Survey
* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project
Data, models & computation for stochastic dynamic cellular networks in systems biology Mike West Department of Statistical Science Duke University INI Cambridge, June 24th 2008 Single cell studies - dynamic data Much intra-cellular behaviour (including gene expression) is intrinsically stochastic Cellular systems cannot be properly understood (hence predicted and controlled) unless appropriate stochastic components are incorporated into dynamic cellular network models INI Cambridge, June 24th 2008 Synthetic bacterial gene circuits “emulate” gene networks key to mammalian cell proliferation (and cancer) c.f. Studies on mammalian cells Mammalian Rb/E2f pathway: Mammalian cell development & fate network Feed-Forward (Cancer Systems Biology) Stochastic models: States=RNA levels over time Data - movies: multiple genes over time Positive Feedback Fit, assess, refine models: - evaluate cell-specific stochasticity - multiple cancer cell lines - predict network responses to interventions INI Cambridge, June 24th 2008 Synthetic circuit T7 Partial data over time on elements of yt INI Cambridge, June 24th 2008 Aspects of inference & computation Many (#cells): stochastic cell-specific effects, experimental noise Parameters (rate constants) Unobserved (latent) time series of (1,2,..) RNAs Fine time scale model: crude time scale data Imputation of uncertain state variables Model fitting, assessment, comparison Simulation-based Bayesian analysis: parameters and latent states Markov chain Monte Carlo methods for dynamic, non-linear systems Integration of time course, single cell data with “marginal” data from flow cytometry - “snapshots in time on 105+ cells INI Cambridge, June 24th 2008 Stochastic imputation of latent processes HMM: Forward filtering backward sampling (FFBS) yt+k yt Latent process xt xt+1 t t+1 xt+k-1 xt+k t+k Filtering: Sampling: Latent “missing” states imputed INI Cambridge, June 24th 2008 Mixture modelling Metropolis MCMC mixture INI Cambridge, June 24th 2008 Mixture modelling Metropolis MCMC mixture INI Cambridge, June 24th 2008 Imputed trajectories + data Posterior for parameters Information content: prior posterior INI Cambridge, June 24th 2008 Data extraction: single cell dynamic imaging Novel hybrid-image-based segmentation algorithms & neighborhood-based cell tracking • E-coli • Budding yeast • Mammalian cells Open source software Cell lineage reconstruction INI Cambridge, June 24th 2008 People, papers, software etc Jarad Niemi Quanli Wang www.stat.duke.edu/~mw NSF-NIH Duke (NCI) Systems Cancer Biology Center Statistical Science Lingchong You Chee-Meng Tan Bioengineering NIH Duke (NIH) Systems Biology Center INI Cambridge, June 24th 2008 Stochastic imputation of latent processes INI Cambridge, June 24th 2008 Raw single cell data – snapshot images Frame 11 Frame 17 Frame 26 10 mins between frames - technical limit of time resolution INI Cambridge, June 24th 2008