear translocation available for experiment 1 (see Figure 4). Upon activation, NF- B is transported into the nucleus, where it acts to regulate the expression of many genes 7,8 . B 400 450 500 550 600 650 700 FI 0 0.1 1 10 100 0.01 TNFα (ng/ml) *** *** *** *** * Figure 3: Overview of NPA Scoring. Gene expression profiles are projected on the aggregated network model nodes. A single NPA score is then computed. Blue edges indicate causal relations, red edges indicate causal relationships between model nodes and gene expression (green = up-regulation and red = down-regulation). Figure 6: NPA score versus NF- B nuclear translocation. Score error bars represent the 95% confidence interval. Y-axis shows NPA scores, X-axis shows fluorescence intensity. Error bars in NF- B nuclear translocation represent standard deviation of the mean nuclear translocation for three different fields of view of the same population of cells. 0.5h 2h 4h 24h 2. How does NPA scoring compare with phenotypic readouts? 1. How does NPA scoring accord with individual experimental conditions, here, time-response and dose-response? 3. How does NPA scoring compare across independent datasets? [1] Salomonis, N., et al., GenMAPP 2: new features and resources for pathway analysis. BMC Bioinformatics, 2007. 8: p. 217. PMID 17588266. [2] Subramanian, A., et al., Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. Proc Natl Acad Sci U S A, 2005. 102(43): p. 15545-50. PMID 16199517. [3] Selventa. (2010). Reverse Causal Reasoning Methods Whitepaper [White paper]. Retrieved from http://www.selventa.com/technology/white-papers [4] Nelson, D.E., et al., Oscillations in NF-kappaB signaling control the dynamics of gene expression. Science, 2004. 306(5696): p. 704-8. PMID 15499023 [5] Tay, S., et al., Single-cell NF-kappaB dynamics reveal digital activation and analogue information processing. Nature, 2010. 466(7303): p. 267-71. PMID 20581820. [6] Gentleman, R., Bioinformatics and computational biology solutions using R and Bioconductor. Statistics for Biology and Health, 2005. xix(473): p. 397-420. [7] Nelson, D.E., et al., Oscillations in NF-kappaB signaling control the dynamics of gene expression. Science, 2004. 306(5696): p. 704-8. PMID 15499023. [8] Tay, S., et al., Single-cell NF-kappaB dynamics reveal digital activation and analogue information processing. Nature, 2010. 466(7303): p. 267-71. PMID 20581820. A CONTROL TNFα 100ng/ml A Abstract BACKGROUND: Holistic approaches such as gene expression profiling, complemented with prior knowledge captured in network models, provide an efficient way of identifying molecular processes systems-wide. PURPOSE: The purpose of this study was to develop an approach to quantify the global impact of a treatment applied to a biological sub-system. This approach is termed “Network Perturbation Amplitude” (NPA). METHODS: Network : Three fully curated TNF and NF-κB signaling related network models were constructed from the Selventa Knowledge Assembly. The nodes of the network correspond to causally related biological processes, some of which directly regulate the expression of specific sets of genes. Perturbation applied : Gene expression profiles were collected from two experiments where Normal Human Bronchial Epithelial (NHBE) cells were exposed to TNFα at different concentrations and times. Algorithm : To score the TNFα perturbation amplitude, all genes regulated by each network model nodes were aggregated. Specific scoring methods were then developed to score the aggregated network effect based on the individual gene response and the overall directionality of the network regulation. RESULTS: NPA scores were concordant when applied to the NF-κB network and gene expression profiles from the NHBE perturbation experiments. They all correctly indicated dose-dependent responses to TNFα. In order to validate the methods, NPA results for identical treatments (same concentration and exposure time) between two experiments were also compared. CONCLUSION: The NPA method provides an objective metric to quantify the global impact of external perturbations on a biological system by combining the knowledge contained in causal network models and systems response profiles such as gene expression. CHUK CHUK P@S176 CHUK P@S180 kaof(CHUK) [41] CHUK:IKBKB CHUK:IKBKB:IKBKG IKBKB IKBKB P@S177 IKBKB P@S181 IKBKB P@ST kaof(IKBKB) [24] IKBKG IKBKG ubiquitinated at ? IkappaB kinase complex Hs kaof(IkappaB kinase complex Hs)[58] NFKBIA [25] NFKBIA P@S NFKBIA P@S32 NFKBIA P@S36 NFKBIA ubiquitinated at K21 NFKBIA ubiquitinated at K22 degradationof(NFKBIA) NFKBIB NFKBIB P@S degradationof(NFKBIB) NFKBIE NFKBIE P@S degradationof(NFKBIE) NFKB1 [5] taof(NFKB1) [26] NFKB1:NFKB2 NFKB1:RELA RELA [48] RELA P@S276 RELA P@S311 RELA P@S536 RELA P@S529 taof(RELA) [130] NFKB Complex Hs [16] taof(NFKB Complex Hs) [795] NFKBIA NFKBIB NFKBIE CHUK IKBKB IKBKG NFKB1 NFKB2 RELA Experiment NPA Scoring NPA Score Looking at Figure 5, response to TNFα was dose- dependent for both datasets and all time-points, whereas only dataset 1 showed clear time- dependence. This holds true for all networks. Time-dependent response is not clear in dataset 2 due to a shift in the gene expression profiles. Time points 2 and 4 hours show higher contrast values then the corresponding values in dataset 1 (volcano plot results not shown). This was not the case for time points 0.5 and 24 hours where the volcano plots of dataset 1 and 2 looked comparable. Potential reasons for this observation could be: independent data normalization, different cell responsiveness, batch effect confounded with treatments, smaller dynamic range of the two experimental factors dose and time in experiment 2. NPA Score NPA Score NPA Score The NF- B.direct NPA scores for four time points plotted against NF- B nuclear translocation at 30 minutes (see Figure 6) shows, that each scoring method produced a monotonic relationship between score and nuclear translocation and the relationship was preserved at different times after TNFα treatment. These findings are consistent with the hypothesis that the NPA scores can quantify NF- B activity. This side by side comparison shows reasonable agreement between the NPA scores of two independent experiments involving the same cell lines and applied perturbations. The observed discrepancies were already visible in the differential expressions of the individual genes underlying the network and described previously (see results on Figure 5). cted. F1000 Posters. Copyr opyright protected. F1000 Posters. Copyright protected 00 Posters. Copyright protected. F1000 Posters. Copyright protected. F1000 Poste rotected. F1000 Posters. Copyright protected. F1000 Posters. Copyright protected. F1000 Posters. Copyright s. Copyright protected. F1000 Posters. Copyright protected. F1000 Posters. Copyright protected. F1000 Poste ters. Copyright protected. F1000 Posters. Copyright protected. F1000 Posters. Copyright protected ht protected. F1000 Posters. Copyright protected. F1000 Posters. Copyr 1000 Posters. Copyright protected. F1000 P . Copyright prote