How did GenieUs prepare for the challenge?
Illustration by Henris Kas with permission
You may be wondering the process behind the 12-week challenge. So, let’s get into a bit more detail.
The challenge will be split up into two portions. One delegated to GenieUs and the other to Arctoris.
This post will focus on GenieUs
GenieUs will start by utilising their machine learning and molecular dynamic modelling tools. Their unique machine learning tool uses ALS genomic data to find genes that are differentially expressed. Once genes are identified, they know which genes/proteins to target.
Now they have a set of targets/proteins that are possibly contributing to Amyotrophic Lateral Sclerosis (ALS). GenieUs will model each protein structure using their unique molecular dynamic modeling tool.
The next step is to filter the target candidates. A protein is usually very flexible and dynamic. Its shape can change in response to its environment or other factors. They want to choose a protein that is in the same conformation as it is in the human body. This selection process is achieved through a higher temperature simulation.
Once a target is selected, drug binding pockets are identified. Their scientists will introduce compounds/ligands and assess which ones bind most efficiently/in the most stable manner. This process is done in two phases.
1. Blind screening
2. Targeted screening
Blind screening is when you allow a compound to freely bind at any site of the protein. Once this phase is complete, analysis is performed is to find the compounds binding closest to or to the binding site. These compounds are taken to the next screening phase – targeted screening. Here, the compounds are programmed to bind only to the binding site.
Once targeted screening is complete, researchers will obtain a ‘pose’ count for each ligand (the number of different orientations of binding to the protein). Poses indicate the flexibility to the ligand during docking. A statistical scoring technique will be used to rank the ‘poses’ of each ligand. Poses are ranked based on how energetically favourable the binding position is (most stable) when forming the protein complex. Let’s take a scenario:
You have two ligands. The first, you have 1 pose which is highly stable, but the rest are unstable. The second, most poses are stable, but none are as stable as the one pose of the first ligand. GenieUs recommends using the ligand that has more stable poses over the ligand with one highly stable pose. This is because there is more room for error as the ligand can bind to the target in different ways and still have an effect.
GenieUs have two screening pipelines:
Fragment Screening – Here they performed screening on small fragment like approved drug compounds (~500) with additional set of chemical fragments (~800)
Small Drugs Screening – In this screening pipeline, GenieUs used all approved small drug compounds (~2000) irrespective of their size and mass.