Page 90 - TD-4-1

Tumor Discovery Identification of a potential KRAS(G12C) inhibitor

2.5. Post-MD simulations analyses COVDOCK protocol was utilized to investigate the

The analysis employed Bio3D, an R library-based tool, to redocking capabilities as well as the binding of the
44
generate time-correlated DCCM for various ligand-bound co-crystallized Sotorasib to the KRAS(G12C) with GDP.
forms of KRAS(G12C). This approach aimed to discern the As a result, we observed a similar binding mechanism to
α
localization and nature of fluctuations in C atoms within that observed in the crystallized structure, validating the
residues crucial to KRAS(G12C) function. Given the covalent docking protocol (Figure 1A). Next, the binding
limitations of DCCM, which may overlook correlated but modes of these four compounds were compared with the
perpendicular motions, the analysis was complemented co-crystallized Sotorasib according to their 2-D interaction
with PCA. In PCA, simulation frames are organized based diagram (Figure 1B). The precision of the COVDOCK
on their principal components (eigenvectors), which protocol in reproducing the binding mechanism of the
represent the key directions capturing the most significant co-crystallized ligand with the KRAS(G12C) protein has
22,45
variability in the data. This approach allows for effective been demonstrated by prior researchers.
separation of structures by emphasizing dominant patterns The set of 174 selected compounds was then subjected
and variations within the dataset. For both analyses, the to a covalent docking-based virtual screening, with their
final 400 ns of all MD simulations were utilized at 20 ps binding energies ranked according to their Glide scores.
intervals per Bio3D file size constraints. This comprehensive While none of the ranked compounds displayed a higher
approach provides a nuanced understanding of the dynamic estimated binding energy compared to Sotorasib, which
behavior of KRAS(G12C) ligand-bound forms, integrating possesses a value of −8.1 kcal/mol, four compounds
both normal mode analysis and PCA techniques to capture namely C01, C02a, C02b, and C03 exhibited closer and
a broad spectrum of structural fluctuations. notable Glide scores of −7.8 kcal/mol, −7.5 kcal/mol,
−7.3 kcal/mol, and −7.0 kcal/mol, respectively (Table 2
3. Results and Figure A1).

3.1. Covalent docking-based virtual screening The comparison of binding modes for the four
It is widely accepted that molecules with comparable compounds with co-crystalized Sotorasib exhibited Pi-Pi
chemical structures will have similar pharmacological stacking interactions with the residue TYR96. Similarly,
properties. In ligand-based studies, Sotorasib, an FDA- compounds C01, C02a, C02b, and C03 exhibited similar
approved drug for treating KRAS(G12C) protein in interactions (Figure 2). Furthermore, these compounds
NSCLC, was reported to be a reliable template. To evaluate formed hydrogen bonds with additional residues (LYS16,
ligand interactions, we used the InfiniSee, a platform ALA59, and GLN61) in the binding pocket of KRAS(G12C)
capable of examining billions of compounds across (Figure 2). Taken together, the four candidates show
chemical spaces, including eXplore, Freedom Space, promise as covalent binders of KRAS(G12C) in its inactive
GalaXi, CHEMriya, and REAL, to identify potential GDP-bound configuration.
chemical scaffolds similar to the template. This platform
also provides access to KnowledgeSpace, a literature-based 3.3. MD simulations of C01, C02a, C02b, C03, and
virtual chemical space with a strong emphasis on synthetic Sotorasib with KRAS(G12C) protein
accessibility. These technological advancements offer great The root RMSD analysis allows us to monitor the fluctuation
potential in the fields of drug discovery and chemical in the three-dimensional structure over time, offering
research due to its capacity to provide a greater spectrum valuable insight into the mobility of binding pocket residues
of molecules, allowing the design of more effective during the MD simulation. RMSD is a measure used in
and innovative therapeutic solutions. As a result, 174 computer simulations to determine how far a molecule
molecules exhibiting structural similarities to Sotorasib or part of it has moved from its initial position. To assess
were found from billions of compounds in the chemical possible fluctuations in the 3D structure over time, we ran
spaces (Table A1 in Appendix). Subsequently, the selected simulations on various KRAS(G12C) protein complexes,
compounds were further evaluated using a covalent including KRAS-C01, KRAS-C02a, KRAS(G12C)-C02b,
docking-based virtual screening approach. KRAS-C03, and KRAS(G12C)-Sotorasib. Initially, we
ran simulations for all complexes for 100 ns. Following
3.2. Covalent docking-based virtual screening that, we increased the simulation time for KRAS-C01,
identifies four compounds KRAS(G12C)-C02b, and KRAS(G12C)-Sotorasib to 300
For the covalent-based virtual screening approach, the ns. Finally, we ran simulations for KRAS(G12C)-C02b
X-ray structure of the KRAS(G12C) (PDB ID: 6OIM), and KRAS(G12C)-Sotorasib for 500 ns. This enabled us
forming a complex with Sotorasib, was selected. The to compare the structural stability of potential therapeutic
29

Volume 4 Issue 1 (2025) 82 doi: 10.36922/td.5163