De Novo Drug Design Of Caspase-6 Inhibitors And Molecular Simulation Researches

Cysteine-aspartic proteases(Caspases),also known as apoptotic protease Mch-2,is a cysteinylaspartate-specific protease that plays essential roles in cell regulatory of apoptosis and inflammation.Accumulated evidences suggested that caspase-6participates in a series of neurodegenerative pathways by cleaving neuronal substrates,e.g.,microtubule-associated protein Tau,amyloid precursorprotein(APP).Thus,caspase-6 has aroused widespread attentions as a promising molecular target for the treatment of neurodegeneration.Over the last decade,deep learning(DL)technologies have been gradually applied in drug design,and is proven to be promising approaches for artificial intelligence-based drug design.How to adopt DL methods as well as traditional drug design and molecular simulation techniques to guide drug design has become an urgent demand for caspase-6 inhibitor development.Herein,we focused on the molecular characterization and novel molecular design of caspase-6 competitive inhibitors,and illustrate the molecular mechanism of caspase-6 conformational changes between the non-canonical and canonical conformations.In general,this work can provide valuable references for the development of caspase-6 inhibitor and mechanism researches.The main results and conclusions are summarized as follows:(1)Emerging chemical pattern applied to discriminate the caspase-6 inhibitorsEmerging chemical pattern(ECP)was used to construct the predictive models of566 caspase-6 competitive inhibitors and 491 non-inhibitors.The results showed that the optimal ECP model with only 4 Volsurf descriptors exhibits good predictive performances,of which the accuracy,sensitivity and specificity for 422 test samples are74%,74% and 75%,respectively.The 4 Volsurf descriptors derived by ECP model are molecular hydrophilic regions(W4-O::),elongation(Elon),hydrophobic regions(D6-DRY)and capacity factors(Cw4-OH2).Furthermore,ECP-based hierarchical cluster analysis showed that the W4-O:: is the most important feature for distinguishing caspase-6 inhibitors;and the caspase-6 inhibitor had larger hydrophilic regions and more significant hydrogen bonding features when comparing with non-inhibitors.(2)de nove drug design and virtual screening of caspase-6 competitive inhibitorsBased on 2,393,029 chemical molecules and 433 know caspase-6 competitive inhibitors collected from Pub Chem database,recurrent neural network(RNN)-based molecular generator and a machine learning(ML)-based classifier were established for the de novo drug design of caspase-6 competitive inhibitors.The results showed RNN models can accurately learn the SMILES grammars of 2,393,029 chemical molecules,including ionic and isomeric compounds,and can generate novel potential caspase-6inhibitors with similar chemical space after transfer learning of the known 433caspase-6 inhibitors.After transfer learning,the valid percentages of of 128 SMILES strings sampled by RNN model is up to 0.99.After 50,000 randomly sampling,the recall value of the 144 independent test caspase-6 inhibitors is 13.19%.Based on the molecules generated by the RNN models,five ligand-based ML together with the receptor-based docking methods,were employed for screening the potential caspase-6inhibitors.The results showed that the obtained potential caspase-6 inhibitors are mainly generated by substituent modification,scaffold hopping,and chiral transformation,etc.,from the known inhibitors on the level of SMILES stings.Then,virtual screening was performed by using logistic regression(LR)molecular docking,ECP model,and ADME properties prediction.Seven potential caspase-6 inhibitors were finally selected as the most promising candidates for the further research.Binding free-energy analysis showed that molecules 5129 and 5804 have the highest affinities for caspase-6,of which the predicted binding free-energies are-15.34±5.08 and-16.75±3.35 kcal/mol,respectively.In general,the framework presented in this paper provides an efficient combinational strategy for de novo molecular design of caspase-6inhibitors.(3)Targeted molecular dynamics reveals caspase-6 conformational transitionHerein,taking advantages of crystallized caspase-6 structures between non-canonical and canonical states,perturbation-response scanning(PRS)combined with targeted molecular dynamics(TMD)simulations was used to elucidate the conformational transition of caspase-6.On the basis of MD simulations,a total of 22 residues were identified as crucial residues in caspase-6 conformational transition process by using PRS,which agree well with the functional residues reported previously.Then,TMD simulation was performed to explore the conformational transition process of caspase-6.The results showed that the conformational transition of caspase-6 is mainly driven by the structural rearrangement of the active site as well as the 60’s and 130’s extended helices,which results in the inward rotation of 90’s helix.This work provides a dynamic perspective for the conformational transition of caspase-6 from the non-canonical to the canonical states.However,the conformational transition of the biomolecule is an extremely complicated process,and the biased TMD simulations only characterize some representative features in the conformational transition process.The precise mechanism and details require validated by exploring free-energy landscapes.(4)Free-energy landscape of caspase-6 conformational transitionHerein,well-tempered metadynamics(WT-Mt D)simulations were further used to explore the conformational transition pathway and free-energy landscape of caspase-6at the microsecond-scale level.The results showed that the L1-L4 loops,90’s and 130’s helices of caspase-6 have higher structural flexibility,indicating that the above regions are the key regions for the conformational transition of caspase-6.Along the caspase-6conformational transition,four transition-state conformations were obtained.The transition-state VII can reproduce the canonical caspase-6 very well,of which the root-mean-square deviation(RMSD)for the canonical caspase-6 is 2.37 (?).Further analysis revealed that the maximum free-energy barrier is about 10 kcal/mol,which is mainly contributed to the formation of β-sheet of the 130’s helix.Dictionary of secondary structure of proteins(DSSP)analysis showed that the structural rearrangement of the top of the 130’s helix is mainly driven by the complicated hydrogen-bonds network between L1,90’s and 130’s helices.In conclusion,the de novo drug design strategy and the mechanism study of caspase-6 proposed by this thesis may provide valuable references for the drug development in neurodegenerative diseases such as Alzheimer’s disease.