This question is often debated since the apparition of the first 3D computational tools in the late 80s – early 90s.

It generates a wide range of reactions from (medicinal) chemists but also from their peers, computational chemists, or molecular modelers. In both parts of a SBDD (Structure-Based Drug Design) team, people can be enthusiastic, or hostile, and even afraid about this idea.

The debate always goes around the reliability of the predictions, the cost in time and effort to produce results, the cost of learning to use the tools, the cost of interpreting them and finally and most importantly, its usefulness. Some argue that it is faster, more useful, and reliable to conduct the appropriate experiment directly in the wet-lab than running some in-silico predictions. Some other people believe in-silico predictions are a useful tool that shouldn’t be neglected and that the best approach is to combine both.

In this article, we sat down with Dr. Alexis Denis, Ph.D. Medicinal Chemistry Director at Galapagos to collect his thoughts on this question, to understand how he manages the collaboration in SBDD teams, but also to understand his vision of how these positions/roles could evolve in the future.  

Could you introduce yourself, your training, your background?

Hello! I am Dr. Alexis Denis. I am a medicinal chemist with 32 years of experience in the big pharma and the biotech industry, and I have a Ph.D. in organic chemistry.

I started my career as a medicinal chemist at Aventis. There, I had the chance to work with a chemist who became a modeler and taught me how to have pragmatic thinking based on the 3D structures of proteins and ligands.

I evolved from researcher to project leader at Aventis then, joined Pfizer as a Chemistry Department Manager. In 2004, I became Chemistry Director at Mutabilis, GSK, and Oncodesign.

Today, I am a Medicinal Chemistry Director at Galapagos, I work with a team of ~ 35 people, chemists (org & syn), and modelers. In my job as a medicinal chemist, my goal is to find new drugs by using all means at my disposal. This includes understanding how the drug interacts with the target (protein or RNA) in 3 dimensions.

With this approach of using 3D, I was able to develop more than 15 drug candidates, 2 molecules in clinical phase, and one on the market.

What is your personal experience working with 3D information?

At the beginning of my career, 3D tools were not user friendly at all! We had to enter commands in a basic Linux environment. I also worked with Tripos, but I really understood the modeling world by using a tool called Flow-plus-QXP. It was the first tool that was doing in-situ minimization illustrating Van Der Walls forces with simple color representations, in a very didactic way.

In other experiences, I also used Discovery Studio from Accelrys and MOE, which is the most simple and complete tool for my pragmatic approach.

Did it bring value to your project(s)? Was it an easy experience?

My first experience with this approach was in 1995. We used 3D to make some kind of a mini-docking of fragments. From these results, we were doing 3D minimizations that allowed us to understand the 3D structure of the molecule and generate new ideas. We validated those hypotheses by synthesizing and testing these ideas.

Since then, I use this approach and train chemists I work with to use it. It helps us accomplish lead optimizations and de-novo designs.

How were medicinal chemists using 3D information in the past? Are Medicinal Chemists and Molecular Modelers using it differently?

I was always opposed to the classic organization where Medicinal Chemists and Modelers poorly interact with each other. I always wanted medicinal chemists to generate hypotheses using 3D and to think in 3D.

Sometimes, Medicinal Chemists can be reluctant to use this approach and Modelers are under the impression that their job is at risk. But often, these limits are more psychological than technical. We will always need Modelers to develop, use, and understand the algorithm and Chemists to draw, synthesize, and test molecules.

To overcome these limits, I show them that it is quicker for the Medicinal Chemist to minimize in-situ directly than to (in)validate with a docking result from the Modeler. Only a few people have success with the classical method.

For example, on our last project, in nearly 9 months, starting from scratch we have been able to generate by design 120 products to obtain one in the nanomolar scale! This has been made possible by using our brain first and directly analyzing the interactions and understanding Van Der Walls forces.

Nevertheless, we will always need Molecular Modelers for complicated projects, lead optimization stages, and when using a fragment-based hit generation approach.

How do you organize the work with 3D data? How is the work split between crystallographers, medicinal chemists, and molecular modelers?

It depends on the context.

If we already have the 3D structure, a co-crystallized protein & ligand. We will study it and do in-situ minimization of fragments or modified leads.

If we only have a model, we are doing minimization in-situ and outside of the pocket (void situation).  If we find some low energy conformers inside the binding site that were identified thanks to the void minimization, then we consider our hypothesis as “good” and synthesize the compound.

As opposed to the classic approach where the docking will make the compound fit in the pocket despite its internal instability which generates a lot of waste.

By using this method, we want the 3D analysis to give the maximum of arguments (Van Der Walls forces, angles, rotations,…) to guide the synthesis.

In the future, what is your vision about the collaboration between Medicinal Chemists & Modelers?

Like I said earlier, Modelers sometimes think that Medicinal Chemists (by using 3D in their reflection) could put their jobs at risk. But technology is a bigger threat from my point of view. Innovation creates new user-friendly tools that could do their job more effectively, and cheaper.

However, there is still a large area of the modelization that we did not explore yet! I believe it is in this direction that the modeler's job is moving.

With the rapid increase in computing power, the number of structures available, etc… How do you envision the Structure Base Drug Discovery activity?

It is hard to predict but today, I see potential in one challenge: capturing knowledge generated by the scientist (Medicinal Chemists & Modelers).

Today, we do not save, nor reuse ideas, or hypotheses generated during a project. Even if the hypothesis is working.

Thanks to this knowledge and the explanation of why one molecule is working or not. I think it could be very interesting to develop an algorithm able to generate hypotheses in-situ. Thanks to the evolution of AI or ML, this algorithm could become reality, but I do not see this in the near future in our field. It will probably take a long time.

What will be Medicinal Chemist and Modeller profile will become? How positions will evolve?

The impact of AI on the Medicinal Chemists and Modelers world isn’t easy to evaluate.

There are major differences in these worlds. Modelers are developing algorithms, which links them to a world closer to equations and mathematics. On the other hand, Medicinal Chemists, even if they feel more at ease with computers, they are not trained for this kind of activity.

Also, even if tomorrow, AI can generate new ideas, Medicinal chemists are facing other challenges understanding the distribution and the metabolism of a molecule.

There is also still a lot of work in the world of modeling. We are still far from reality with our current tools.

Positions will evolve of course, but there is still plenty of work to do!

Quick wrap up!

Dr. Alexis Denis's point of view is very clear on the question. Medicinal chemists should work with 3D tools. He expresses a strong belief in this system, and it has been a source of success for him.

Moreover, even if medicinal chemists move to using 3D tools in their position, this does not necessarily mean than computational chemists or modelers’ positions will disappear. On the contrary, Alexis Denis believes that both roles will evolve and that there is still a lot to discover and develop for this computational workforce.

And you? What are your thoughts on this question?
Should (medicinal) chemists use 3D molecular modeling tools?

Feel free to interact in the comment sections (respecting each other and their opinions).

Alexis Denis, PhD - Medicinal Chemistry Director at Galapagos

3decision’s blog has been moved!