Our Innovative structure based in silico drug design using Artificial Intelligence (AI)
Structure based in silico drug design became widely used in the past dedicate but it was not often employed in the real life, i.e. from the big pharma companies. This has been mainly changed during the last few years by the introduction of FEP+ technology by Schrödinger. Many projects are under development now in which most of the big pharmaceutical companies were involved. In this way, many other in silico approaches became also well recognized. The main reason for this progress is the new computation capability and in particular the use of modern GPUs instead CPUs for the calculations, which makes possible an acceleration of about 30-100 times. Unfortunately, although the cosmetics active substance and drug discovery are similar in their initial phase the in silico design is rare used by cosmetics companies.
Our innovative approach includes all modern approaches but also our own innovations. For instance, it is not secret that the initial compounds pose prediction and scoring is very important step and without a good starting ligand conformation and equilibrated protein structure in solution all other computational steps may fail. Moreover, it has been recently established that more and more of the drug candidate compounds have a mutable binding modes in a solution. Thus, we have developed a new strategy for both the binding mode and free energy scoring prediction based on the modified accelerated molecular dynamics (aMD). By our approach even the binding constants such as kin and koff can be more easily calculated with help of the computational power of the new GPUs such as Nvidia Tesla generation. Further, we greatly improved the in silico hit compounds identification from an average of 10-20% success rate up to 70-80% by our new combination of structural based pharmacophore and docking virtual screen. This is in a great help during the initial stage of any drug discovery pipeline. Finally, although the replica exchange solute tempering (REST) method is indeed a good sampling approach and has been already introduced in the FEP+ calculations during the lead optimization process, the short FEP/REST simulation time does not guarantee impressive results neither in a case of flexible receptor binding sites nor homology modelling/MD derived structures. Thus we recently improved the FEP+ sampling protocol and achieved much better predictions of the binding affinities which will be of help in any hit to lead drug discovery project. Moreover, the ligands binding modes are often not well established thus our preliminary MD simulations are of great help. Because of the fact that the lead optimization is the most expensive step in the drug discovery we combine the FEP approach and the above mentioned our own method to achieve much better results.
Many companies claim that provide in silico drug discovery but a close view of the technologies used shows that only some steps of this complicate process are covered and mainly not well established methods are employed. Instead, we cover all steps by all sophisticated technology available. We can screen a databases with millions chemical compounds and precisely predict the lead drug candidates by many scoring ways. Further, for the lead optimization step the FEP methodology is used but the results are also supposed by our own developments.
Our technology combines but is not limited to following methods:
Artificial Intelligence (AI)
Docking
Pharmacophore
Scoring by all available advanced functions and methods
Molecular dynamics
Accelerated Molecular dynamics
New core recognizing
Free energy perturbation (FEP)
GPU calculations
