CN108804868B - Protein two-stage conformation space optimization method based on dihedral angle entropy value - Google Patents

Abstract

A two-stage protein conformation space optimization method based on dihedral angle entropy values comprises the steps of firstly, initializing a population by utilizing the first stage and the second stage of a Rosetta protocol; in the exploration phase, the third phase of the iterative Rosetta protocol carries out large-range conformation search, and the change of dihedral angle entropy values in each generation of population is used as the standard of phase switching; in the enhancement stage, random local disturbance is carried out on the dihedral angle of the residues in the loop region on the basis of the fourth stage of the Rosetta protocol, dihedral angle energy is introduced to guide conformation updating in combination with the Rosetta score3 energy function, and adverse effects caused by inaccuracy of the energy function are reduced while exploration on the loop region is enhanced. The invention provides a protein two-stage conformation space optimization method based on dihedral angle entropy with higher prediction precision.

Description

Protein two-stage conformation space optimization method based on dihedral angle entropy value

Technical Field

The invention relates to the fields of bioinformatics and computer application, in particular to a protein two-stage conformation space optimization method based on dihedral angle entropy.

Background

Protein molecules are important components for composing all cells and tissues of human body. All important components of the body require the involvement of proteins. The protein has abundant functions and plays an important role in the normal operation of the organism. The three-dimensional structure of a protein determines the function of the protein, and the protein can only be correctly folded into a specific three-dimensional structure to generate a specific biological function. Diseases due to protein misfolding are not uncommon. Therefore, it is necessary to obtain a three-dimensional structure of a protein in order to understand the function of the protein and cure various diseases related to the protein.

Since the end of the twentieth century, the field of life science has developed rapidly, and proteins, which are macromolecules with the widest distribution and the most complex functions in organisms, are particularly and widely concerned and researched. Prediction of the three-dimensional structure of proteins is an important task of bioinformatics. Different proteins have different amino acid sequences, and all proteins are folded on the basis of one-dimensional sequences to form specific three-dimensional structures, and the understanding of the three-dimensional structures of the proteins is the basis for researching the biological functions of the proteins. At present, the three-dimensional structure of the protein is mainly obtained by X-ray crystal diffraction and nuclear magnetic resonance imaging technologies, but the two protein structures obtained by experimental methods have different costs and respective application limitations. Therefore, it is important to predict the three-dimensional structure of a protein by simulating the process of folding the protein from an amino acid sequence into a specific spatial structure by a computer technology in combination with bioinformatics. Predicting the tertiary structure of a protein based on sequence information remains a significant challenge. The technology based on the fragment assembly principle is the most advanced method for solving the problem, but the method has insufficient prediction precision for larger and harder proteins.

Rosetta is a relatively successful protein structure prediction method based on fragment assembly at present, but the method also has limitations. The main technical difficulties at present come from two aspects. On one hand, the conformation sampling capability of the existing protein structure prediction method is not strong enough, and the exploration efficiency of a loop area is extremely low; another aspect is that the energy function is not accurate enough and the low energy conformation does not necessarily correspond to the near-native conformation, which makes it impossible to produce a prediction result with high accuracy using only the energy function as a criterion for the conformational update.

Therefore, the current protein structure prediction methods have disadvantages and need to be improved.

Disclosure of Invention

In order to overcome the defects of the conventional protein structure prediction method, the invention provides a two-stage conformation space optimization method with higher prediction precision based on dihedral angle entropy. Firstly, initializing a population by utilizing a first phase and a second phase of a Rosetta protocol; in the exploration phase, the third phase of the iterative Rosetta protocol carries out large-range conformation search, and the change of dihedral angle entropy values in each generation of population is used as the standard of phase switching; in the enhancement stage, random local perturbation is carried out on the dihedral angle of the residues in the loop region on the basis of the fourth stage of the Rosetta protocol, and dihedral angle energy is introduced to guide conformation update in combination with the Rosetta score3 energy function.

The technical scheme adopted by the invention for solving the technical problems is as follows:

a method for two-stage conformational space optimization of a protein based on dihedral entropy, the method comprising the steps of:

1) inputting sequence information and secondary structure information of a predicted protein;

2) setting parameters: population size NP, maximum number of iterations in exploration phase

Maximum number of iterations in enhancement phase

An entropy convergence threshold α;

3) population initialization: the first and second phases of the Rosetta protocol are iterated to generate an initial population with NP individuals

And let P be { P ═ P₁,P₂,...,P_NP}＝P^init；

4) Calculating the entropy value of the dihedral angle of the population by the following process:

4.1) dividing the dihedral angle range [ -180 °,180 °) into 36 dihedral angle blocks with 10 ° as a reference;

4.2) statistics of the probability that dihedral angles φ and ψ fall on each block for each residue of all conformations in the population

And

wherein i denotes the residue number, i ∈ {1, 2., l }, l denotes the protein sequence length, j denotes the dihedral corner block, j ∈ {1, 2., 36 }; obtaining the distribution probability of dihedral angles of each residue;

4.3) according to the formula

And

calculating the entropy value of dihedral angles phi and psi of each residue;

4.4) according to the formula

Solving a population dihedral angle entropy value;

5) the exploration phase comprises the following processes:

5.1) start iteration, set g₁1, where the number of iterations

5.2) performing Rosetta protocol third-phase assembly on each individual in the population P;

5.3) executing the step 4), and calculating the entropy value of the dihedral angle of the population

5.4) if g₁And if the entropy value is more than 20, judging whether the entropy value is converged, wherein the process is as follows:

5.4.1) according to the formula

Wherein k belongs to {1, 2.,. 20}, and each entropy change of the current generation and the previous 20 generations is calculated;

5.4.2) according to the formula ε_max＝max{ε₁,ε₂,...,ε₂₀Calculating the maximum entropy change;

5.4.3) if ε_max＜α，Ending the exploration phase, and entering the enhancement phase in the step 6); otherwise, executing step 5.5);

5.5)g₁＝g₁+1；

5.6) if

Turning to step 5.2) to execute the next iteration; otherwise, ending the exploration phase and entering the enhancement phase in the step 6);

6) the enhancement stage comprises the following processes:

6.1) start iteration, let g2 be 1, where the number of iterations is

6.2) performing Rosetta protocol fourth phase assembly on each individual in the population P;

6.3) locally perturbing each individual of the population P, as follows:

6.3.1) order P^*＝P_nIn which P is_nRepresents individuals in the population P, and n belongs to {1, 2.., NP };

6.3.2) randomly selecting a residue position r of a loop region;

6.3.3) generating a random decimal rand1, rand1 belongs to [0,1 ];

6.3.4) if rand1 is less than 0.5, execute step 6.3.5); otherwise, go to step 6.3.6)

6.3.5) Pair of conformations P^*The dihedral angle phi of the r-th residue in (b) is locally perturbed as follows:

6.3.5.1) reading conformation P^*The value phi of the dihedral angle phi of the r-th residue in (1);

6.3.5.2) generates a random number rand2, rand2 ∈ [ -5 °,5 ° ];

6.3.5.3)phi＝phi+rand2；

6.3.5.4) will form a conformation P^*Replacing the value of dihedral angle phi for the residue at residue r in (1) with phi;

6.3.6) Pair of conformations P^*The dihedral angle ψ of the r-th residue in (a) is locally perturbed as follows:

6.3.6.1) readTaking conformation P^*The dihedral angle psi value of the r-th residue psi;

6.3.6.2) generates a random number rand3, rand3 ∈ [ -5 °,5 ° ];

6.3.6.3)psi＝psi+rand3；

6.3.6.4) will form a conformation P^*The value of dihedral angle ψ of the r-th residue in (a) is replaced with psi;

6.3.7) if P^*Dihedral angle of medium perturbation and P_nThe corresponding dihedral angles in the block are in the same dihedral angle block, and step 6.3.8) is executed; otherwise, go to step 6.3.9)

6.3.8) computing P with the Rosetta score3 energy function^*And P_nThe energy of (a); deciding whether to use P according to Metropolis criterion^*Replacement of P_nI.e. to determine whether to accept the pair of conformations P_nLocal perturbation of (2);

6.3.9) directed conformational update according to dihedral energy, the procedure is as follows:

6.3.9.1) determining P based on the dihedral distribution probability of the last generation residue in the search stage^*Dihedral angle and P of medium perturbation_nDistribution probability p of dihedral angle block in which corresponding dihedral angle is located^*And p_n；

6.3.9.2) according to formula e^*＝-ln(p^*) And e_n＝-ln(p_n) Calculating dihedral angle energy;

6.3.9.3) determine whether to use P according to Metropolis criteria^*Replacement of P_nI.e. to determine whether to accept the pair of conformations P_nLocal perturbation of (2);

6.4)g₂＝g₂+1；

6.5) if

Skipping to step 6.2) to execute the next iteration; otherwise, ending the enhancement stage; 7) and outputting a prediction result according to a Rosetta protocol.

The invention has the beneficial effects that: guiding conformation search by counting entropy value information of two-sided angles of conformation population residues in the prediction process, and realizing exploration and enhancement switching according to the change of entropy values; randomly selecting a residue dihedral angle of a loop region in the conformation to carry out local disturbance, selecting a Rosetta score3 energy function or dihedral angle energy to guide conformation update according to the distribution of the dihedral angle, and reducing adverse effects brought by inaccurate energy function while enhancing the search for the loop region; thereby improving the accuracy of protein structure prediction.

Drawings

FIG. 1 is a schematic diagram of conformational update in the structural prediction of protein 1ctf by a protein two-stage conformational space optimization method based on dihedral entropy.

FIG. 2 is a three-dimensional structure diagram of protein 1ctf obtained by structure prediction based on dihedral entropy two-stage conformational space optimization method.

Detailed Description

The invention is further described below with reference to the accompanying drawings.

Referring to fig. 1 and2, a two-stage conformation space optimization method of protein based on dihedral angle entropy includes the following steps:

1) inputting sequence information and secondary structure information of a predicted protein;

2) setting parameters: population size NP, maximum number of iterations in exploration phase

Maximum number of iterations in enhancement phase

An entropy convergence threshold α;

3) population initialization: the first and second phases of the Rosetta protocol are iterated to generate an initial population with NP individuals

And let P be { P ═ P₁,P₂,...,P_NP}＝P^init；

4) Calculating the entropy value of the dihedral angle of the population by the following process:

4.1) dividing the dihedral angle range [ -180 °,180 °) into 36 dihedral angle blocks with 10 ° as a reference;

4.2) statistics of the probability that dihedral angles φ and ψ fall on each block for each residue of all conformations in the population

And

wherein i denotes the residue number, i ∈ {1, 2., l }, l denotes the protein sequence length, j denotes the dihedral corner block, j ∈ {1, 2., 36 }; obtaining the distribution probability of dihedral angles of each residue;

4.3) according to the formula

And

calculating the entropy value of dihedral angles phi and psi of each residue;

4.4) according to the formula

Solving a population dihedral angle entropy value;

5) the exploration phase comprises the following processes:

5.1) start iteration, set g₁1, where the number of iterations

5.2) performing Rosetta protocol third-phase assembly on each individual in the population P;

5.3) executing the step 4), and calculating the entropy value of the dihedral angle of the population

5.4) if g₁And if the entropy value is more than 20, judging whether the entropy value is converged, wherein the process is as follows:

5.4.1) according to the formula

Wherein k belongs to {1, 2.,. 20}, and each entropy change of the current generation and the previous 20 generations is calculated;

5.4.2) according to the formula ε_max＝max{ε₁,ε₂,...,ε₂₀Calculating the maximum entropy change;

5.4.3) if ε_maxIf the alpha is less than alpha, ending the exploration stage and entering the step 6) enhancement stage; otherwise, executing step 5.5);

5.5)g₁＝g₁+1；

5.6) if

Turning to step 5.2) to execute the next iteration; otherwise, ending the exploration phase and entering the enhancement phase in the step 6);

6) the enhancement stage comprises the following processes:

6.1) begin iteration, set g₂1, where the number of iterations

6.2) performing Rosetta protocol fourth phase assembly on each individual in the population P;

6.3) locally perturbing each individual of the population P, as follows:

6.3.1) order P^*＝P_nIn which P is_nRepresents individuals in the population P, and n belongs to {1, 2.., NP };

6.3.2) randomly selecting a residue position r of a loop region;

6.3.3) generating a random decimal rand1, rand1 belongs to [0,1 ];

6.3.4) if rand1 is less than 0.5, execute step 6.3.5); otherwise, go to step 6.3.6)

6.3.5) Pair of conformations P^*The dihedral angle phi of the r-th residue in (b) is locally perturbed as follows:

6.3.5.1) reading conformation P^*The value phi of the dihedral angle phi of the r-th residue in (1);

6.3.5.2) generates a random number rand2, rand2 ∈ [ -5 °,5 ° ];

6.3.5.3)phi＝phi+rand2；

6.3.5.4) will form a conformation P^*Replacing the value of dihedral angle phi for the residue at residue r in (1) with phi;

6.3.6) Pair of conformations P^*The dihedral angle ψ of the r-th residue in (a) is locally perturbed as follows:

6.3.6.1) reading conformation P^*The dihedral angle psi value of the r-th residue psi;

6.3.6.2) generates a random number rand3, rand3 ∈ [ -5 °,5 ° ];

6.3.6.3)psi＝psi+rand3；

6.3.6.4) will form a conformation P^*The value of dihedral angle ψ of the r-th residue in (a) is replaced with psi;

6.3.7) if P^*Dihedral angle of medium perturbation and P_nThe corresponding dihedral angles in the block are in the same dihedral angle block, and step 6.3.8) is executed; otherwise, go to step 6.3.9)

6.3.8) computing P with the Rosetta score3 energy function^*And P_nThe energy of (a); deciding whether to use P according to Metropolis criterion^*Replacement of P_nI.e. to determine whether to accept the pair of conformations P_nLocal perturbation of (2);

6.3.9) directed conformational update according to dihedral energy, the procedure is as follows:

6.3.9.1) determining P based on the dihedral distribution probability of the last generation residue in the search stage^*Dihedral angle and P of medium perturbation_nDistribution probability p of dihedral angle block in which corresponding dihedral angle is located^*And p_n；

6.3.9.2) according to formula e^*＝-ln(p^*) And e_n＝-ln(p_n) Calculating dihedral angle energy;

6.3.9.3) determine whether to use P according to Metropolis criteria^*Replacement of P_nI.e. to determine whether to accept the pair of conformations P_nLocal perturbation of (2);

6.4)g₂＝g₂+1；

6.5) if

Jump toTo step 6.2) executing next iteration; otherwise, ending the enhancement stage; 7) and outputting a prediction result according to a Rosetta protocol.

In this embodiment, taking protein 1ctf with a sequence length of 68 as an example, a two-stage conformational space optimization method of protein based on dihedral entropy comprises the following steps:

1) inputting sequence information and secondary structure information of protein 1 ctf;

2) setting parameters: population size NP is 100, maximum number of iterations in exploration phase

Maximum number of iterations in enhancement phase

The entropy convergence threshold α is 0.005;

3) population initialization: the first and second phases of the Rosetta protocol are iterated to generate an initial population with NP individuals

And let P be { P ═ P₁,P₂,...,P_NP}＝P^init；

4) Calculating the entropy value of the dihedral angle of the population by the following process:

4.1) dividing the dihedral angle range [ -180 °,180 °) into 36 dihedral angle blocks with 10 ° as a reference;

4.2) statistics of the probability that dihedral angles φ and ψ fall on each block for each residue of all conformations in the population

And

wherein i denotes the residue number, i ∈ {1, 2., l }, l denotes the protein sequence length, j denotes the dihedral corner block, j ∈ {1, 2., 36 }; obtaining the distribution probability of dihedral angles of each residue;

4.3) according to the formula

And

calculating the entropy value of dihedral angles phi and psi of each residue;

4.4) according to the formula

Solving a population dihedral angle entropy value;

5) the exploration phase comprises the following processes:

5.1) start iteration, set g₁1, where the number of iterations

5.2) performing Rosetta protocol third-phase assembly on each individual in the population P;

5.3) executing the step 4), and calculating the entropy value of the dihedral angle of the population

5.4) if g₁And if the entropy value is more than 20, judging whether the entropy value is converged, wherein the process is as follows:

5.4.1) according to the formula

Wherein k belongs to {1, 2.,. 20}, and each entropy change of the current generation and the previous 20 generations is calculated;

5.4.2) according to the formula ε_max＝max{ε₁,ε₂,...,ε₂₀Calculating the maximum entropy change;

5.4.3) if ε_maxIf the alpha is less than alpha, ending the exploration stage and entering the step 6) enhancement stage; otherwise, executing step 5.5);

5.5)g₁＝g₁+1；

5.6) if

Go to the step5.2) executing the next iteration; otherwise, ending the exploration phase and entering the enhancement phase in the step 6);

6) the enhancement stage comprises the following processes:

6.1) begin iteration, set g₂1, where the number of iterations

6.2) performing Rosetta protocol fourth phase assembly on each individual in the population P;

6.3) locally perturbing each individual of the population P, as follows:

6.3.1) order P^*＝P_nIn which P is_nRepresents individuals in the population P, and n belongs to {1, 2.., NP };

6.3.2) randomly selecting a residue position r of a loop region;

6.3.3) generating a random decimal rand1, rand1 belongs to [0,1 ];

6.3.4) if rand1 is less than 0.5, execute step 6.3.5); otherwise, go to step 6.3.6)

6.3.5) Pair of conformations P^*The dihedral angle phi of the r-th residue in (b) is locally perturbed as follows:

6.3.5.1) reading conformation P^*The value phi of the dihedral angle phi of the r-th residue in (1);

6.3.5.2) generates a random number rand2, rand2 ∈ [ -5 °,5 ° ];

6.3.5.3)phi＝phi+rand2；

6.3.5.4) will form a conformation P^*Replacing the value of dihedral angle phi for the residue at residue r in (1) with phi;

6.3.6) Pair of conformations P^*The dihedral angle ψ of the r-th residue in (a) is locally perturbed as follows:

6.3.6.1) reading conformation P^*The dihedral angle psi value of the r-th residue psi;

6.3.6.2) generates a random number rand3, rand3 ∈ [ -5 °,5 ° ];

6.3.6.3)psi＝psi+rand3；

6.3.6.4) will form a conformation P^*Substitution of the value of dihedral angle ψ for the r-th residue in (1)Is psi;

6.3.7) if P^*Dihedral angle of medium perturbation and P_nThe corresponding dihedral angles in the block are in the same dihedral angle block, and step 6.3.8) is executed; otherwise, go to step 6.3.9)

6.3.8) computing P with the Rosetta score3 energy function^*And P_nThe energy of (a); deciding whether to use P according to Metropolis criterion^*Replacement of P_nI.e. to determine whether to accept the pair of conformations P_nLocal perturbation of (2);

6.3.9) directed conformational update according to dihedral energy, the procedure is as follows:

6.3.9.1) determining P based on the dihedral distribution probability of the last generation residue in the search stage^*Dihedral angle and P of medium perturbation_nDistribution probability p of dihedral angle block in which corresponding dihedral angle is located^*And p_n；

6.3.9.2) according to formula e^*＝-ln(p^*) And e_n＝-ln(p_n) Calculating dihedral angle energy;

6.3.9.3) determine whether to use P according to Metropolis criteria^*Replacement of P_nI.e. to determine whether to accept the pair of conformations P_nLocal perturbation of (2);

6.4)g₂＝g₂+1；

6.5) if

Skipping to step 6.2) to execute the next iteration; otherwise, ending the enhancement stage; 7) and outputting a prediction result according to a Rosetta protocol.

Taking the protein 1ctf with the amino acid sequence length of 68 as an example, the near-native conformation of the protein is obtained by the method, and the conformation renewal scheme is shown in FIG. 1; minimum root mean square deviation of

The prediction structure is shown in fig. 2.

The foregoing is a predictive description of the invention as embodied in one embodiment, and it will be apparent that the invention is not limited to the embodiment described above, but may be embodied with various modifications without departing from the basic inventive concept and without departing from the spirit thereof.

Claims (1)

1. A protein two-stage conformation space optimization method based on dihedral angle entropy is characterized in that: the two-stage conformation space optimization method of the protein comprises the following steps:

1) inputting sequence information and secondary structure information of a predicted protein;

2) setting parameters: population size NP, maximum number of iterations in exploration phase

Maximum number of iterations in enhancement phase

An entropy convergence threshold α;

3) population initialization: the first and second phases of the Rosetta protocol are iterated to generate an initial population with NP individuals

And let P be { P ═ P₁,P₂,...,P_NP}＝P^init；

4) Calculating the entropy value of the dihedral angle of the population by the following process:

4.1) dividing the dihedral angle range [ -180 °,180 °) into 36 dihedral angle blocks with 10 ° as a reference;

4.2) statistics of the probability that dihedral angles φ and ψ fall on each block for each residue of all conformations in the population

And

wherein i denotes the residue number, i ∈ {1, 2., l }, l denotes the protein sequence length, j denotes the dihedral corner block, j ∈ {1, 2., 36 }; obtaining the distribution probability of dihedral angle of each residue；

4.3) according to the formula

And

calculating the entropy value of dihedral angles phi and psi of each residue;

4.4) according to the formula

Solving a population dihedral angle entropy value;

5) the exploration phase comprises the following processes:

5.1) start iteration, set g₁1, where the number of iterations

5.2) performing Rosetta protocol third-phase assembly on each individual in the population P;

5.3) executing the step 4), and calculating the entropy value of the dihedral angle of the population

5.4) if g₁And if the entropy value is more than 20, judging whether the entropy value is converged, wherein the process is as follows:

5.4.1) according to the formula

Wherein k belongs to {1, 2.,. 20}, and each entropy change of the current generation and the previous 20 generations is calculated;

5.4.2) according to the formula ε_max＝max{ε₁,ε₂,...,ε₂₀Calculating the maximum entropy change;

5.4.3) if ε_maxIf the alpha is less than alpha, ending the exploration stage and entering the step 6) enhancement stage; otherwise, executing step 5.5);

5.5)g₁＝g₁+1；

5.6) if

Turning to step 5.2) to execute the next iteration; otherwise, ending the exploration phase and entering the enhancement phase in the step 6);

6) the enhancement stage comprises the following processes:

6.1) begin iteration, set g₂1, where the number of iterations

6.2) performing Rosetta protocol fourth phase assembly on each individual in the population P;

6.3) locally perturbing each individual of the population P, as follows:

6.3.1) order P^*＝P_nIn which P is_nRepresents individuals in the population P, and n belongs to {1, 2.., NP };

6.3.2) randomly selecting a residue position r of a loop region;

6.3.3) generating a random decimal rand1, rand1 belongs to [0,1 ];

6.3.4) if rand1 is less than 0.5, execute step 6.3.5); otherwise, go to step 6.3.6)

6.3.5) Pair of conformations P^*The dihedral angle phi of the r-th residue in (b) is locally perturbed as follows:

6.3.5.1) reading conformation P^*The value phi of the dihedral angle phi of the r-th residue in (1);

6.3.5.2) generates a random number rand2, rand2 ∈ [ -5 °,5 ° ];

6.3.5.3)phi＝phi+rand2；

6.3.5.4) will form a conformation P^*Replacing the value of dihedral angle phi for the residue at residue r in (1) with phi;

6.3.6) Pair of conformations P^*The dihedral angle ψ of the r-th residue in (a) is locally perturbed as follows:

6.3.6.1) reading conformation P^*The dihedral angle psi value of the r-th residue psi;

6.3.6.2) generates a random number rand3, rand3 ∈ [ -5 °,5 ° ];

6.3.6.3)psi＝psi+rand3；

6.3.6.4) will form a conformation P^*The value of dihedral angle ψ of the r-th residue in (a) is replaced with psi;

6.3.7) if P^*Dihedral angle of medium perturbation and P_nThe corresponding dihedral angles in the block are in the same dihedral angle block, and step 6.3.8) is executed; otherwise, go to step 6.3.9)

6.3.8) computing P with the Rosetta score3 energy function^*And P_nThe energy of (a); deciding whether to use P according to Metropolis criterion^*Replacement of P_nI.e. to determine whether to accept the pair of conformations P_nLocal perturbation of (2);

6.3.9) directed conformational update according to dihedral energy, the procedure is as follows:

6.3.9.1) determining P based on the dihedral distribution probability of the last generation residue in the search stage^*Dihedral angle and P of medium perturbation_nDistribution probability p of dihedral angle block in which corresponding dihedral angle is located^*And p_n；

6.3.9.2) according to formula e^*＝-ln(p^*) And e_n＝-ln(p_n) Calculating dihedral angle energy;

6.3.9.3) determine whether to use P according to Metropolis criteria^*Replacement of P_nI.e. to determine whether to accept the pair of conformations P_nLocal perturbation of (2);

6.4)g₂＝g₂+1；

6.5) if

Skipping to step 6.2) to execute the next iteration; otherwise, the enhancement phase is ended

7) And outputting a prediction result according to a Rosetta protocol.

Images