Vassilios Stathopoulos

Markov Chain Monte Carlo from Lagrangian Dynamics.

Hamiltonian Monte Carlo (HMC) improves the computational efficiency of the Metropolis–Hastings algorithm by reducing its random walk behavior. Riemannian HMC (RHMC) further improves the performance of HMC by exploiting the geometric properties of the parameter space. However, the geometric integrator used for RHMC involves implicit equations that require fixed-point iterations. In some cases, the computational overhead for solving implicit equations undermines RHMC’s benefits. In an attempt to circumvent this problem, we propose an explicit integrator that replaces the momentum variable in RHMC by velocity. We show that the resulting transformation is equivalent to transforming Riemannian Hamiltonian dynamics to Lagrangian dynamics. Experimental results suggest that our method improves RHMC’s overall computational efficiency in the cases considered. All computer programs and datasets are available online (http://www.ics.uci.edu/babaks/Site/Codes.html) to allow replication of the results reported in this article.Hamiltonian Monte Carlo (HMC) improves the computational efficiency of the Metropolis algorithm by reducing its random walk behavior. Riemannian Manifold HMC (RMHMC) further improves HMCs performance by exploiting the geometric properties of the parameter space. However, the geometric integrator used for RMHMC involves implicit equations that require costly numerical analysis (e.g., fixed-point iteration). In some cases, the computational overhead for solving implicit equations undermines RMHMCs benefits. To avoid this problem, we propose an explicit geometric integrator that replaces the momentum variable in RMHMC by velocity. We show that the resulting transformation is equivalent to transforming Riemannian Hamilton dynamics to Lagrangian dynamics. Experimental results show that our method improves RMHMCs overall computational efficiency. All computer programs and data sets are available online (http://www.ics.uci.edu/~babaks/Site/Codes.html) in order to allow replications of the results reported in this paper.

Markov chain Monte Carlo inference for Markov jump processes via the linear noise approximation.

Bayesian analysis for Markov jump processes (MJPs) is a non-trivial and challenging problem. Although exact inference is theoretically possible, it is computationally demanding, thus its applicability is limited to a small class of problems. In this paper, we describe the application of Riemann manifold Markov chain Monte Carlo (MCMC) methods using an approximation to the likelihood of the MJP that is valid when the system modelled is near its thermodynamic limit. The proposed approach is both statistically and computationally efficient whereas the convergence rate and mixing of the chains allow for fast MCMC inference. The methodology is evaluated using numerical simulations on two problems from chemical kinetics and one from systems biology.

EWS-FLI1 employs an E2F switch to drive target gene expression

Cell cycle progression is orchestrated by E2F factors. We previously reported that in ETS-driven cancers of the bone and prostate, activating E2F3 cooperates with ETS on target promoters. The mechanism of target co-regulation remained unknown. Using RNAi and time-resolved chromatin-immunoprecipitation in Ewing sarcoma we report replacement of E2F3/pRB by constitutively expressed repressive E2F4/p130 complexes on target genes upon EWS-FLI1 modulation. Using mathematical modeling we interrogated four alternative explanatory models for the observed EWS-FLI1/E2F3 cooperation based on longitudinal E2F target and regulating transcription factor expression analysis. Bayesian model selection revealed the formation of a synergistic complex between EWS-FLI1 and E2F3 as the by far most likely mechanism explaining the observed kinetics of E2F target induction. Consequently we propose that aberrant cell cycle activation in Ewing sarcoma is due to the de-repression of E2F targets as a consequence of transcriptional induction and physical recruitment of E2F3 by EWS-FLI1 replacing E2F4 on their target promoters.

Acoustic identification of Mexican bats based on taxonomic and ecological constraints on call design

1. Monitoring global biodiversity is critical for understanding responses to anthropogenic change, but biodiversity monitoring is often biased away from tropical, megadiverse areas that are experiencing more rapid environmental change. Acoustic surveys are increasingly used to monitor biodiversity change, especially for bats as they are important indicator species and most use sound to detect, localise and classify objects. However, using bat acoustic surveys for monitoring poses several challenges, particularly in mega-diverse regions. Many species lack reference recordings, some species have high call similarity or differ in call detectability, and quantitative classification tools, such as machine learning algorithms, have rarely been applied to data from these areas. 2. Here, we collate a reference call library for bat species that occur in a megadiverse country, Mexico. We use 4,685 search-phase calls from 1,378 individual sequences of 59 bat species to create automatic species identification tools generated by machine learning algorithms (Random Forest). We evaluate the improvement in species-level classification rates gained by using hierarchical classifications, reflecting either taxonomic or ecological constraints (guilds) on call design, and examine how classification rate accuracy changes at different hierarchical levels (family, genus, and guild). 3. Species-level classification of calls had a mean accuracy of 66% and the use of hierarchies improved mean species-level classification accuracy by up to 6% (species within families 72%, species within genera 71.2% and species within guilds 69.1%). Classification accuracy to family, genus and guild-level was 91.7%, 77.8% and 82.5%, respectively. 4. The bioacoustic identification tools we have developed are accurate for rapid biodiversity assessments in a megadiverse region and can also be used effectively to classify species at broader taxonomic or ecological levels. This flexibility increases their usefulness when there are incomplete species reference recordings and also offers the opportunity to characterise and track changes in bat community structure. Our results show that bat bioacoustic surveys in megadiverse countries have more potential than previously thought to monitor biodiversity changes and can be used to direct further developments of bioacoustic monitoring programs in Mexico.

Putting the Scientist in the Loop -- Accelerating Scientific Progress with Interactive Machine Learning

Technology drives advances in science. Giving scientists access to more powerful tools for collecting and understanding data enables them to both ask and answer new kinds questions that were previously beyond their reach. Of these new tools at their disposal, machine learning offers the opportunity to understand and analyze data at unprecedented scales and levels of detail. The standard machine learning pipeline consists of data labeling, feature extraction, training, and evaluation. However, without expert machine learning knowledge, it is difficult for scientists to optimally construct this pipeline to fully leverage machine learning in their work. Using ecology as a motivating example, we analyze a typical scientists data collection and processing workflow and highlight many problems facing practitioners when attempting to capitalize on advances in machine learning and pattern recognition. Understanding these shortcomings allows us to outline several novel and underexplored research directions. We end with recommendations to motivate progress in future cross-disciplinary work.

mcmc_clib–an advanced MCMC sampling package for ode models

SUMMARY We present a new C implementation of an advanced Markov chain Monte Carlo (MCMC) method for the sampling of ordinary differential equation (ode) model parameters. The software mcmc_clib uses the simplified manifold Metropolis-adjusted Langevin algorithm (SMMALA), which is locally adaptive; it uses the parameter manifolds geometry (the Fisher information) to make efficient moves. This adaptation does not diminish with MC length, which is highly advantageous compared with adaptive Metropolis techniques when the parameters have large correlations and/or posteriors substantially differ from multivariate Gaussians. The software is standalone (not a toolbox), though dependencies include the GNU scientific library and sundials libraries for ode integration and sensitivity analysis. AVAILABILITY AND IMPLEMENTATION The source code and binary files are freely available for download at http://a-kramer.github.io/mcmc_clib/. This also includes example files and data. A detailed documentation, an example model and user manual are provided with the software. CONTACT andrei.kramer@ist.uni-stuttgart.de.

Abstract 2104: Evidence for E2F/EWS-FLI1 oncoprotein synergism using systems biology

Cell cycle deregulation and enhanced proliferation of cells is one of the hallmarks of oncogenesis. In Ewing Sarcoma, a highly aggressive pediatric cancer, the chimeric transcription factor EWS-FLI1 deregulates cell cycle by targeting several cell cycle regulators including the E2F family of transcription factors. ChIP-seq studies showed a significant overlap of EWS-FLI1 and E2F3/4 binding in Ewing sarcoma cells. We show that EWS-FLI1 is able to directly activate E2F3, followed by the combinatorial binding of EWS-FLI1 and E2F3 on their target genes synergistically activating their transcription. Furthermore we propose a model where EWS-FLI1 directly exchanges repressive E2F4/p130 by E2F3/pRB thereby driving cells into enhanced cell proliferation. However, so far we were unable to experimentally demonstrate a physical interaction of EWS-FLI1 with any E2F or pocket protein. As an alternative approach to the study of functional synergy between EWS-FLI1 and E2F3, we used time resolved RNA and protein expression data as a basis for mathematical modeling of EWS-FLI1 dependent E2F target gene regulation. By Bayesian model selection, we were able to postulate the formation of a complex between EWS-FLI1 and E2F3 as the most likely explanation for the observed ETS/E2F synergy. This study therefore provides an example of how computational systems approaches can complement experimental data in the discovery of disease mechanisms. This study was supported in part by the Austrian Science Fund (FWF), [grant 22328-B09], and by the 7th framework program of the European Commission, [grant 259348] (‘ASSET’). Citation Format: Raphaela Schwentner, Theodore Papamarkou, Maximilian Kauer, Vassilios Stathopoulos, Fan Yang, Sven Bilke, Paul S. Meltzer, Mark Girolami, Heinrich Kovar. Evidence for E2F/EWS-FLI1 oncoprotein synergism using systems biology. [abstract]. In: Proceedings of the 106th Annual Meeting of the American Association for Cancer Research; 2015 Apr 18-22; Philadelphia, PA. Philadelphia (PA): AACR; Cancer Res 2015;75(15 Suppl):Abstract nr 2104. doi:10.1158/1538-7445.AM2015-2104