Artificial Intelligence Applied In Targeted Drugs Design And Auxiliary Diagnosis Of Non Small Cell Lung Cancer

ObjectiveRecent studies have identified the receptor tyrosine kinase Mer as a potential therapeutic target for non small cell lung cancer(NSCLC). To deeply understand the structure–activity correlation of a series of potent Mer inhibitors, explore molecular structure features to their bioactivity and study the interaction between the targeted protein and their inhibitors, molecular modeling techniques were carried out that will aid in the mechanism research and development of novel potent Mer kinase inhibitors. Basis on a number of serum biomarkers, the gene expression programming(GEP) method was used in the auxiliary diagnosis of non small cell lung cancer to improve the accuracy rate of diagnosis.Methods1. Based on the structures of a series of analogues as potent Mer inhibitors,three-dimensional quantitative structure–activity relationship modeling was carried out to explore structural features that may affect the Mer inhibitory activities. A molecular docking study was performed to analysis the conformations of these inhibitors in the binding site of Mer kinase, and study the molecular mechanism of receptor-ligand interactions. Then, molecular dynamics simulations including binding free energy calculations and per-residue energy decomposition were used to study the different binding modes of inhibitors with different activities, and identify the crucial residues responsible for the selectivity of ligands. Finally, several derivatives were designed and validated by the QSAR model.2. Some tumor biomarkers for lung cancer that are routinely tested in most hospitals.The lactate dehydrogenase(LDH), C reactive protein(CRP), human fibrinogen(FIB),carcino embryonie antigen(CEA), neuron specific enolase(NSE) and some ions(K、Na、Cl、Ca、Mg、P) were selected in our study. Statistical analysis was performed to analyzethe differences of each biomarker between patients and control group. The ROC curves for discovery the sensitivity/specificity in each index were also introduced to identify which one with high sensitivity. Then, satisfied diagnosis models by GEP were built using these biochemical indice.Results1. Based on structures and activities of 43 potential Mer inhibitors, a optimum model was built that exhibited statistically significant results: the cross-validated correlation coefficient2 q was 0.599, non-cross-validated2 r value was 0.984, and external predictive correlation coefficient 728.02=extr. And the model shows that the hydrophobic, hydrogen bond donor, and hydrogen bond acceptor fields are essential for the inhibition activity of Mer. In the docking processes, we found several key residues that have hydrogen bonding interactions with ligands, such as GLU595, PRO672,MET674, and LYS675. The hydrophobic residues LEU593, ILE650, LEU671, PRO672,MET674 and VAL804 could interact with hydrophobic groups in Mer inhibitors. Around the residues ILE650, LEU671, PRO672, and MET674, a hydrophobic clamp exists that could also be occupied by hydrophobic groups. And there is an important solvent region that may influence the physical and pharmacokinetic(PK) properties. Then, MD simulations and MM/GBSA free energy calculations were employed to determine the dynamic binding processes and compare the binding modes between the inhibitors with different activities. The binding free energies predicted by MM/GBSA are in good agreement with the experimental bioactivities, and some key residues were identified by per-residue energy decomposition that was in accordance with the results of docking.2. CRP, FIB, CEA, Ca and Mg that have significant differences between patients and control group(P<0.05) and high sensitivity were determined by statistical analysis and ROC curves analysis, respectively. Two different combination were determined: one includes all of the 5 indice, and the other one model does not have Ca and Mg. Based on the two combination, two GEP models were built and both of them show very high accuracy rate of diagnosis: 91%(245/270) and 89%(240/270), respectively.ConclusionThe multistep framework combining of some molecular modeling techniques, such as3D-QSAR, molecular docking and molecular dynamics simulation can explore molecular structure features of both the Mer kinase and its inhibtors, and study the interaction mode of them accurately, which aids the designing and synthesis of highly activity inhibitors.GEP is new method that is used in the diagnosis of NSCLC, and has high accuracy rate of diagnosis. With the aid of auxiliary GEP model, the NSCLC patients can receive timely treatment.