Benchmarks for interpretation of QSAR models

Interpretation of QSAR models is useful to understand the complex nature of biological or physicochemical processes, guide structural optimization or perform knowledge-based validation of QSAR models. Highly predictive models are usually complex and their interpretation is non-trivial. This is particularly true for modern neural networks. Various approaches to interpretation of these models exist. However, it is difficult to evaluate and compare performance and applicability of these ever-emerging methods. Herein, we developed several benchmark data sets with end-points determined by pre-defined patterns. These data sets are purposed for evaluation of the ability of interpretation approaches to retrieve these patterns. They represent tasks with different complexity levels: from simple atom-based additive properties to pharmacophore hypotheses. We proposed several quantitative metrics of interpretation performance. Applicability of benchmarks and metrics was demonstrated on a set of conventional models and end-to-end graph convolutional neural networks interpreted by the previously suggested universal ML-agnostic approach for structural interpretation. We anticipate these benchmarks to be useful in evaluation of new interpretation approaches and investigation of decision making of complex “black box” models.

Elaboration of Novel TTK1 Inhibitory Leads via QSAR-Guided Selection of Crystallographic Pharmacophores Followed By In vitro Assay

Introduction: Tyrosine threonine kinase (TTK1) is a key regulator of chromosome segregation. TTK targeting received recent concern for the enhancement of possible anticancer therapies. Objective: In this regard we employed our well-known method of QSAR-guided selection of best crystallographic pharmacophore(s) to discover considerable binding interactions that anchore inhibitors into TTK1 binding site. Method:Sixtyone TTK1 crystallographic complexes were used to extract 315 pharmacophore hypotheses. QSAR modeling was subsequently used to choose a single crystallographic pharmacophore that when combined with other physicochemical descriptors elucidates bioactivity discrepancy within a list of 55 miscellaneous inhibitors. Results: The best QSAR model was robust and predictive (r2(55) = 0.75, r2LOO = 0.72 , r2press against external testing list of 12 compounds = 0.67), Standard error of estimate (training set) (S)= 0.63 , Standard error of estimate (testing set)(Stest) = 0.62. The resulting pharmacophore and QSAR models were used to scan the National Cancer Institute (NCI) database for new TTK1 inhibitors. Conclusion: Five hits confirmed significant TTK1 inhibitory profiles with IC50 values ranging between 11.7 and 76.6 micM.

Pharmacophore Modeling and Docking Study of Pyrazolylaminoquinazoline Derivatives as Highly Potent Fibroblast Growth Factor Receptor Inhibitors2 (FGFR2)

In this study pharmacophore modeling and molecular docking investigations have been performed on pyrazolylaminoquinazoline derivatives, highly potent fibroblast growth factor receptor2 (FGFR2) inhibitors. The best pharmacophore hypotheses displaying five features (ADHRR.2051 and AADHR.798) were generated using a set of 28 compounds. The associated 3D atom-based quantitative structure � activity relationships (QSAR) models were statistically robust showing high correlation coefficients (R-squared = 0.981 / 0.982), and cross validation coefficients (Q-squared = 0.645 / 0.671). The R-Pearson values for the test set of 0.805 / 0.820 indicate that the models are robust and exhibit good predictive power. The interactions of pyrazolylaminoquinazoline with FGFR2 binding site revealed two hydrogen bonds with Ala567. The obtained pharmacophore, 3D atom-based QSAR models and binding features resulted from docking studies can help medicinal chemists to design new pyrazolylaminoquinazoline inhibitors with improved potency.

Multi-Pharmacophore Modeling of Caspase-3 Inhibitors using Crystal, Dock and Flexible Conformation Schemes

Aim and Objective: Numerous caspase-3 drug discovery projects were found to have relied on single receptor as the template to recognize most promising small molecule candidates using docking approach. Alternatively, some researchers were contingent upon ligand-based alignment to build up an empirical relationship between ligand functional groups and caspase-3 inhibitory activity quantitatively. To connect both caspase-3 receptor details and its inhibitors chemical functionalities, this study was undertaken to develop receptor- and ligand-pharmacophore models based on different conformational schemes. Material and Methods: A multi-pharmacophore modeling strategy is carried out based on three conformational schemes of pharmacophore hypothesis generation to screen caspase-3 inhibitors from database. The schemes include (i) flexible (conformations unrestricted or flexible during pharmacophore mapping), (ii) dock (conformations obtained using FlexX docking method) and (iii) crystal (extracted from multiple caspase-3-ligand complexes from PDB repository) conformations of query ligands. The pharmacophore models developed using these conformational schemes were then used to identify probable caspase-3 inhibitors from ZINC database. Results: We noticed better sensitivity with good specificity measures returned by candidate pharmacophore hypotheses across each conformation type and recognized crucial pharmacophore features that enable caspase-3 binding. Pharmacophore modeling based on flexible conformational scheme indicated that the crystal structure 3KJF (AAAADH) is the best receptor structure to perform receptor-based pharmacophore screening of caspase-3 inhibitors. When multiple crystal structures were included, the hypothesis (HAAA) is more generalized. Superimposition of multiple co-crystal ligands from various caspase-3 PDB entries in crystallographic binding mode revealed similar hypothesis (HAAA). Further, FlexX-guided dock conformations of validation dataset showed that the crystal structure 1RE1 is the best-suited for dock-based pharmacophore models. Database screening using these pharmacophore hypotheses identified N'-[6-(benzimidazol-1-yl)-5-nitro-pyrimidin-4-yl]-4 methylbenzenesulfonohydrazide and 2-nitro-N'-[5-nitro-6-[N'-(p-tolylsulfonyl)hydrazino]pyrimidin-4- yl]benzohydrazide as the probable caspase-3 inhibitors. Conclusion: N'-[6-(benzimidazol-1-yl)-5-nitro-pyrimidin-4-yl]-4 methylbenzenesulfonohydrazide and 2-nitro-N'-[5-nitro-6-[N'-(p-tolylsulfonyl)hydrazino]pyrimidin-4-yl]benzohydrazide may be tested for caspase-3 inhibition. We believe that potential caspase-3 inhibitors can be recognized efficiently by adapting multi-pharmacophore models in database screening.

In Silico Design of Human IMPDH Inhibitors Using Pharmacophore Mapping and Molecular Docking Approaches

Inosine 5′-monophosphate dehydrogenase (IMPDH) is one of the crucial enzymes in thede novobiosynthesis of guanosine nucleotides. It has served as an attractive target in immunosuppressive, anticancer, antiviral, and antiparasitic therapeutic strategies. In this study, pharmacophore mapping and molecular docking approaches were employed to discover novel Homo sapiens IMPDH (hIMPDH) inhibitors. The Güner-Henry (GH) scoring method was used to evaluate the quality of generated pharmacophore hypotheses. One of the generated pharmacophore hypotheses was found to possess a GH score of 0.67. Ten potential compounds were selected from the ZINC database using a pharmacophore mapping approach and docked into the IMPDH active site. We find two hits (i.e., ZINC02090792 and ZINC00048033) that match well the optimal pharmacophore features used in this investigation, and it is found that they form interactions with key residues of IMPDH. We propose that these two hits are lead compounds for the development of novel hIMPDH inhibitors.

Dual Binding Site and Selective Acetylcholinesterase Inhibitors Derived from Integrated Pharmacophore Models and Sequential Virtual Screening

In this study, we have employedin silicomethodology combining double pharmacophore based screening, molecular docking, and ADME/T filtering to identify dual binding site acetylcholinesterase inhibitors that can preferentially inhibit acetylcholinesterase and simultaneously inhibit the butyrylcholinesterase also but in the lesser extent than acetylcholinesterase. 3D-pharmacophore models of AChE and BuChE enzyme inhibitors have been developed from xanthostigmine derivatives through HypoGen and validated using test set, Fischer’s randomization technique. The best acetylcholinesterase and butyrylcholinesterase inhibitors pharmacophore hypotheses Hypo1_A and Hypo1_B, with high correlation coefficient of 0.96 and 0.94, respectively, were used as 3D query for screening the Zinc database. The screened hits were then subjected to the ADME/T and molecular docking study to prioritise the compounds. Finally, 18 compounds were identified as potential leads against AChE enzyme, showing good predicted activities and promising ADME/T properties.