Page 118 - ITPS-7-2

INNOSC Theranostics and
Pharmacological Sciences PI3K-α inhibitors for cancer immunotherapy

fraction of compounds in the later stages, the workflow ligand. The reference ligand was obtained after virtual
could ensure that the final hits are diverse, relevant, and screening-molecular docking, and the interactive pose
reliable, as docking and post-processing are more accurate prediction (IPP) panel offered in Schrödinger Maestro
and informative than ligand preparation. Therefore, this was employed for a new ligand design. The IPP operated
ratio was chosen as a default value that balances the speed under the maximum common substructure (MCS)
and accuracy of the virtual screening workflow. In addition, constrained docking type, which simultaneously docked
the virtual screening workflow aided in the computation of the compounds into the binding site of proteins using a
performance scores, including Glide, docking, interaction, grid-based approach. The GlideScore value and predicted
and penalty scores, as well as similarity scores, to validate biological activity from the 3D-QSAR model were the
the refined ligands. Furthermore, during virtual screening, metrics employed for performance verification between
interaction scores for residues within 12 Å of the grid the newly designed compound and the reference lead
center were considered. compound. Furthermore, the interaction pattern of the
designed ligand within the protein was also studied. In
2.7. Implementation of 3D-QSAR addition, the ADMET-related indices of the new ligand
The computational modeling technique employed in this were assessed using the QikProp program in Schrödinger
study was field-based 3D-QSAR to analyze and predict Maestro at normal mode to evaluate its pharmacokinetics,
the relationship between the 3D structure of the refined efficacy, and safety profiles. All ADMET-related indices
ligands based on their alignment, similarity to a known for the new compounds were evaluated using a total of 50
pharmacophore, and biological activity. The pIC values descriptors with a #star parameter as an indicator of several
50
served as a measure of the potency or biological activity of property descriptors computed by QikProp that violate a
the ligands. given optimum range of values for 95% of known drugs.
The pIC values of the refined ligands ranged from 4.866 3. Results and discussion
50
to 9.398. In Schrödinger Maestro, structural alignment was
deployed to identify similar ligand structures, focusing 3.1. Data preprocessing
on identifying the core for each structure to align the Several ligands obtained from the database were dropped
molecules effectively. The field-based model was built on during the data preprocessing step due to missing column
the Gaussian field domain that utilized a training set of 70% information, inconsistencies, and ambiguities in the data
of the total input of refined ligands and a random seed set structure. This step resulted in reducing the initial dataset
to 0, with a maximum of four partial least square factors. size from 3994 rows of ligands to 2972 rows and 48 columns,
Both the steric and the electrostatic force fields were set ensuring a clean dataset for analysis. In addition, the computed
truncated at 30.0 kcal/mol, and the cross-validation was pIC values (activity) for each ligand ranged from 4.54 to
50
performed by leaving out just one ligand. 10.15 throughout the dataset. Using pIC as a measure of
50
Statistical analysis methods, such as comparative activity guarantees that the potency of different compounds
molecular field analysis (CoMFA) or comparative can be precisely compared, facilitating the evaluation of their
molecular similarity indices analysis (CoMSIA), as well efficacy and the selection of the most promising candidates
as partial least squares (PLS), were applied to correlate for further research or drug development.
the calculated descriptors with the activity values of the 3.2. Protein complex refinement
ligands in the training set. The generated correlation was
utilized to predict the activity of new compounds based In mechanistic studies involving drug design, molecular
on their 3D structures. In addition, five Gaussian field docking, and prediction of protein functions, refining and
fractions, including steric, electrostatic, hydrophobic, minimizing protein structures play a significant role in
49
hydrogen bond acceptor, and hydrogen bond donor, were improving their utility in pharmaceutical applications.
evaluated to provide insight into the field interactions of In Figure 4A, the 6PYS protein structure representing the
the ligands within the binding pocket of the receptor. An human PI3K-α protein complex possesses inherent local
optimal number of PLS factors that can balance the trade- and global errors, including irregular contacts or hydrogen
off between data fitting and model prediction was chosen. bonds, chain breaks and atomic clashes, and unusual bond
angles and lengths. However, refining a protein obtained
50
2.8. Rational design of a new ligand from a database before docking improves the accuracy and
The rational design of a new ligand in this study involved reliability of docking results.
a robust approach to iteratively modifying the skeletal Figure 4B illustrates the schematics of the refined
structure of a lead compound, considered the reference 6PYS protein complex with the necessary side-chain

Volume 7 Issue 2 (2024) 8 doi: 10.36922/itps.2340