Mathematical Modelling with Multidisciplinary Applications (M3A)http://hdl.handle.net/20.500.11824/132019-08-19T22:35:35Z2019-08-19T22:35:35ZModified Hamiltonian Monte Carlo for Bayesian InferenceRadivojevic T.Akhmatskaya E.http://hdl.handle.net/20.500.11824/10012019-08-07T01:00:11Z2019-07-22T00:00:00ZModified Hamiltonian Monte Carlo for Bayesian Inference
Radivojevic T.; Akhmatskaya E.
The Hamiltonian Monte Carlo (HMC) method has been recognized as a powerful sampling tool in computational statistics. We show that performance of HMC can be significantly improved by incorporating importance sampling and an irreversible part of the dynamics into a chain. This is achieved by replacing Hamiltonians in the Metropolis test with modified Hamiltonians, and a complete momentum update with a partial momentum refreshment. We call the resulting generalized HMC importance sampler—Mix & Match Hamiltonian Monte Carlo (MMHMC). The method is irreversible by construction and further benefits from (i) the efficient algorithms for computation of modified Hamiltonians; (ii) the implicit momentum update procedure and (iii) the multi-stage splitting integrators specially derived for the methods sampling with modified Hamiltonians. MMHMC has been implemented, tested on the popular statistical models and compared in sampling efficiency with HMC, Riemann
Manifold Hamiltonian Monte Carlo, Generalized Hybrid Monte Carlo, Generalized Shadow Hybrid Monte Carlo,
Metropolis Adjusted Langevin Algorithm and Random Walk Metropolis-Hastings. To make a fair comparison, we propose a metric that accounts for correlations among samples and weights, and can be readily used for all
methods which generate such samples. The experiments reveal the superiority of MMHMC over popular sampling
techniques, especially in solving high dimensional problems.
2019-07-22T00:00:00ZOpportunities at the Intersection of Synthetic Biology, Machine Learning, and AutomationCarbonell P.Radivojević T.Garcia-Martin H.http://hdl.handle.net/20.500.11824/9982019-08-05T01:00:11Z2019-07-01T00:00:00ZOpportunities at the Intersection of Synthetic Biology, Machine Learning, and Automation
Carbonell P.; Radivojević T.; Garcia-Martin H.
Our inability to predict the behavior of biological systems severely hampers progress in bioengineering and biomedical applications. We cannot predict the effect of genotype changes on phenotype, nor extrapolate the large-scale behavior from small-scale experiments. Machine learning techniques recently reached a new level of maturity, and are capable of providing the needed predictive power without a detailed mechanistic understanding. However, they require large amounts of data to be trained. The amount and quality of data required can only be produced through a combination of synthetic biology and automation, so as to generate a large diversity of biological systems with high reproducibility. A sustained investment in the intersection of synthetic biology, machine learning, and automation will drive forward predictive biology, and produce improved machine learning algorithms.
2019-07-01T00:00:00ZA concerted systems biology analysis of phenol metabolism in Rhodococcus opacus PD630Roell G.W.Carr R.R.Campbell T.Shang Z.Henson W.R.Czajka J.J.Garcia-Martin H.Zhang F.Foston M.Dantas G.Moon T.S.Tang Y.J.http://hdl.handle.net/20.500.11824/9972019-08-05T01:00:10Z2019-06-01T00:00:00ZA concerted systems biology analysis of phenol metabolism in Rhodococcus opacus PD630
Roell G.W.; Carr R.R.; Campbell T.; Shang Z.; Henson W.R.; Czajka J.J.; Garcia-Martin H.; Zhang F.; Foston M.; Dantas G.; Moon T.S.; Tang Y.J.
Rhodococcus opacus PD630 metabolizes aromatic substrates and naturally produces branched-chain lipids, which are advantageous traits for lignin valorization. To provide insights into its lignocellulose hydrolysate utilization, we performed 13C-pathway tracing, 13C-pulse-tracing, transcriptional profiling, biomass composition analysis, and metabolite profiling in conjunction with 13C-metabolic flux analysis (13C-MFA) of phenol metabolism. We found that 1) phenol is metabolized mainly through the ortho–cleavage pathway; 2) phenol utilization requires a highly active TCA cycle; 3) NADPH is generated mainly via NADPH-dependent isocitrate dehydrogenase; 4) active cataplerotic fluxes increase plasticity in the TCA cycle; and 5) gluconeogenesis occurs partially through the reversed Entner–Doudoroff pathway (EDP). We also found that phenol-fed R. opacus PD630 generally has lower sugar phosphate concentrations (e.g., fructose 1,6-bisphosphatase) compared to metabolite pools in 13C-glucose-fed Escherichia coli (set as internal standards), while its TCA metabolites (e.g., malate, succinate, and α-ketoglutarate) accumulate intracellularly with measurable succinate secretion. In addition, we found that phenol utilization was inhibited by benzoate, while catabolite repressions by other tested carbon substrates (e.g., glucose and acetate) were absent in R. opacus PD630. Three adaptively-evolved strains display very different growth rates when fed with phenol as a sole carbon source, but they maintain a conserved flux network. These findings improve our understanding of R. opacus’ metabolism for future lignin valorization.
2019-06-01T00:00:00ZConductance-Based Refractory Density Approach for a Population of Bursting NeuronsChizhov A.Campillo F.Desroches M.Guillamon ARodrigues S.http://hdl.handle.net/20.500.11824/9932019-07-12T01:00:11Z2019-01-01T00:00:00ZConductance-Based Refractory Density Approach for a Population of Bursting Neurons
Chizhov A.; Campillo F.; Desroches M.; Guillamon A; Rodrigues S.
The conductance-based refractory density (CBRD) approach is a parsimonious mathematical-computational framework for modeling interact- ing populations of regular spiking neurons, which, however, has not been yet extended for a population of bursting neurons. The canonical CBRD method allows to describe the firing activity of a statistical ensemble of uncoupled Hodgkin-Huxley-like neurons (differentiated by noise) and has demonstrated its validity against experimental data. The present manuscript generalises the CBRD for a population of bursting neurons; however, in this pilot computational study we consider the simplest setting in which each individual neuron is governed by a piecewise linear bursting dynamics. The resulting popula- tion model makes use of slow-fast analysis, which leads to a novel method- ology that combines CBRD with the theory of multiple timescale dynamics. The main prospect is that it opens novel avenues for mathematical explo- rations, as well as, the derivation of more sophisticated population activity from Hodgkin-Huxley-like bursting neurons, which will allow to capture the activity of synchronised bursting activity in hyper-excitable brain states (e.g. onset of epilepsy).
2019-01-01T00:00:00Z