Drug discovery is a complex and resource-intensive process that involves synthesizing new compounds, often using well-established chemical reactions. While these methods have proven effective, there can be certain biases in the types of molecular structures they produce. For example, common reactions such as Suzuki or amide coupling tend to favor specific molecular frameworks, potentially limiting the diversity of compound libraries.
Skeletal editing, however, provides a distinct strategy.
This blog post is the result of a collaboration between myself, Dr. Phyo Phyo Zin, and Dr. Jeremy E. Monat. While I focus on the high-level principles and applications of skeletal editing in drug discovery, Jeremy has developed a hands-on tutorial [skeletal-editing.ipynb] that guides readers through implementing skeletal editing transformations using Python and RDKit. Together, our goal is to provide both a conceptual understanding and practical approach to this emerging field.
What is Skeletal Editing?
Rather than building molecules from scratch or focusing on peripheral modifications, skeletal editing enables chemists to modify the core structure of existing molecules by swapping, deleting, inserting, or rearranging atoms (Woo et al., 2023).
This can allow for a more efficient diversification of compounds by modifying the foundational framework of the molecule. This can generate novel structures that may be difficult to achieve using traditional methods. While skeletal editing presents clear advantages in specific cases, its applications are still being explored and expanded.
Why Skeletal Editing Matters
Traditional synthetic methods have successfully led to diverse molecular libraries. But, they are often constrained by reaction biases that favor specific frameworks. Skeletal editing addresses these limitations by:
- Modifying Core Scaffolds: This approach involves making targeted changes to the central framework of a molecule—its core scaffold. By altering the shape or composition of this core, researchers can fine-tune a compound’s potency, selectivity, and other key properties, potentially reducing side effects or improving overall performance as a drug.
- Expanding Chemical Space: Skeletal editing enables the creation of molecular frameworks that are not easily accessible through conventional synthetic methods. This helps open up new corners of chemical space—areas with unique shapes and properties. This allows chemists to explore diverse and underrepresented chemical structures.
- Simplifying Synthesis: Skeletal editing is especially useful later in the synthesis process, where it can make major changes without starting from scratch. This helps avoid the time, cost, and effort of building new molecules from the ground up—speeding up the discovery and fine-tuning of complex drug candidates. (Sharma et al., 2025, Xu et al., 2025)
Skeletal editing doesn’t replace traditional methods—it adds a powerful new option to the medicinal chemist’s toolkit. By focusing on the core structure of molecules, it offers a more direct and efficient way to explore chemical space and fine-tune potential drug candidates.
Real-World Examples of Skeletal Editing in Drug Discovery
1. More Precise Drug Modifications
One of the biggest strengths of skeletal editing is how it lets chemists make pinpoint changes to a molecule’s structure—especially in the later stages of drug development when starting from scratch isn’t practical. Instead of rebuilding a whole molecule, skeletal editing allows you to tweak its framework in a way that’s faster and more efficient.
A great example is the carbon-to-nitrogen transmutation used to convert quinolines into quinazolines—a key step in the synthesis of several drugs, including belumosudil, talnetant, and brequinar (Woo et al., 2023).
Figure credit: Adapted from “Carbon-to-nitrogen single-atom transmutation of azaarenes,” Jisoo Woo, Colin Stein, Alec H. Christian, and Mark D. Levin, Nature, Volume 623, pages 77–82 (2023).
The researchers showed how their carbon-to-nitrogen transmutation method works by using it in the synthesis of Talnetant. They started with O-methyl talnetant N-oxide and transformed it into its quinazoline form. This key step reshapes the molecule’s core, helping build the structural foundation needed for Talnetant.
These kinds of single-atom tweaks, like replacing a carbon with a nitrogen in an aromatic ring, can help create molecules with distinct biological activity.
Figure credit: Adapted from “Carbon-to-nitrogen single-atom transmutation of azaarenes,” Jisoo Woo, Colin Stein, Alec H. Christian, and Mark D. Levin, Nature, Volume 623, pages 77–82 (2023).
Extracted from Jeremy’s notebook:
quinoline_to_quinazoline_modified = "[cX3:1]1[n:2][c:3]2[c:4]([c:5][c:6][c:7][c:8]2)[c:9][c:10]1>>[cX3:1]1[n:2][c:3]2[c:4]([c:5][c:6][c:7][c:8]2)[c:9][n:10]1"
talnetant_smi = 'c1ccccc1c2cc(C(=O)NC(CC)c4ccccc4)c3ccccc3n2'
talnetant_mol = Chem.MolFromSmiles(talnetant_smi)
product_smls = plot_rxn(quinoline_to_quinazoline_modified, talnetant_mol)
Reaction:
Reactant:
Distinct products:
CCC(NC(=O)c1nc(-c2ccccc2)nc2ccccc12)c1ccccc1
Re-engineering existing drugs – Skeletal editing at later stages lets scientists fine-tune a drug’s safety and effectiveness—or even give it a new purpose. It’s a quick and efficient way to improve how a drug works, especially when time and resources are limited.
2. Exploring New Chemical Space
Skeletal editing opens the door to molecular shapes that aren’t easy to reach with traditional chemistry. For instance, swapping out a pyrrolidine ring for a cyclobutane ring—by removing a nitrogen atom—can dramatically reshape the molecule. That shift in 3D structure can boost how well a drug binds to its target or moves through the body (Hui et al., 2021).
Figure credit: Adapted from Table 1 of “Stereoselective Synthesis of Cyclobutanes by Contraction of Pyrrolidines,” Chunngai Hui, Lukas Brieger, Carsten Strohmann, and Andrey P. Antonchick, Journal of the American Chemical Society, Volume 143, Issue 45, (2021).
A good example? Hui and colleagues used this editing strategy early in the synthesis of piperarborenine B, a cytotoxic compound that features a cyclobutane core—a structure made possible by this skeletal transformation.
Figure credit: Scheme 3 of “Stereoselective Synthesis of Cyclobutanes by Contraction of Pyrrolidines,” Chunngai Hui, Lukas Brieger, Carsten Strohmann, and Andrey P. Antonchick, Journal of the American Chemical Society, Volume 143, Issue 45, (2021).
Extracted from Jeremy’s notebook:
pyrrolidine_to_cyclobutane_modified = "[C:5](=O)-[C:1]1-[C:3]-[C:4]-[C:2]-N-1>>[C:5](=O)-[C:1]1-[C:3]-[C:4]-[C:2]-1"
starting_smi = 'c1cc(Br)ccc1C2NC(C(=O)OC)C(c3ccccc3)C(C(=O)OC)2'
starting_mol = Chem.MolFromSmiles(starting_smi)
product_smls = plot_rxn(pyrrolidine_to_cyclobutane_modified, starting_mol)
Reaction:
Reactant:
Distinct products:
COC(=O)C1C(c2ccccc2)C(C(=O)OC)C1c1ccc(Br)cc1
3. Streamlining Complex Syntheses
Some drugs are difficult to make and can take up a lot of time. Skeletal editing makes these processes simpler, helping scientists quickly create a wider variety of drug candidates.
Take, for example, the transformation of nitroarenes into azepanes—a seven-membered ring. This opens up the possibility of tweaking piperidine-based drugs like melperone (used to treat psychosis) and fentanyl (a synthetic opioid). By making these changes, researchers can develop new drug versions with different properties and effects (Mykura et al., 2024).
The reaction scheme below is inspired from Figure 5 in Mykura et al., 2024.
The following code snippet demonstrates the transformation of precursor 51 into product 52. If 52 undergoes a further reaction, it can yield 53, an azepane-based drug analogue of fentanyl.
Extracted from Jeremy’s notebook:
nitroarene_to_azepane_modified = "O=[N+](-[O-])-[c:2]1[c:3][c:4][c:5][c:6][c:7]1>>[C:2]-1-[N]-[C:3]-[C:4]-[C:5]-[C:6]-[C:7]1"
starting_smi = 'CCC(=O)N(c1ccccc1)c3cccc([N+](=O)[O-])c3'
starting_mol = Chem.MolFromSmiles(starting_smi)
product_smls = plot_rxn(nitroarene_to_azepane_modified, starting_mol)
Reaction:
Reactant:
Distinct products:
CCC(=O)N(c1ccccc1)C1CCCNCC1
CCC(=O)N(c1ccccc1)C1CCCCNC1
The difference between these two products is where the nitrogen (N) attaches to the ring. One product forms when the nitrogen bonds to a carbon next to the nitro group, while the other forms when the nitrogen bonds to the second neighboring carbon. This can depend on the nature of the substituent and is influenced by factors like electronic effects, steric effects and reaction conditions (Dherange et al., 2022).
Integrating Cheminformatics, Machine Learning, and Computational Chemistry in Skeletal Editing
Skeletal editing is a powerful technique that allows chemists to modify the core structure of a molecule. When combined with advanced fields like cheminformatics, machine learning (ML), and computational chemistry, it becomes even more effective. Together, these tools help researchers speed up drug discovery, optimize molecular designs, and cut down on experimental work.
Cheminformatics: Guiding Rational Selection of Transformable Scaffolds
Cheminformatics uses vast chemical datasets and algorithms to spot patterns between molecular structures and their properties. For skeletal editing, it can:
- Identify the best starting materials for editing reactions based on things like reactivity, availability, and structural compatibility.
- Map out less-explored chemical space, highlighting core structures that could be transformed into novel scaffolds with drug-like properties.
- Help with scaffold hopping—finding new, distinct cores that still have similar biological effects. This can be done through direct skeletal changes rather than designing from scratch.
For instance, if you’re aiming to create azepane-based drug analogs, cheminformatics tools can quickly scan chemical libraries to find nitroarenes that are not only easy to synthesize but also likely to undergo ring expansion, reducing the experimental work involved.
Machine Learning: Optimizing Modifications and Predicting Impacts
Once you’ve chosen potential structures, machine learning models can help figure out which modifications are most likely to succeed. These models, trained on structure-activity relationship (SAR) data, can:
- Predict how a modification will affect things like biological activity, selectivity, and ADMET properties even before the molecule is synthesized.
- Rank various transformation options based on their likelihood of success.
- Use methods like transfer learning and graph neural networks to apply predictions to related scaffolds, even if limited data is available.
For example, when evaluating several azepane products made from different nitroarenes, ML can prioritize which ones to synthesize first based on predicted activity or stability—saving time and effort.
Computational Chemistry: Enhancing Predictions
Computational chemistry tools work hand-in-hand with ML and cheminformatics to help researchers simulate how skeletal edits might affect a molecule’s behavior.
- Researchers can model the energetics and feasibility of specific transformations using methods like DFT or semi-empirical techniques.
- Molecular docking or dynamics simulations can predict how changes to the molecule might affect its binding to target proteins.
- They can also visualize how changes in shape, polarity, or charge distribution might influence a drug’s pharmacokinetics or off-target effects.
For example, converting piperidine to an azepane could alter a drug’s shape or binding behavior. Simulations can reveal whether this improves its fit at the target site or if it disrupts key interactions, helping chemists make more informed decisions.
Challenges and Future Directions
Skeletal editing holds a lot of promise, but there are still some hurdles to overcome:
- Reaction Specificity: Many transformations work only on certain structures. Expanding their use to a wider range of molecules through new catalysts and reaction conditions is a big area of ongoing research.
- Unintended Effects: Some modifications may weaken molecular stability or mess with biological activity. Computational models can help predict these issues, though the accuracy of these models is still improving.
- Early-stage Field: Skeletal editing is still a developing area, and not all methods are ready for widespread use. As the field matures, new techniques will emerge, making skeletal editing even more useful for medicinal chemists.
Conclusion
Skeletal editing isn’t here to replace traditional chemistry methods—it’s meant to complement them. By allowing precise modifications to molecular structures, it enables chemists to explore new chemical spaces, simplify drug synthesis, and optimize drug candidates more efficiently.
The combination of cheminformatics, machine learning, and computational chemistry with skeletal editing is an exciting direction for the future of drug discovery. These tools make it possible to predict, refine, and expand skeletal editing strategies, ultimately speeding up drug development. In the future, skeletal editing will likely play an increasingly important role in creating safer, more effective drugs, and doing it faster than traditional methods.
Thanks for reading this far, and I hope you found this post helpful! If you’re interested in similar projects, I’ve developed a few tools for chemical structure generation, transformation, and enumeration. They might be just what you need to make molecular design faster and more flexible. Check them out below!
- PKS Enumerator
- This is a specialized tool for designing virtual libraries of macrolide scaffolds. It takes modular building blocks and combines them in different ways to create a variety of macrocycles. These diverse structures can be used to explore chemical space or integrate into workflows like docking and QSAR (Quantitative Structure-Activity Relationship).
- SIME
- SIME blends biosynthetic logic with enumeration to design macrolide libraries that are both realistic and easy to synthesize. It’s built around well-known macrolide motifs and sugar decorations, focusing on generating analogs that are both useful and feasible.
- ChemX
- ChemX is a Python-based program that creates virtual libraries based on a target molecule. It works by breaking the molecule into smaller parts, swapping those parts with chemically similar analogs using RDKit, and then building new structures for screening. It was developed at a hackathon in 2019.
- Rxn-Based Enumeration
- This is a guide on how to apply specific reaction-based chemical transformations (like amide coupling) using RDKit and Python.
- Rxn-SMARTS
- A hands-on tutorial for writing and applying reaction SMARTS patterns. You’ll learn how to write the syntax, map atoms, and apply transformations using RDKit. This is a great resource for automating reaction steps in silico.