Myelosuppression is one of the most common and severe adverse events associated with anti-cancer therapies and can be a source of drug attrition. Current mathematical modeling methods for assessing cytopenia risk rely on indirect measurements of drug effects and primarily focus on single lineage responses to drugs. However, anti-cancer therapies have diverse mechanisms with varying degrees of effect across hematopoietic lineages. To improve predictive understanding of drug-induced myelosuppression, we developed a quantitative systems pharmacology (QSP) model of hematopoiesis in vitro for quantifying the effects of anti-cancer agents on multiple hematopoietic cell lineages. We calibrated the system parameters of the model to cell kinetics data without treatment and then validated the model by showing that the inferred mechanisms of anti-proliferation and/or cell-killing are consistent with the published mechanisms for three classes of drugs with different mechanisms of action. Using a set of compounds as a reference set, we then analyzed novel compounds to predict their mechanisms and magnitude of myelosuppression. Further, these quantitative mechanisms are valuable for the development of translational in vivo models to predict clinical cytopenia effects.

Electronic Health Record (EHR) phenotyping utilizes patient data captured through normal medical practice, to identify features that may represent computational medical phenotypes. These features may be used to identify at-risk patients and improve prediction of patient morbidity and mortality. We present a novel deep multi-modality architecture for EHR analysis (applicable to joint analysis of multiple forms of EHR data), based on Poisson Factor Analysis (PFA) modules. Each modality, composed of observed counts, is represented as a Poisson distribution, parameterized in terms of hidden binary units. Information from different modalities is shared via a deep hierarchy of common hidden units. Activation of these binary units occurs with probability characterized as Bernoulli- Poisson link functions, instead of more traditional logistic link functions. In addition, we demonstrate that PFA modules can be adapted to discriminative modalities. To compute model parameters, we derive efficient Markov Chain Monte Carlo (MCMC) inference that scales efficiently, with significant computational gains when compared to related models based on logistic link functions. To explore the utility of these models, we apply them to a subset of patients from the Duke-Durham patient cohort. We identified a cohort of over 16,000 patients with Type 2 Diabetes Mellitus (T2DM) based on diagnosis codes and laboratory tests out of our patient population of over 240,000. Examining the common hidden units uniting the PFA modules, we identify patient features that represent medical concepts. Experiments indicate that our learned features are better able to predict mortality and morbidity than clinical features identified previously in a large-scale clinical trial.

We propose a new deep architecture for topic modeling, based on Poisson Factor Analysis (PFA) modules. The model is composed of a Poisson distribution to model observed vectors of counts, as well as a deep hierarchy of hidden binary units. Rather than using logistic functions to characterize the probability that a latent binary unit is on, we employ a Bernoulli-Poisson link, which allows PFA modules to be used repeatedly in the deep architecture. We also describe an approach to build discriminative topic models, by adapting PFA modules. We derive efficient inference via MCMC and stochastic variational methods, that scale with the number of non-zeros in the data and binary units, yielding significant efficiency, relative to models based on logistic links. Experiments on several corpora demonstrate the advantages of our model when compared to related deep models.

High-density lipoprotein (HDL) is believed to play an important role in lowering cardiovascular disease (CVD) risk by mediating the process of reverse cholesterol transport (RCT). Via RCT, excess cholesterol from peripheral tissues is carried back to the liver and hence should lead to the reduction of atherosclerotic plaques. The recent failures of HDL-cholesterol (HDL-C) raising therapies have initiated a re-examination of the link between CVD risk and the rate of RCT, and have brought into question whether all target modulations that raise HDL-C would be atheroprotective. To help address these issues, a novel in-silico model has been built to incorporate modern concepts of HDL biology, including: the geometric structure of HDL linking the core radius with the number of ApoA-I molecules on it, and the regeneration of lipid-poor ApoA-I from spherical HDL due to remodeling processes. The ODE model has been calibrated using data from the literature and validated by simulating additional experiments not used in the calibration. Using a virtual population, we show that the model provides possible explanations for a number of well-known relationships in cholesterol metabolism, including the epidemiological relationship between HDL-C and CVD risk and the correlations between some HDL-related lipoprotein markers. In particular, the model has been used to explore two HDL-C raising target modulations, Cholesteryl Ester Transfer Protein (CETP) inhibition and ATP-binding cassette transporter member 1 (ABCA1) up-regulation. It predicts that while CETP inhibition would not result in an increased RCT rate, ABCA1 up-regulation should increase both HDL-C and RCT rate. Furthermore, the model predicts the two target modulations result in distinct changes in the lipoprotein measures. Finally, the model also allows for an evaluation of two candidate biomarkers for in-vivo whole-body ABCA1 activity: the absolute concentration and the % lipid-poor ApoA-I. These findings illustrate the potential utility of the model in drug development.

With recent improvements of protocols for the assembly of transcriptional parts, synthetic biological devices can now more reliably be assembled according to a given design. The standardization of parts open up the way for in silico design tools that improve the construct and optimize devices with respect to given formal design specifications. The simplest such optimization is the selection of kinetic parameters and protein abundances such that the specified design constraints are robustly satisfied. In this work we address the problem of determining parameter values that fulfill specifications expressed in terms of a functional on the trajectories of a dynamical model. We solve this inverse problem by linearizing the forward operator that maps parameter sets to specifications, and then inverting it locally. This approach has two advantages over brute-force random sampling. First, the linearization approach allows us to map back intervals instead of points and second, every obtained value in the parameter region is satisfying the specifications by construction. The method is general and can hence be incorporated in a pipeline for the rational forward design of arbitrary devices in synthetic biology.

The SBML ODE Solver Library (SOSlib) is a programming library for symbolic and numerical analysis of chemical reaction network models encoded in the Systems Biology Markup Language (SBML). It is written in ISO C and distributed under the open source LGPL license. The package employs libSBML structures for formula representation and associated functions to construct a system of ordinary differential equations, their Jacobian matrix and other derivatives. SUNDIALS' CVODES is incorporated for numerical integration and sensitivity analysis. Preliminary benchmarking results give a rough overview on the behavior of different tools and are discussed in the Supplementary Material. The native application program interface provides fine-grained interfaces to all internal data structures, symbolic operations and numerical routines, enabling the construction of very efficient analytic applications and hybrid or multi-scale solvers with interfaces to SBML and non SBML data sources. Optional modules based on XMGrace and Graphviz allow quick inspection of structure and dynamics.