From receptor dynamics to therapeutic candidates.

in particular G protein-coupled receptors (GPCRs), are the most important drug targets. As an example, about 40% of the clinically used drugs are targeting GPCRs.
In some pharmaceutical companies, this figure can be even as high as 70-80%. Therefore, the understanding of membrane protein functions is of fundamental importance for modern drug discovery.

Our team members have made important contributions to understand receptor mediated signalling at the molecular level. They have published many innovative works in top journals including Cell, Nature,
Nature Communications, Chemical Science, Trends in Biochemical Sciences, Angewandte Chemie and others.

Modern drug discovery and development is a complicated process, which costs about $2-3 billion and lasts for 12 years on average. How to decrease the costs and speed up the process is the challenge. Computational biology and Artificial Intelligence, combined with new technologies in experimental drug screening, are expected to make the hunt for new drugs quicker, cheaper and more efficient.
Our team has established a sophisticated computer-aided drug discovery platform, which combines Artificial Intelligence, WebGL, cloud computing and big data. It can perform a wide range of drug discovery tasks such as molecular generation, virtual screening, lead optimization, drug property optimization, molecular dynamics simulation, accurate binding energy prediction and others. We have applied our tools successfully to discover and design potent drug candidates for various targets.
Members of our team have the experience of advancing designed molecules into clinical trials. More recently, we discovered a series of potent compounds for different orphan receptors. These compounds could result in "me-only" drug candidates.

Our team has developed an absolute binding energy prediction tool, which is known as AlphaE. With the new algorithm of AlphaE, we can predict accurately the binding activity of a compound. The RMSE (root-mean-square-energy) is as low as 1 kcal/mol in comparison with experimental data. AlphaE also overcomes many difficulties in currently available tools such as: (1) predictions are not reliable when the molecule changes too much; (2) calculations cannot be performed if molecules bind to different pockets; (3) predictions need large amount of CPU and GPU resources.
Furthermore, AlphaMol also developed many new computational algorithm and tools to predict the physical properties and drug-likeness of molecules. These tools demonstrated perfect correlations with experimental data, of which the R2 is up to 0.99.