| Due to the rapid progress of computer technology and molecular simulation technology,molecular modeling has permeated into many different fields of life science and achieved remarkable achievements in numerous researches.Since the complex nature of biological systems,experimental methods face great challenges investigating molecule mechanism in life,which exceeds the physical limit of experiment.On the other hand,computer assistant molecular modeling technology could comprehensively resolve the molecule level mechanism of living organism.As the foundation of life activities,the function and structure investigations of enzymes demands the help from molecular modeling particularly.Associated with with experimental measures,molecular modeling has promoted a new stage of life science research.In bio-catalysis research field,enzymes are used as effective catalysts for chemical reactions.Therefore,it demands vast experiment tests looking for suitable enzymes.The catalytic performance of enzymes is vital for their application in industry.However,the traditional experimental methods could not describe the molecular level mechanism of enzyme's catalytic selectivity,therefore the only way to locate feasible substrates is to take parallel tests on wide-range different compounds.Since,the specific binding between enzyme and substrates is the key of enzyme's catalytic selectivity,which includes many different interactions such as hydrogen bond,Van der Waals interactions,electrostatic interactions,solvation energy and etc.,molecular simulation is the only valid measure to analysis these atomic level interactions.Although there are many computation algorithms and programs that can take simulations on enzyme systems,none of them meet the requirement of guiding daily experiment practically.This paper has focused on this important scientific problem of enzyme's catalytic selectivity,trying to create novel algorithms and programs to work out a new way for enzyme development.The study of this paper could be divided into two major parts.In the first part,an algorithm is designed to predict feasible substrates for enzyme,which could quickly filter potential substrates from database of chemical compounds.In the second part,the focus had been shifted to discover the potential of enzymes that is finding even more substrates in a fragment-based chemical space by a novel algorithm.Both parts are hoping to serve the future of enzyme research by introducing applicable methods and tools.The first part of current paper starts from analyzing of mechanism of enzyme catalysis.Based on a NAC(Near attack conformation)model and molecular docking technology,a universal workflow for prediction of substrates of enzyme has been designed and coded into a program called Themis.Candida Antarctic lipase B was used in the verification step,which is an enzyme with plenty of academic investigations and broad commercial applications in food,pharmaceutical,chemical and energy industries.CALB catalyzes the hydrolysis of ester or its corresponding reverse reaction,in which CALB shows highly substrate selectivity,region-selectivity and stereo-selectivity.CALB has rich literature experimental data and clear mechanism,which makes it suitable for the verification of our method.An in depth analysis shows that there is a direct relationship between the activity of substrates and their interaction model with CALB active site.By summarizing the interaction mode between enzyme and substrate,we designed a score function that could distinguish substrate and non-substrate for CALB.The score function is called CALB-SP score,which is used in the Themis program.In a test of nighty-four compounds,our algorithm could predict real substrate with a true positive rate as high as 92.7%and the speed of calculation reaches over a thousand compounds per day.In the validation of our method with a nitrilase the accuracy has also reached 91.7%.In the second part,we created a novel protocol called de novo substrate design;a method could extensively investigate potential substrate structures based on structure and mechanism of enzyme.Referring to the fragment-based drug discovery methods,we build up the algorithm called Crius.Metropolis Monte Carlo(MC)sampling strategy was implemented and a molecular mechanics(MM)free energy based score function was applied in the Crius.Our algorithm provides ability of in depth search of potential substrates without any chemical database.It's mainly written in Python language integrating several third part programs.Using MPI parallel programing model,our algorithm can be accelerated by over 512 CPUs with near linear acceleration.To demonstrate the utility of the algorithm,Crius is used to perform de novo substrate design upon three different enzymes separately.First,our algorithm is tested by reproducing experimental verified aldehyde substrates of an aldo-keto reductase from Gluconobacter oxydans(Gox0644)and nine out of sixteen substrates were successfully reproduced.Then,a nitrilase from Syechocystis sp.PCC6803(Nit6803)was used and eighteen out of thirty-one known nitrile substrates were successfully acquired by our in silico approach.At last,in order to prove the applicability of our algorithm,an extensive test was taken by exploring acetate ester substrates of lipase B from Candida Antarctica(CALB)with Crius.Eight substrates were conformed active by literatures.Another twelve novel substrates without any previous report were verified by experiment as expected. |
PDF Full Text Request