Multiscale modeling of biomolecular systems: Computer modeling is becoming increasingly valuable for understanding protein function and ligand-receptor interactions. Atomistic molecular dynamics, coarse-grained Brownian dynamics and continuum simulations are combined to study detailed molecular interactions, large scale protein motions, bio-molecular dynamics, and complex biochemical network. We continue developing new methods, as well as applying them to various problems, e.g. mechanics of coupled binding and folding of disordered proteins (see Figure 1), protein conformational changes in allosteric binding, and functions of signaling and multifunctional protein complexes.
Figure 1: Coupled folding and binding: a complex encounter between disorder and order. (a) free pKID (disorder) + KIX (order) proteins (b) encounter state (c) intermediate (d) final complex; both pKID and KIX proteins are folded. This figure was adapted and modified from Eliezer and Palmer Nature 2007.
(a) (b) (c) (d)
Kinetics of binding: The association of two free molecules to form a complex is one of the most important processes in chemical and biological systems. The binding affinity, or the standard free energy change of binding, is simply an alternative way of expressing its equilibrium constant Keq=exp(-ΔGo /RT), which in turn is the ratio of the rate-constants for association (kon) and dissociation (koff). It has been shown experimentally that different molecules that bind to the same chemical receptor may have similar binding free energies (ΔG), but very different binding kinetics (kon, koff). It is unclear why their kinetic features are very different. Therefore, our lab studies not only equilibrium properties, such as the free energy of binding, but also the kinetics of binding. Since basic research, e.g. computational methodology and theories, is needed in this field, we start from tractable simplified models, and then move to more complicated chemical systems and biomedically relevant systems.
Host 1: X = O
Host 2: X = S
Guest host kon (1/M s) koff (1/s) D G(kcal/mol)
Me3NH 1 76.6 4.81x 10-2 4.3
2 2.32x10-3 1.99x10-6 4.1
Figure 2: Cryptophane hosts 1 and 2. Experiments showed that the association and dissociation rates of a guest binding to host 2 can be slowed down by a factor of 103 to 104 with respect to host 1, but the net binding affinities ΔG are essentially the same (see Table for details). There is also a striking difference in the entropy and enthalpy changes on binding (ΔS and ΔH) to these similar hosts.
Figure 3: Binding process diagram for ligand-protein association. ΔG is the free energy difference between the unbound and the final bound states. During the binding process, there may be a free energy barrier ΔG* between the unbound state and the intermediate state.
Stochastic simulations in cell signaling complexes: Computer simulations of complex biochemical networks can be useful tools to rigorously determine if a proposed model is consistent with observed experimental results. We will use Brownian dynamics simulations to model stochastic processes of multiple molecules, in particular multi-protein complexes that mediate signal transduction. Our work aims to provide insights into the complex relationships between the stimuli and the cellular responses, and reveal the mechanisms that are responsible for signal amplification and cellular communications. Important proteins can be represented using coarse-grained models, together with key intermolecular interactions if necessary. It is expected that these simulations will provide useful information regarding the biophysical mechanisms of “signaling diseases”, such as cancer, asthma and diabetes. The method may also be applied to model other systems-level processes in a cell.
Computer-aided ligand/receptor design and discovery: In drug design and discovery, finding a small molecule that maximizes binding free energy is very important and is an interesting challenge. A thorough understanding of driving forces, binding penalties, and conformational changes induced by ligand binding, should enable more accurate prediction of binding affinities. Our lab assembles state of the art methods, e.g. docking and scoring, QSAR and applies our work described above to improve the ligand-design work. We will not only focus on finding tight binding ligands, but also consider drug resistance and early ADME (Absorption, Distribution, Metabolism, Excretion, critical for the optimization of pharmacokinetic properties) in the calculations. For example, before docking, “far-from-drug” compounds in the dataset should be eliminated by applying commercialized pre-screening programs or simple filters, e.g. molecular weight, number of hydrogen bond donors and acceptors. Since we still do not fully understand what features endow a compound with favorable ADME properties, these filtering tools are not highly accurate. I will seek to address this problem with methods based on neural network simulations to distinguish “drug-like” and “nondrug-like” molecules.
Figure 4: Role of computation in drug design/discovery. Computational work is shown in yellow; experimental work is indicated in blue.
Recent projects
- Gated Binding of Ligands to HIV-1 Protease: Brownian Dynamics Simulations in a Coarse-Grained Model
- Multiscale Modeling of Biological Systems--from Atomistic to Coarse-Grained Representation. Application to acetylcholinesterase tetramer, drugs binding to HIV-1 protease and macromolecular crowding effects
- Studies of Free Energy, Entropy, and Induced Fit in Molecular Recognition
- Theory and Modeling of Biomolecular Diffusion and Association Rate Constants
- Computer-Aided Drug Design and Discovery
- Algorithm Development
Modification and application for coarse-grained
Brownian dynamics simulations in UHBD package
Second generation Mining-Minima algorithm (M2) -- a powerful free energy calculation method
HA/MS algorithm -- fast method to compute the configuration integral
Tork conformational search method