CN103886225A - Method for designing proteins on basis of polarizable force fields and pso (particle swarm optimization) - Google Patents

Abstract

The invention relates to a method for designing proteins on the basis of polarizable force fields and pso (particle swarm optimization). The method includes steps of acquiring a plurality of proteins by the aid of a protein design template; optimizing each protein, in other words, searching certain rotamers in a rotamerlib, and creating a temporary particle by the aid of each found rotamer; computing the fitness of each temporary particle, and updating local optimal particles and neighbor optimal particles according to the fitness of each temporary particle; updating positions and speeds of all non-temporary particles; comparing the positions of all the neighbor optimal particles to acquire the global optimal positions; searching side-chain conformation combined spaces by the aid of particle swarms to acquire the optimized proteins. Each certain rotamer is the closest to sub-vectors of the corresponding particle which continuously moves in a 4n (four-dimensional) space. The method has the advantages that discrete problems are solved by the aid of pso, and the polarizable force fields are used as evaluation foundations by the aid of the high efficiency of pso; modification on the proteins can be predicted theoretically, and the proteins can be screened theoretically.

Description

Based on the Design of Protein of the polarizable field of force and pso optimization

Technical field

What the present invention relates to is a kind of method of computer forecast protein function, belongs to biology information technology, computing method and computer virtual reality technology.

Background technology

Area of computer aided protein design is to carry out means important in protein design.But complete sequence protein design has two difficult problems.First difficult problem is exactly shot array.Simply give an example, have 100 amino acid whose polypeptied chains to design for one, the interblock space of sequence is pow(20 so, 100), this numeral is too large, and therefore selecting the optimized algorithm having developed is exactly the problem that must face.Second Problem is that speed and the degree of accuracy of energy computing function is conflicting, is often difficult to get both.If neither one accurately energy function is identified sequence and the conformation of optimization, better optimized algorithm can not play a role.

Optimisation strategy: in whole computation process, suppose that Conformation of the main chain does not change, can reduce greatly the complexity of calculating, also make continuous problem change into discrete problem simultaneously.So just can find reasonable conformation with various conventional optimized algorithms.Generally speaking, in protein design, optimized algorithm is divided into two classes, and first kind algorithm is non-exhaust algorithm, can not guarantee to find Optimum configuration, and this class algorithm comprises MC algorithm, simulated annealing, genetic algorithm.This class algorithm is applicable to search for reasonable conformation in larger search volume conventionally, especially when conformational space is too large to such an extent as to when using exhaust algorithm infeasible.Equations of The Second Kind algorithm can obtain best conformation, and this class algorithm comprises, the method for exhaustion, branch and bound algorithm, DEE algorithm, divide and conquer algorithm.

Mc algorithm: random selection, then selected the rotamer on this site according to roulette at a site in a structure.If the energy after changing has diminished, so just accept current variation, otherwise just according to Metropolis rule, determine whether to accept current variation.Metropolis rule: by temperature, the energy before and after changing, obtains a probability numeral, then by this numeral and a random number comparison, if random number is less than probability numeral, just accepts current change, otherwise does not just accept current change.When making conception search, this rule can jump out minimum point.Can cross energy barrier by adjusting temperature.Or use the mode of simulated annealing progressively to reduce the possibility of crossing energy barrier.

Genetic algorithm: genetic algorithm derives from genetic evolution theory.First produce the population of random series, then select low energy conformations to hybridize, exchange sequence and rotamer.The result of hybridization often can obtain more low energy conformations.Such low energy conformations is selected, and carries out follow-on breeding.Energized conformation has just been abandoned.This process comprises restructuring, and sudden change is selected.Repeat this process, until convergence.

DEE algorithm: DEE algorithm can find globally optimal solution.A key factor of DEE algorithm is exactly filtering rule.Different filtering rules significant difference in efficiency.Using the prerequisite of DEE algorithm is that gross energy is decomposed.Resolve into the energy of the right interaction energy of any side chain and side chain inside, and the energy of side chain and constant part.Gross energy obtains by summation when calculating.

The selection of energy function.Energy function is interactional for describing protein interior.This function is used for the structure quality of assess proteins.The selection of energy function is another important factor in protein design.The molecular force field amber of current main-stream, charrm, the energy function of opls. has Van der Waals conventionally, static, hydrogen bond, bond distance, bond angle, dihedral angle etc.

Summary of the invention

In order to overcome the deficiency of currently available technology, the invention provides a kind of Design of Protein based on particle swarm optimization, the technical solution used in the present invention is: based on the Design of Protein of the polarizable field of force and pso optimization, comprise the following steps:

Obtain several protein by protein design template; Each protein is optimized as follows:

In 4n dimension space, the subvector of each particle of continuous moving is found nearest rotamer in rotamerlib; Then the interim particle of rotamer structure finding with these;

Calculate the fitness of each interim particle, and upgrade local optimum particle and neighbours' optimal particle according to the fitness of these interim particles; And upgrade all non-interim particle position speed; Obtain global optimum position by the position of more all neighbours' optimal particle; Particle group optimizing finishes;

Protein after carrying out side chain conformation interblock space search and be optimized by population.

The subvector of described each particle represents the side chain conformation of one 4 dimension.

The fitness that calculates each interim particle comprises the following steps:

First corresponding protein is write in a character string; Then call the built-in pdbxyz module of tinker and carry out format conversion and obtain the character string of the manageable xyz form of analyze; Call analyze module and obtain the gross energy that this protein is corresponding, this gross energy is fitness.

The described fitness according to these interim particles upgrades local optimum particle and neighbours' optimal particle comprises the following steps:

Each particle has recorded a local optimum position and neighbours' optimal location, can be according to an interim particle tpi of principle generation recently after this particle pi moves, if the fitness of tpi is better than the fitness of the local optimum position of pi record, so just upgrade the record of local optimum; If the fitness of tpi is greater than the fitness of the optimum record of neighbours of pi record, the so just record of new neighbor optimum more.

The described position by more all neighbours' optimal particle obtains global optimum position and is specially by obtaining global optimum position take polarizable field of force calculating fitness as basis, pso algorithm do conformational space search, comprises the following steps:

Initialization bit dot information, field of force information, file path, controls parameter;

Use pso algorithm to control whole side chain conformation search procedure, be specially: scale, the calculation times of first specifying population; Particle is to move according to continuous space in mobile, is to corresponding rotamer in marking by continuous problem discretize; The calculating of each fitness is all used analyze(tinker) energy value that calculates gives a mark; 10 protein conformations that marking value is the highest are global optimum position.

The present invention has following beneficial effect and advantage:

1. the present invention can be used in current newer theory (the polarizable field of force) on protein design.

2. the present invention can predict protein transformation in theory, screens in theory, designs the protein that nature does not evolve out.

3. the present invention can do pharmaceutical grade protein design.

4. the present invention can judge to the current various field of forces in the performance in protein design.The current field of force is a lot, but that field of force is more suitable for protein design field, and this is not also come to a conclusion at present.

5. advantage of the present invention is exactly, and feels free to try and uses pso to process dispersed problem, has utilized pso high efficiency, adopts the polarizable field of force as the basis of evaluating simultaneously.

Accompanying drawing explanation

Fig. 1 is method flow diagram of the present invention;

Fig. 2 is ZDL_Pso algorithm flow chart.

Embodiment

Below in conjunction with embodiment, the present invention is described in further detail.

Content of the present invention is to provide a kind of protein design scheme based on particle swarm optimization.Particle cluster algorithm is efficiently famous with it.But particle cluster algorithm is all good at conventionally in continuous problem.And the problem of protein design discrete problem exactly.We combine the connection attribute of the dispersed problem in protein design and particle cluster algorithm.Particle cluster algorithm can be played a significant role in protein design, for protein design provides a new probing direction.

The basic procedure of the traditional particle cluster algorithm indicating in figure as shown in Figure 1.Particle cluster algorithm derives from bionical---and group's wisdom, flock of birds is looked for food.The mobile space of (bird) particle is continuous space.Each point of continuous space is significant.

The present invention is achieved in that

(1) Pso algorithm ultimate principle:

In D dimension space, there is N particle;

Particle i position: x _i=(x _i1, x _i2, x _i3... x _id), by x _isubstitution fitness function f (x _i) ask adaptive value;

Particle i speed: v _i=(v _i1, v _i2, v _i3..., v _id)

The individual extreme point of particle i position: pbest _i=(p _i1, p _i2, p _i3..., p _id)

Neighbours' extreme point position of population: gbest=(g ₁, g ₂, g ₃..., g _d)

N dimension speed and the more new formula of position of particle i:

v _{in_update}＝w×v _in+c ₁×r ₁×(pbest _in-x _in)+c ₂×r ₂×(gbest _n-x _in)

x _{in_update}＝x _in+v _{in_update}

C ₁, c ₂be the study factor, empirical value is got c ₁=c ₂=2, regulate the maximum step-length of study; r ₁, r ₂-two random numbers, span (0,1), to increase search randomness; W-inertial factor, nonnegative number, regulates the hunting zone to solution space; x _infor the n dimension position numerical value of particle i, v _infor the n dimension speed numerical value of particle i.

(2) discretize of continuous problem in protein design

The transform mode (dispersed problem changes into continuously) that we adopt is as follows: suppose that a protein template has n site can change rotamer, the vector representation that particle use 4n ties up so, the element representation in vector be the dihedral angle of rotamer.4 that start anew continuous are one group and represent all dihedral angles of rotamer (dihedral angle in rotamerlib in side chain has at most four, and all side chains all distribute four storage spaces).

Mobile: particle, in mobile, moves in continuous coordinate space.The mode of tolerance adopts Euclidean distance.

Two n-dimensional vector a=(a ₁, a ₂, a ₃..., a _n) and b=(b ₁, b ₂, b ₃..., b _n) between Euclidean distance:

$d_{\mathrm{ab}}=\sqrt{\underset{k=1}{\overset{n}{\mathrm{\Σ}}}{(a_k-b_k)}^2}$

Marking: marking is given a mark by discrete space.First will separate particle vector (4n) and resolve into n subvector, four dihedral angles of the corresponding residue of each subvector.Then seek the rotamer vector in discrete space according to nearest principle.Each subvector is merged into particle vector by subvector after having determined nearest rotamer vector.Finally calculate energy scoring.Move again, then give a mark ...We adopt molecule entirety potential energy the score value of marking.We adopt current relatively forward position theory---calculating energy is carried out in the polarizable field of force.

The mode that global optimum's point adopts has two kinds: one is to use all particles and current particle to communicate definite optimum point.Another kind is only to use limited information to communicate by letter with current particle.We adopt the second, and benefit is: do like this and more easily jump out minimal value, but the speed of convergence is slow.Speed of convergence and jump out minimal value and can not get both.The related communication particle of each particle is dynamic change.

The use in the polarizable field of force: the polarizable field of force is the focus of current research is comparison forward position.The classical field of force (AMBER, CHARMM, OPLS-AA, GROMOS) all based on point of fixity charge model, as everyone knows, molecule all has very strong Quantum Properties, is difficult to their electrostatic property of accurate description with point of fixity charge model, and this Quantum Properties can change along with the change of surrounding environment-be, and the induced polarization effect of atom and molecule cannot accurately calculate with point of fixity charge model especially.2006, the people's such as He research is found, while utilizing the polarizable field of force to make energy calculation for the different angles conformation of a certain dihedral angle in protein, its result of calculation and quantum chemistry result meet finely, but the result of calculation of AMBER point of fixity charge model and quantum chemistry calculation result exist very large difference.The polarizable field of force is applied in protein design neighborhood by we.In the process of design, we combine the tinker software and the protein design program (c++) that use fortran77 to write.Tinker analyze pdbxyz module is revised accordingly.And the form that is compiled into static library is used to c++.In calculating energy, we call the analyze(revising and call the polarizable field of force).The electrostatic interaction in the polarizable field of force of AMOEBA comprises dipole moment and quadrupole moment, and its potential function comprises into that key interacts and non-bonded interaction (list of references: J.Phys Chem B.2010,114 (8), 2549).The expression-form in its field of force is as follows:

$U=U_{\mathrm{bond}}+U_{\mathrm{angle}}+U_{\mathrm{b\θ}}+U_{\mathrm{oop}}+U_{\mathrm{torsion}}+U_{\mathrm{vdW}}+U_{\mathrm{ele}}^{\mathrm{perm}}+U_{\mathrm{ele}}^{\mathrm{ind}}$

First five items represents respectively bond energy, bond angle energy, bond distance's bond angle mixing energy, from face potential energy, dihedral angle potential energy: the one-tenth key that is short distance interacts; Latter three represent respectively model ylid bloom action energy, electrostatic potential energy, induction potential energy.Wherein become the potential function expression-form of key interaction phase identical with the classical MM3 field of force, its expression-form is as follows:

U _bond＝K _b(b-b ₀) ²[1-2.55(b-b ₀)+3.793125(b-b ₀) ²

U _angle＝K _θ(θ-θ ₀) ²[1-0.014(θ-θ ₀)+5.6×10 ^-5(θ-θ ₀) ²

-7.0×10 ^-7(θ-θ ₀) ³+2.2×10 ^-8(θ-θ ₀) ⁴]

U _bθ＝K _bθ[(b-b ₀)+(b'-b' _θ)](θ-θ ₀)

$U_{\mathrm{torsion}}=\underset{n}{\mathrm{\Σ}}{K}_{\mathrm{n\φ}}[1+\cos (\mathrm{n\φ}\±\mathrm{\δ})]$

U _oop＝K _χχ ²

K _bfor bond energy parameter, b is bond distance, b ₀for balance bond distance, K _θfor bond angle energy parameter, θ is bond angle, bond angle when θ 0 is balance, K _{b θ}be bond distance's bond angle mixing energy parameter, b' is bond distance, b ' ₀for balance bond distance,

bond distance's bond angle mixing energy parameter, the degree that n is multiple development,

for dihedral angle, K _χfor from face potential parameter, the informal dihedral angle of χ, δ is that phase place is relevant.

What vdw interaction adopted in the polarizable field of force of AMOEBA is that buffered14-7 interaction potential represents with following form:

$U_{\mathrm{vdw}}\left(\mathrm{ij}\right)={\mathrm{\ϵ}}_{\mathrm{ij}}{\left(\frac{1.07}{{\mathrm{\ρ}}_{\mathrm{ij}}+0.07}\right)}^7(\frac{1.12}{{\mathrm{\ρ}}_{\mathrm{ij}}^7+0.12}-2)$

Wherein: ε _ijfor Van der Waals gesture well, R _ij ⁰two atom equilibrium distances, R _ijtwo atom equilibrium distances.

But the still description to electrostatic interaction of the difference of the polarizable field of force of AMOEBA and classical field of force maximum is the practice of in having replaced the classical field of force by dipole moment and quadrupole moment, electric charge having been regarded as to fixed charge in the polarizable field of force of AMOEBA.The benefit of doing is like this that the polarization effect of having considered electric charge replaces energy interaction only, and electrostatic potential that can accurate Calculation atom, impact and some other interaction of hydrogen bond.Two interatomic electrostatic interaction expression formulas are:

$U_{\mathrm{ele}}^{\mathrm{perm}}=M_i^t{T}_{\mathrm{ij}}{M}_j$

Wherein, T _ijfor i, the interatomic interaction tensor of j matrix, i, the span of j is determined according to interaction between atoms situation.In the polarizable field of force of AMOEBA, electrostatic interaction item is by added multipole moments item on the basis in the classical field of force:

M _i＝[q _i,d _ix,d _iy,d _iz,Q _ixx,Q _ixy,Q _ixz,Q _iyx,Q _iyy,Q _iyz,Q _izx,Q _izy,Q _izz] ^T

M _irepresent multipole distance, q _irepresent electric charge, d _ixrepresent dipole moment, the d of x direction _iyrepresent dipole moment, the d of y direction _izrepresent the dipole moment of z direction, Q _ixxrepresent four pole spans, the Q of xx direction _{ix γ}represent four pole span Q of xy direction _ixzrepresent four pole span Q of xz direction _iyxrepresent four pole spans, the Q of yx direction _{iy γ}represent four pole span Q of yy direction _iyzrepresent four pole span Q of yz direction _izxrepresent four pole span Q of zx direction _{iz γ}represent four pole spans, the Q of zy direction _izzrepresent four pole spans of zz direction; Wherein, multipole moments comprise electric charge in above formula, dipole moment and quadrupole moment.

Electrostatic interaction expression formula between any two sites is:

Wherein,

x, y, z represents three directions of dipole.

In this, the multipole moments of each atom can obtain by high-caliber quantum chemistry calculation.

Method for designing of the present invention comprises the following steps:

One. obtain several protein by protein design template; Each protein is optimized as follows.Algorithm at first, is specified side chain at random to the site entirety that will replace.Often do an initialization and just obtained a protein molecule completing, be and just obtained m different protein molecule for m time.Each site side chain dihedral angle represents with four-dimensional number, and n site just represent with 4n dimension, and this 4n number is exactly a particle, it and a protein correspondence.

Two. in 4n dimension space, the subvector of each particle of continuous moving is found nearest rotamer in rotamerlib, and wherein, the subvector of each particle represents the side chain conformation of one 4 dimension; Then the interim particle of rotamer structure finding with these;

Three. calculate the fitness of each interim particle, and upgrade local optimum particle and neighbours' optimal particle according to the fitness of these interim particles; And upgrade all non-interim particle position speed; Obtain global optimum position by the position of more all neighbours' optimal particle; Finally judge whether the number of times that calculates fitness reaches setting value (setting according to time restriction and designing requirement), if reached, particle group optimizing finishes, and does not reach, and continues to use pso algorithm search;

Four. finally carry out the protein after the search of side chain conformation interblock space is optimized by population.

In step 3, the fitness that calculates each interim particle comprises the following steps: first corresponding protein is write in a character string; Then call tinker(and be a molecular simulation software of supporting the polarizable field of force) built-in pdbxyz module carries out format conversion and obtains the character string of the manageable xyz form of analyze; Call analyze module and obtain the gross energy that this protein is corresponding, this gross energy is fitness, and energy value more low-protein is more stable, and fitness is just better.In the process of calculating energy, select amoeba(direct) field of force, solvation Model Selection gk model.

Analyze is the module in tinker, for calculating energy; Pdbxyz is also the module in tinker, and function is format conversion, and pdb format conversion is become to xyz form.

In step 3, comprise the following steps according to fitness renewal local optimum particle and neighbours' optimal particle of these interim particles:

Each particle has recorded a local optimum position and neighbours' optimal location, after this particle (pi) is mobile, can produce an interim particle (tpi) according to nearest principle, if the fitness of tpi is better than the fitness of the local optimum position of pi record, so just upgrade the record of local optimum.

Upgrade the optimum record of neighbours of each particle according to the same mode.

In step 3, by calculating fitness take the polarizable field of force as basis, pso algorithm does conformational space search and obtains global optimum position and comprise the following steps:

Set site information, field of force information (amoba), file path, various control parameters ... etc..

The process of whole optimization is used pso algorithm to control: the scale that need to first formulate population.The information such as the number of times calculating, acquiescence is just specified 10000 times.Particle moves according to continuous space in mobile, is to corresponding rotamer in marking by continuous problem discretize.The calculating of each fitness is all used analyze(tinker) energy value that calculates gives a mark.

Program can constantly be upgraded 10 the best protein conformations that search.

After predetermined number of times (evaluation fitness) reaches, program is by 10 best protein conformation outputs, and 10 protein conformations that marking value is the highest are global optimum position (be not real optimal location, be just relatively satisfied with optimal location), optimize and finish.

Above-mentioned program is " the polarizable field of force of pso protein design software ".

The concrete enforcement of this method comprises the following steps:

1. select the single chain protein matter of high-res, as the template of protein design.In pdb website http:// www.rcsb.org/pdb/home/home.doclick the advanced on the right.Add search condition: whether comprise protein: comprise; Whether comprise DNA: do not comprise; Whether comprise RNA: do not comprise; Whether comprise RNA: do not comprise; The number of chain: 1; Resolution scope is greater than 0, is less than 2,

The quantity of Models is greater than 0, is less than 2; Do not comprise part; The length of chain is greater than 50, is less than 100; Similarity is deleted and is selected 30%;

We can obtain search condition by above, and we have obtained pdbidlist.This list has deleted some pdb according to similarity.We to deleting the pdblist of similarity, carry out homology deletion again.Login homology is deleted website http://dunbrack.fccc.edu/PISCES.php; Click Accessthe server to create your own lists, select Cull from your own list of PDB chains at What do you want to do:

Screening conditions are set:

Maximum consistance number percent: 30

Minimum resolution: 0

Maximum resolution: 2

Maximum R-value:0.3

Minimum residue number: 50

Maximum residue number: 300

Skip the protein crystal that non-X ray is resolved

Skip the protein crystal that only has CA

Choice for use entries screening protein

The pdblist obtaining by such screening can carry out protein design, also can the performance in protein design evaluate different position.The mode of evaluating is to use the protein that designs under the different field of forces and the rmsd comparison of crystal structure.Under the prerequisite of identical algorithm and computing scale, look at that protein and crystal structure which field of force designs are more approaching, here comprising two layers of meaning, ground floor implication represents the protein and the similarity of crystal structure protein in sequence that design.Second layer implication is illustrated in the situation that assurance sequence is identical, the difference of the protein designing and crystal structure protein rmsd.

2. read in protein template.Extract the information of protein dihedral angle.According to bond distance's bond angle of opls field of force standard, the template protein of input is carried out to rebuild again.

3. read in rotamerlib

First process backbone dependent rotamerlib.According to the information of the main chain dihedral angle of input file and template protein, by mutable site selectively rotamer be put into rotamerlib2 in new storehouse.This rotamerlib2 is for this input file of this protein.This rotamerlib2 has just become backbone independent rotamerlib.

4. use pso algorithm in conformational space, to search for satisfactory solution.In internal memory, retain 10 conformations of best low energy.10 conformations of this best low energy are constantly upgraded.When searching algorithm finishes, by these 10 pdb outputs.

Pso algorithm is as shown in Figure 2:

First formulate the scale of population and initialization is done in the locus of particle.

Next allows particle move in space, goes to the position of having found.Particle moves according to continuous space in mobile.

In marking, be by continuous problem discretize: because our side chain conformation is all taken out from a storehouse.That is to say that side chain angle is not continuous in our problem.So we need to allow the side chain conformation comprising in particle, find corresponding side chain conformation according to nearest principle in storehouse.The interim particle of side chain structure simultaneously finding with these is evaluated for marking.

Call analyze(tinker) calculate the fitness (energy) of the interim particle that finds.Upgrade local optimum and neighbours' optimal value of each particle (not comprising interim particle).

Judge whether termination condition meets, and just exits if met, otherwise continue search.

Claims (5)

1. the Design of Protein based on the polarizable field of force and pso optimization, is characterized in that comprising the following steps:

Obtain several protein by protein design template; Each protein is optimized as follows:

In 4n dimension space, the subvector of each particle of continuous moving is found nearest rotamer in rotamerlib; Then the interim particle of rotamer structure finding with these;

Calculate the fitness of each interim particle, and upgrade local optimum particle and neighbours' optimal particle according to the fitness of these interim particles; And upgrade all non-interim particle position speed; Obtain global optimum position by the position of more all neighbours' optimal particle; Particle group optimizing finishes;

Protein after carrying out side chain conformation interblock space search and be optimized by population.

2. the Design of Protein based on the polarizable field of force and pso optimization according to claim 1, is characterized in that: the subvector of described each particle represents the side chain conformation of one 4 dimension.

3. the Design of Protein based on the polarizable field of force and pso optimization according to claim 1, is characterized in that: the fitness that calculates each interim particle comprises the following steps:

First corresponding protein is write in a character string; Then call the built-in pdbxyz module of tinker and carry out format conversion and obtain the character string of the manageable xyz form of analyze; Call analyze module and obtain the gross energy that this protein is corresponding, this gross energy is fitness.

4. the Design of Protein based on the polarizable field of force and pso optimization according to claim 1, is characterized in that: the described fitness according to these interim particles upgrades local optimum particle and neighbours' optimal particle comprises the following steps:

Each particle has recorded a local optimum position and neighbours' optimal location, can be according to an interim particle tpi of principle generation recently after this particle pi moves, if the fitness of tpi is better than the fitness of the local optimum position of pi record, so just upgrade the record of local optimum; If the fitness of tpi is greater than the fitness of the optimum record of neighbours of pi record, the so just record of new neighbor optimum more.

5. the Design of Protein based on the polarizable field of force and pso optimization according to claim 1, it is characterized in that: the described position by more all neighbours' optimal particle obtains global optimum position and is specially by obtaining global optimum position take polarizable field of force calculating fitness as basis, pso algorithm do conformational space search, comprises the following steps:

Initialization bit dot information, field of force information, file path, controls parameter;

Use pso algorithm to control whole side chain conformation search procedure, be specially: scale, the calculation times of first specifying population; Particle is to move according to continuous space in mobile, is to corresponding rotamer in marking by continuous problem discretize; The calculating of each fitness is all used analyze(tinker) energy value that calculates gives a mark; 10 protein conformations that marking value is the highest are global optimum position.