Problem: How to use Makya to generate new and novel molecules around a given starting point when you have no models (QSAR and/or docking)?
Here, we have a scenario where we have a reference compound coming from the competition or a project, and no data is associated to it nor any structure-based information. In this context, you cannot build a QSAR model or a docking score to guide the generation. The only possible strategy is then to find easily synthesizable molecules, with close similarity to the reference molecule, but still different enough to get IP/new ideas. This similarity can encompass 3D information, such as 3D shape similarity. As mentioned by Sir James Black, Laureate of the 1988 Nobel Prize in Physiology and Medicine: “The most fruitful basis for the discovery of a new drug is to start with an old drug”.
We will show you how to do that in a few clicks inside Makya.
For this use-case, you can use either the Fragment Growing or the Fragment Linking generators, depending on your strategy and which part(s) of the molecule you want to keep. Since we have no model, so no guarantee regarding the activity of the generated molecules, we wish to use easy chemistry, which is a strong asset of the Growing/Linking generators.
Here, we will start from this molecule :
We wish to replace some parts of the molecule (either the central core or branches) with commercially available building blocks.
Step 1: Create the 3D Ligand-based parameters
Since we have a reference ligand, if we know a bioactive conformation for it, we can use 3D Ligand-based parameters in Makya. This let you generate molecules that maximize the 3D Shape Similarity and 3D Pharmacophore Similarity to your reference molecule, thus potentially maintaining good activity with respect to the target.
To set up 3D ligand-based parameters, follow the steps described in the documentation. You will simply need the SMILES of your reference ligand:
CN1N=CC(NC(=O)C2=C(N)SC(=N2)C2=C(F)C=CC=C2F)=C1N1CCC(N)C(F)CC1
and the substructure that we will use as an anchor (a reference point for structural alignment in 3D):
Cn1cc(NC(=O)c2nc(-c3c(F)cccc3F)sc2N)cn1
as well as and an idea of what a bioactive conformer looks like. For the purpose of this exercise, select the first generated conformer.
Step 2: Create the Fragment Linking generator
Go directly to the "Generators" tab since you have no data nor any model to upload. Click on the New Generator box. Name the generator, and under "Generation Engine" select Fragment Linking and click on Next. On the following page, specify the various options to set up this generator.
- Chemical Space: Create a new chemical space (top right) by pasting a SMILES or drawing the reference molecule molecule in the sketcher module:
CN1N=CC(NC(=O)C2=C(N)SC(=N2)C2=C(F)C=CC=C2F)=C1N1CCC(N)C(F)CC1
Explicit the hydrogens and click Save. This molecule will be used to guide the generation using Tanimoto similarity.
- Exit Vectors: On this tab, you will see two windows where your initial fragments can be added. Here, we use the Fragment Linking generator so two fragments are needed.
Click on the pencil icon to input the SMILES of the reference molecule:
CN1N=CC(NC(=O)C2=C(N)SC(=N2)C2=C(F)C=CC=C2F)=C1N1CCC(N)C(F)CC1
From this reference molecule, remove and/or add relevant atoms to get your two fragments.
NOTE: the Linking and Growing generators are based on chemistry rules, so it is important to make sure that on the exit vector, an organic reaction can actually takes place. See the dedicated pages of these generators in the documentation for more details.
Click on Set for each of the intermediates to define their exit vectors.
It is only at these exit vectors that the scaffold will be attached while the rest of the fragment will remain unchanged. Click on Save.
- Scorers: You can build a scorer to filter the results with some basic descriptors, for instance:
- 3D Ligand-based: Here, you can select the configuration that you defined earlier. It will help you generate compounds whose 3D shape matches the shape of your reference molecule.
Go back to the "Generators" tab, and the newly set up generator will show up. Click on the Run button. The progress bar will indicate the status of generation.
Step 3: Explore the generated molecules
Visualize the generated molecules by clicking on the eye icon. Sort the grid of generated molecules by Similarity (= Similarity to the reference molecule selected in Chemical Space) in descending order.
You can browse the results and select the most interesting ones, based on your understanding of the project, the overall shape of the molecule, the scores that were calculated... and add them into the cart by simply clicking on Selection > Add selection to the cart.
Also, by clicking on See details, you have access to the commercial building blocks used to generate the given molecule.
The examples from this generation make it clear that the core of the reference compound has been replaced with novel scaffolds, while keeping the rest of the molecule intact. As we did in previous examples, the generated molecules can be exported to a CSV file for further analysis or a subsequent iteration of generation in Makya.
Alternative to Step 2: Create a Fragment Growing generator
Depending on your project and expectation, the Fragment Growing approach can also be envisaged. Here two generations using two different starting points have been considered:
For the first one, we start from the following Exit vector with the same chemical space as previously described in Step 2:
Nc1sc(-c2c(F)cccc2F)nc1C(=O)O
In addition, in the Building blocks tab, we prevent the introduction of the pyrazol motif to make sure that the generated molecules will be different enough compared to the reference molecule.
Similarly to the linking approach, scorers can be envisaged for selection purpose. Moreover, the 3D ligand-based parameters defined above can be selected so as to generate compounds with optimal 3D resemblance to the reference ligand.
Sorting the results by similarity or scores and after some manual selection, interesting ideas can be observed:
Another generation can be considered, using a less elaborate initial fragment:
This generation provides more novel and diverse molecules compared to the previous one since the starting point is simpler:
Note: We recommend using 3D ligand-based parameters to guide the generation towards potentially active molecules. Without 3D ligand-based parameters, and with a simple Tanimoto similarity guiding the generation, there is no guarantee that the generated molecules will be active. However, such an approach still allows you to generate diverse and novel ideas with a high synthetic feasibility, and as a result, to converge on active molecules at a low cost and after few design, make and test cycles.