eSSENCE Workshop on Frontiers in Computational Biomedicine

The enormous advances in structural and biochemical studies of biological systems challenges one to develop simulations approaches that can correlate the structure and function of biological molecules. The development of such approaches will be reviewed in this lecture and related to current and further advances. We will review the developments in the field in terms of three main directions, structure, dynamics and function. The review will start with the development of molecular mechanics computer programs and their application to structure refinement. We will then consider the move from energy minimization the molecular dynamics (MD), pointing out that most early MD studies did not explored truly functional problems , focusing on trying to reproduce observables which are no related to biological problems. We will then move to studies that focused on protein functions. This part of the lecture will consider the development of combined quantum mechanical molecular mechanics (QM/MM) approaches, the introduction of free energy perturbation approaches, the developments of effective electrostatic models and other advances major. After presenting a perspective of the early developments and subsequent advances will consider the current state of the field and the need for a balance between brute force use of computer power and the requirement of obtaining clear descriptions of functional properties with the ability to analyze conveniently the meaning of the calculated results. Finally we will provide a perspective of the future direction of the field.

Serum paraoxonase 1 (PON1) is a calcium-dependent lipo-lactonase that promiscuously catalyzes the hydrolysis of various organophosphate pesticides and nerve agents. PON1 interacts with high-density lipoprotein (HDL) affecting both PON1’s stability and activity. However, the role of PON1-HDL interactions in regulating PON1’s activity remains unclear.
We present here a combination of kinetic, crystallographic and computational work (using the empirical valence bond approach) of the PON1 catalyzed hydrolysis of phosphotriester and lactone substrates. We provide a detailed structural and mutational analysis of the native lactonase activity of the enzyme, and for the promiscuous phosphatase activity using paraoxon as a model substrate. Our results provide a detailed model for the activation of PON1 by HDL and discuss the role of flexibility on PON 1’s catalytic activity.

The speed of chemical reactions in water and in enzymes varies with temperature, depending on how the free energy of activation is partioned into enthalpy and entropy. In enzymes, this partioning is also optimized as a consequence of the organism’s adaptation to the environment. We will show how the temperature dependence of chemical reaction rates can be obtained from brute force computer simulations. Such calculations shed new light on how protein structures have evolved in differently adapted species.

Our nervous system is a remarkable machinery. In a fraction of a second, signals are sent e.g. from our brain to our fingers through a long sequence of nerve cells, and mistakes almost never occur. This signaling between nerve cells is controlled by ligand-gated ion channels that either create or inhibit signals in a new cell based on the release of neurotransmitter substances from the originating cell. However, the location between cells exposes these channels to other molecules, and this is how a wide range of drugs such as alcohols or anesthetics influence our nervous system. These molecules work as allosteric modulators and alter the behavior of the neurotransmitters, but the response is extremely complex with modulators either potentiating or inhibiting single channels. It is a fascinating topic, since the biological effects cannot be explained by a single structure, but they are due to the conformational transitions and alterations in the equilibrium between different states of the macromolecules. I will describe how we are using electrophysiology in combination with molecular dynamics simulations and free energy calculations to explain the molecular effects of allosteric modulation in ligand-gated ion channels. In particular, we have used computational techniques to identify separate sites responsible for channel opening and closing, we have been able to design specific mutations that even invert the modulation behavior – and we have confirmed these predictions to be correct with electrophysiology. Simulations and free energy predictions are becoming critical tools not merely for understanding complex biological macromolecules, but also to actively design new properties and interactions.

In mammals, the Suprachiasmatic Nucleus (SCN), a brain region of about 20,000 neurons, serves as the master circadian clock, coordinating timing throughout the body and entraining the body to daily light cycles. The extent to which cells in the SCN can synchronize and entrain depends on the communication network between individual cell oscillators. Characterization of that network is challenging, due to the dynamics of the circadian oscillators and the stochastic noise inherent in discrete molecular chemical reactions. Statistical models based on information theoretic measures are well-suited for the analysis of stochastic information flow across networks. We have developed a methodology that uses information-theoretic measures on time course data to infer network structure, and tested its performance on data from computational models of networks of stochastic circadian oscillators with known connectivity. We then applied the method to experimental data, to infer the functional network for synchronization in mouse SCN slices. We will discuss the properties of the inferred net works.

Our recent progress in single molecule tracking makes it possible to follow individual biomolecules in living cells at a spatial precision of 20nm and a temporal resolution of a few milliseconds. It is possible to obtain thousands of short trajectories in a few five minute experiments. The trajectories contain a wealth of information about life at the molecular level, i.e. where molecules interact, for how long and how fast they move, but novel efficient computational methods are needed to extract this information. I will present vbSPT, variational Bayesian Single Particle Tracking, as a software that potentially overcomes these challenges. On problem of making inference about in vivo reaction kinetics is however that there is no ground truth information to test the methods against. To overcome this problem, we have extended our software MesoRD for stochastic reaction diffusion simulations, to also simulate the photophysics of the single particle tracking experiment. This includes the blinking and bleaching kinetics of flourophores, the microscope’s 3D point spread function, the detector noise characteristics and the non-uniform cell background. Given simulated ground truth data we can identify errors in the particle localization, trajectory construction and state inference and develop better algorithms to overcome these limitations.

In molecular systems biology, quantitative models of reaction-diffusion kinetics have emerged as an important tool to study cellular control systems. The most commonly used models are the microscale hard sphere Smoluchowski diffusion-limited reaction model and the Reaction-Diffusion Master Equation, a mesoscale discrete stochastic model. In this talk I will discuss some recent developments on the relationship between the two modeling levels. In particular, I will show how multiscale reaction rates can be derived that leads to better agreement between the two models compared to the classical theory following Collins-Kimball. I will also show how efficient and accurate simulations can be conducted by blending the modeling levels in hybrid methods.