Search

Parameterized probabilistic complex computational (P 2 C 2 ) models are being increasingly used in computational systems biology for analyzing biological systems. A key challenge is to build mechanistic P 2 C 2 models by combining prior knowledge and empirical data, given that certain system properties are unknown. These unknown components are incorporated into a model as parameters and determining their values has traditionally been a process of trial and error. We present a new algorithmic procedure for discovering parameters in agent-based models of biological systems against behavioral specifications mined from large data-sets. Our approach uses Bayesian model checking, sequential hypothesis testing, and stochastic optimization to synthesize parameters of P 2 C 2 models. We demonstrate our algorithm by discovering the amount and schedule of doses of bacterial lipopolysaccharide in a clinical agent-based model of the dynamics of acute inflammation that guarantee a set of desired clinical outcomes with high probability.

Stochastic Differential Equation (SDE) models are often used to model the dynamics of complex biological systems. The stochastic nature of these models means that some behaviors are more likely than others. It is often the case that a model's primary purpose is to study rare but interesting or important behaviors, such as the formation of a tumor, or the failure of a cyber-physical system. Unfortunately, due to the limited availability of analytic methods for SDEs, stochastic simulations are the most common means for estimating (or bounding) the probability of rare behaviors. Naturally, the cost of stochastic simulations increases with the rarity of the behavior under consideration. To address this problem, we introduce a new algorithm, RESERCHE, that is specifically designed to quantify the likelihood of rare but interesting behaviors in SDE models. Our approach relies on the use of temporal logics for specifying rare behaviors of possible interest, and on the ability of bit-vector decision procedures to reason exhaustively about fixed precision arithmetic. We also compute the probability of an observed behavior under the assumption of Gaussian noise.

Stochastic models are often used to study the behavior of biochemical systems and biomedical devices. While the structure of such models is often readily available from first principles, several quantitative features of the model are not easily determined. These quantitative features are often incorporated into the model as parameters. The algorithmic discovery of parameter values from experimentally observed facts (including extreme-scale data) remains a challenge for the computational systems biology community. In this paper, we present a new parameter discovery algorithm based on Wald's sequential probability ratio test (SPRT). Our algorithm uses a combination of simulated annealing and sequential hypothesis testing to reduce the number of samples required for parameter discovery of stochastic models. We use probabilistic bounded linear temporal logic (PBLTL) to express the desired behavioral specification of a model. We also present theoretical results on the correctness of our algorithm, and demonstrate the effectiveness of our algorithm by studying a detailed model of glucose and insulin metabolism.

The development of mechanistic computational models of biological systems continues to be an iterative manual process that requires considerable insight and patience from the modeler. An abundance of high throughput experimental methods has caused a rapid explosion of factual information about the behavior of biochemical systems. It is unreasonable to assume that a human can continue to use his limited intellectual faculties to explore the space of all possible mechanistic models and suggest biochemical models that affirm all the observed behaviors.

We introduce a method for solving the following problem: given a Stochastic Differential Equation (SDE) model describing the dynamics of a biological system, algorithmically decide whether the model satisfies a given high-level behavioral specification. Our proposed solution uses a combination of Bayesian statistical hypothesis testing, Girsanov's theorem for change of measures, and independent but non-identically distributed (i.i.d.) sampling algorithms. Our use of non-i.i.d. sampling contributes to the state of the art in statistical verification of stochastic systems by providing an effective means for exposing rare events, while retaining the ability to compute bounds on the probability that those events occur.

Computer science and biology have inspired each other by drawing analogies. Biological molecules are even used directly as new material building blocks for creating next generation computer hard- and software. Recently, biology is moving to larger scales, generating datasets of high complexity and size. This allows moving from studies of individual biological entities such as proteins to their complex networks of interactions. Since cells are able to respond to perturbations over time, it is becoming an important variable in these studies. This is analogous to systems of systems engineering to operate in dynamic environments, requiring the efficient monitoring of external variables and intelligent choice of adequate responses. Here, we survey previous successes in exploiting analogies between biology and computer science and propose novel avenues and potential for cross-fertilization of research in dynamic systems modeling in biology and systems-of-systems engineering.