Functional Genomics and Systems Biology Group and at IBM Gus - PowerPoint PPT Presentation

IBM Computational Biology Center www.research.ibm.com/FunGen IBM Research Functional Genomics and Systems Biology Group and at IBM Gus Stolovitzky Jorge Lepre Accomplices Rich Mushlin Gyan Bhanot Jeremy Rice Yuhai Tu: Phys. Sci Keith Duggar Glenn Held: Phys. Sci. Lan Ma John Wagner Aaron Kershenbaum Computational Biology Center Thomas J. Watson Research Center gustavo@us.ibm.com

IBM Computational Biology Center Start from a known Network Topology Original E-coli Network � 518 actual connections - 423 nodes

IBM Computational Biology Center Simulate a dynamic behavior simulated dynamics using Original E-coli Network known topology � 518 actual connections - 423 nodes Produce a simulated gene expression Dataset: Exp 1 Exp 2 …. Exp N u 12 u 11 u 1N Gene 1 u 21 u 22 u 2N Gene 2 …. Gene 423

IBM Computational Biology Center Reverse Engineer this Exp 1 Exp 2 …. Exp N u 12 u 11 u 1N Gene 1 u 21 u 22 u 2N Gene 2 …. Gene 423 Reconstructed Network Using your favorite algorithm, reconstructed original network from gene expression data

IBM Computational Biology Center Use some metrics to compare inferred to original Original E-coli Network � 518 actual connections - 423 nodes � 495 connections correctly predicted � 85 connections wrongly predicted Reconstructed Network Rice, Tu and Stolovitzky , “Reconstructing synthetic biological network”, Bioinformatics, 21(6):765-73 (2005)

IBM Computational Biology Center Inference of Biological Networks Network topology Original E-coli Network + synthetic dynamics � 518 actual connections - 423 nodes + protocols representing actual � 495 connections correctly predicted � 85 connections wrongly predicted experimental assays Reconstructed Network + conditional correlation algorithms (blind to original network) = Reconstructed network Rice, Tu and Stolovitzky , “Reconstructing synthetic biological network”, Bioinformatics, 21(6):765-73 (2005) Rice and Stolovitzky, Making the most of it: Pathway reconstruction and integrative simulation using the data at hand, Biosilico 2(2):70-7 (2004). Basso, Margolin, Nemenman, Klein, Wiggins, Stolovitzky, Dalla Favera, and Califano , Reverse engineering of regulatory networks in human B cells, 37(4):382-90 (2005).

IBM Computational Biology Center Standardized Datasets for Tool Development DREAM: A Dialogue on Reverse Critical Assessment of Engineering Assessment and Techniques for Protein Methods Structure Prediction (CASP) GSMLISHSDMNQQLKSAGIGFNATELHGFLSGLLCGGLKDQSWLPLLYQFSNDNHA YPTGLVQPVTELYEQISQTLSDVEGFTFELGLTEDENVFTQADSLSDWANQFLLGIG LAQPELAKEKGEIGEAVDDLQDICQLGYDEDDNEEELAEALEEIIEYVRTIAMLFYS HFNEGEIESKPVLH Columbia University (Andrea Califano) & IBM Computational Biology Center (G. Stolovitzky) http://www.nyas.org/dream2

IBM Computational Biology Center Motif Discovery in Biological Networks E. coli regulatory network • Biological Networks have an architecture yet to be understood… External source - 37 nodes Internal source - 29 nodes Intermediary - 21 nodes Internal sink – 94 nodes External sink - 208 nodes • …and functional modules. We designed algorithms for discovery of network motifs using sub-graph isomorphism algorithms. hns Squares Triangles 4 other target In E. coli , some functional genes flhDS modules are composed out of flgB-K fliG-K fliE fliL-R flhBAE … smaller motifs, such as in the Motifs motABcheAW flagella formation pathway. fliC flgAMN flgKL Combination of Motifs fliDST flgNM moaA-E tsr tar Rice, Kershenbaum and Stolovitzky . Analyzing and reconstructing gene regulatory networks. “Specialist review”, The Encyclopedia of Genetics, Genomics, Proteomics and Bioinformatics, John Wiley & Sons, Ltd:Chichester (2005). Rice, Kershenbaum and Stolovitzky , Lasting impressions: Motifs in protein-protein maps may provide footprints of evolutionary events, Proc. Natl. Acad. Sci. USA, 102, 3173-4 (2005).

IBM Computational Biology Center Digital response of tumor suppressor p53 to IR Lahav, Rosenfeld, Sigal, Geva-Zatorsky, Levine, Elowitz, & Alon : Dynamics of the p53-Mdm2 feedback loop in individual cells. Nat Genet. 36 : 147-50 (2004) Irradiation p53 protein DNA DNA ATM: ATM: P53- P53- Cell Cycle Cell Cycle damage damage DNA DNA MDM2 MDM2 Arrest and Arrest and Irradiation Irradiation initiation initiation damage damage oscillator oscillator DNA Repair DNA Repair & repair & repair detection detection ? ?

IBM Computational Biology Center Digital response of tumor suppressor p53 to IR Response to 5Gy Molecular intensity (fold) 5 TP53 (Fast) (Fast) 4 mdm 2 p 53 p 53 T T P P 5 5 3 3 p p 53 53 Basal Basal MDM2 Delay= ? Delay= ? Protein Protein 3 D D N N A A m m R R N N A A 2 DNA A A T T M M * * (Slow) (Slow) damage 1 TP53* TP53* Protein Protein 0 Induced Induced 0 500 1000 1500 (P2) (P2) Time (min) M M D D M M 2 2 m m d d m m 2 2 m m d d m m 2 2 Basal Basal Delay= ? Delay= ? Protein Protein (P1) (P1) D D N N A A m m R R N N A A Ma, Wagner, Rice, Hu, Levine and Stolovitzky , A plausible model for the digital response of p53 to DNA damage, Proc. Natl. Acad. Sci. U S A. 102, 14266 (2005). Lahav et al., Nature Genetics Wagner; Ma; Rice; Hu; Levine; Stolovitzky , p53-Mdm2 loop 2004 controlled by a balance of its feedback strength and effective dampening using ATM and delayed feedback, IEE PROCEEDINGS SYSTEMS BIOLOGY, 152, 3, 109-118 (2005).

IBM Computational Biology Center Predictions Figures 4 and 8 from Ma, Wagner et al ., (Fast) p53 p 53 p 53 Basal Protein DNA mRNA 8 30 p53 concentration (fold) p53 concentration (fold) equilibrium equilibrium ATM* 25 oscillation oscillation 6 (Slow) p53* 20 15 Delay ? ? 4 Protein P53 -induced 10 transcription 2 5 0 0 0 1 2 3 0 2 4 6 8 Mdm2 Mdm2 Mdm2 mdm 2 basal transcription (fold) p53 basal transcription (fold) Basal Delay ? ? Protein DNA mRNA A B Figure 4 (From Ma, Wagner et.al.) - Diagram of the Figure 8 (From Ma, Wagner, et. al.) - One- p53-Mdm2 oscillator. p53 is translated from p53 mRNA dimensional bifurcation diagrams of steady-state and inactive for induction of its targets. Phosphorylated p53 versus single parameter variation of Mdm2 by ATM*, p53 becomes active (p53*), and able to basal transcription rate (A) or p53 basal transcribe (after a time delay) Mdm2 which also has a transcription rate (B). The stable equilibrium is basal transcription rate. Mdm2 protein promotes a fast represented by solid line. The lower and upper degradation of p53 and a slow degradation of p53*. In bounds of stable oscillation are represented by addition to a basal self-degradation, Mdm2 is degraded paired dotted lines. by a mechanism stimulated by ATM*.

IBM Computational Biology Center Validation of Predictions Relative induction fold of Mdm2 3 Relative induction fold of p53 5 2.5 4 2 3 1.5 2 1 1 0.5 0 0 0 2 4 6 8 10 12 0 2 4 6 8 10 12 Hu, Feng, Ma, Wagner, Rice, Stolovitzky, Levine. “A single nucleotide polymorphism in the MDM2 gene disrupts the oscillation of p53 and MDM2 levels in cells.” Cancer Res. 2007 Mar 15;67(6):2757-65.