In silico design

Computational molecule design

By systematically varying and extending the functional groups of the lead blebbistatin molecule we aim to optimize the designed inhibitors binding strength and selectivity for specific myosin isoforms. We apply the Marvin/Chemaxon and Ambertools toolkits to draw, design and approximate the electronic and structural parameters of the possible blebbistatin derivatives. The determined parameters enable the Molecular Dynamics simulations that form the basis of our analyses.

In silico screening

In silico screening involves two steps: finding the best conformation of a compound in the binding site of a protein, and calculating its binding affinity. The first step, the mapping of the conformational space is commonly reffered to as docking while the binding affinity evaluation is called scoring. Our team uses in silico screening to examine potential drug-target interactions, select drug development candidates, develop validating experiments and produce virtual interaction profiles used in Drug Profile Matching.

Molecular dynamics

We use molecular dynamics to create high fidelity models of the interaction between the various myosin isoforms and the inhibitor molecules. We use the state of the art AMBER molecular dynamics program suite to model interactions between myosin and the investigated inhibitors in atomic detail. The modelled protein-inhibitor complex is equilibrated in explicit water solvent and its conformational distribution is sampled for several nanoseconds. Using the distribution we calculate the free energy of the interaction over the entire complex as well as contributions of individual amino acids of the protein. We use the detailed interaction energy profile to predict the inhibitory strength and specificity of the investigated molecules.

Multivariate statistical analyses

Molecular Dynamics runs provide a wealth of data on both the investigated inhibitor molecules and their interaction with the various myosin isoforms. Besides the various interaction energy components between the inhibitor and the protein we analyze the proteins behavior regarding its flexibility, salt-bride and hydrogen bonds. The inhibitor itself is analyzed with regards to the Van der Waals and electric interaction potential it exerts to its surroundings by integrating these interaction potentials on a fine grid. Based on these data we use both linear and non-linear regression models to predict the inhibitory properties of candidate molecules, by training the models on experimental data already measured.

Drug Profile Matching

Drug Profile Matching (DPM) is a novel computational drug development tool applied to quantitatively predict the pharmacological effect and target profiles of drugs and drug candidates in order to gain a comprehensive view of their bioactivity properties. More than 1,000 FDA approved drug molecules were docked to the binding sites of 149 non-target proteins, i.e., proteins that are not known to be involved in the mechanisms of action of any of the studied drugs and their interaction profiles were generated. Available pharmacological effect and target information on the drug molecules was also collected and binary effect and target profiles were created for each molecule. In order to relate the complex interaction profiles to the generated effect and target profiles, a two-step multivariate method (canonical correlation and linear discriminant analyses) was adapted that identifies the best discriminating surfaces (’hyperplanes’) in the multidimensional space of the 149 reference proteins. Using the mathematical equations of the created discriminating surfaces, the probability of each drug-effect and drug-target pair were calculated. Hence, the likelihood of possessing a given effect or interacting with a studied target could be quantitatively described for all drugs in human use. The method is also capable of predicting the bioactivity properties of new molecules after generating their interaction profiles. For further information and testing DPM click here.

Graphical summary of the Drug Profile Matching method: from the atomic structures to the effect probability matrix. The effect pattern (EP) matrix contains the therapeutic effects of the drugs in a binary coded form (blue and white cells represent the presence and the absence of a given effect from the 177 categories, respectively). Then, a canonical correlation analysis is performed with the IP matrix in order to generate highly correlating factor pairs that serve as the input for linear discriminant analysis. This way, classification functions are produced that yield the probability for each drug–effect pair, resulting in the effect probability matrix.

DPM is described in details in our paper “Drug Effect Prediction by Polypharmacology-Based Interaction Profiling” appeared in the Journal of Chemical Information and Modeling (JCIM) as a cover story. The issue is a sample issue of JCIM and the full text paper is available free of charge.

Hardware background

Our consortium partner Printnet Ltd. has set up a new computational cluster, which increases our calculation capacity by 10-fold. This cluster will have the greatest computational capacity ( 50 Teraflop/s) dedicated to a single project in Hungary.