Overview

In this model page, we introduce whole of the modeling work we do in our project.And it is divided into three parts.

To understand the mechanism of peptide binding to metal ions, we performed machine learning-assisted structure optimization. On this basis, we calculated the binding energies of metal ions to peptides and analyzed the selectivity of peptides.Finally , we use the AlphaFold3 to simulate the structure of the peptides polymers to evaluate their potential for self-assembly.

Structural calculations of metal ion-bound peptides bound to metal ions

Metal ion binding to peptides is an ion trapping realized through heteroatoms in the peptide, such as elements like N, O, and S, which form ligand bonds with the metal ion or through weak molecular inter-ion interactions. Metal ion binding to peptides can be studied more intuitively by means of quantum chemical calculations and molecular dynamics simulations, which are expected to reveal the nature of this process.

Machine learning algorithms can save the time for energy calculations in quantum chemical calculations and are suitable for larger systems, here we use the MLatom calculation program of the XACS platform for our calculations.^[1-4]

We will show the process of optimization of the molecular geometry and use it as an analogy to the binding of peptides to ions derived from molecular dynamics simulations. The kinetic study reflects the evolution of the system along the potential energy surface over time, and with the addition of adjustable time steps compared to geometric optimization, it can show the motion of the molecules under the uniform passage of time, and can therefore be called a simulation, as shown in Fig. 2. Geometric optimization, on the other hand, focuses on computationally solving for the minima of the potential energy surface without taking into account the variable of time, and the changes can be inhomogeneous, as shown in Fig. 3. Since both geometric optimization and molecular mechanics (or molecular dynamics simulations) involve calculations of energy and force, and they are optimization and molecular mechanics (or molecular dynamics simulation) both involve the calculation of energy, force, and they are the same molecule's geometric structure changes, these two processes are approximated in the same potential energy surface of the process, so the process of optimizing the molecular geometry can also reflect the molecular dynamics characteristics.

Fig. 1 Potential energy surfaces and calculations

According to the comparison in Figure 2 and Figure 3, it can be seen that the process of geometric optimization is still quite different from the dynamic simulation process, but the overall trend is similar. Although the last arrow in Figure 3 indicates a different trend in the process compared to the kinetics, it has a similar effect when the vibrations and rotations of the molecules are considered in the dynamics process.

In the follow-up video: the white ball is H, the gray ball is C, the blue ball is N, the red ball is O, the flesh-colored ball is Co, the purple ball is Li, the light green ball is Cl, and the dark green ball is Ni.

CBP1

(1) Peptide structure optimization

Here we performed machine learning quantum chemical calculations for the sequence CBP1: DAHKSEVAHRFK. The geometry of the polypeptide chain was first optimized to obtain the folded conformation, as shown below.

(2) Structural analysis of individual ion binding

A Co²⁺ atom was added to the result in (1) to simulate the binding of the folded peptide to the ion. Through structural optimization, we obtained the following results:

It can be seen that the atoms coordinating with Co²⁺ are mainly the carbonyl O of the polypeptide chain and the N atom of the amine.The N=C double bond may also be involved in the coordination.

(3) Binding of multiple ions

We designed this metal-binding peptide with the hope that they would capture multiple metal ions and increase efficiency. In order to analyse the binding of multiple Co²⁺ ions, we added 4 Co²⁺ ions to the molecular model, and to balance their charge we added 4 Cl^- ions for a total system charge of +4. Through structural optimisation, we obtained the following results:

It can be seen that the N and carbonyl O atoms of the senior amine are involved in the coordination, as well as the counterbalance ion Cl^-, and the electrostatic interaction of the counterbalance ion with the H atoms on the polypeptide also plays a role in stabilising the system. Meanwhile, it can be observed that there are still more N and carbonyl O atoms of the senior amine exist as ligand active sites, which can bind more Co²⁺ ions.

CBP2

(1) Peptide structure optimization

Here we performed machine learning quantum chemical calculations for the sequence CBP2:CTQMLGQLCGGG. The geometry of the polypeptide chain was first optimised to obtain the folded conformation as follows.

(2) Structural analysis of individual ion binding

Through structural optimisation, we obtain the following results:

It can be seen that the atoms coordinating with Co²⁺ are mainly S atoms in the polypeptide chain, with O and N atoms also involved in the coordination.

(3) Binding of multiple ions

We designed this metal-binding peptide with the hope that they would capture multiple metal ions and increase efficiency. In order to analyse the binding of multiple Co²⁺ ions, we added 3 Co²⁺ ions to the molecular model, and to balance their charge we added 4 Cl^- ions for a total system charge of +2. Through structural optimisation, we obtained the following results:

It can be seen that the counterbalance ion Cl^- is also involved in the coordination, and the three Co²⁺ form clusters through Cl^- bridge bonding under the encapsulation of the polypeptide, and the counterbalance ion also plays a role in stabilising the system.

LBP3

(1) Peptide structure optimization

Here we performed machine learning quantum chemical calculations for the sequence LBP3:GPGDPGPGDPGPGDPGPGDP. The geometry of the peptide chain was first optimised to obtain the folded conformation as follows.

(2) Structural analysis of individual ion binding

We simulated the binding of a single lithium ion to a folded polypeptide chain. Through structural optimisation, we obtained the following results:

It can be seen that the atoms coordinated to Li⁺ are mainly O and N atoms in the polypeptide chain. Li⁺ ions are encapsulated.

(3) Binding of multiple ions

We designed this metal-binding peptide with the hope that they could capture multiple metal ions and increase efficiency. In order to analyse the binding of multiple Li⁺ ions, we added 4 Li⁺ ions to the molecular model, and in order to balance their charge, we added 4 Cl^- ions, and the total charge of the system was 0. Through structural optimisation, we obtained the following results:

The coordination number of Li is low and no clusters are formed. It can be seen that the counter ion Cl^- is less involved in coordination, and the counter ion is also bound to the carbon skeleton, which can adjust the charge distribution and adjust the structure of the polypeptide.

NBP1

(1) Peptide structure optimization

Here, we perform machine learning quantum chemistry calculations for the sequence NBP1:GLHTWATNLYHM. Firstly, the geometric structure of the polypeptide chain was optimized to obtain the folded configuration, as shown below.

(2) Structural analysis of individual ion binding

We simulated the binding of a single lithium ion to a folded polypeptide chain. Through structural optimization, we get the following results:

It can be seen that the atoms coordinated to Ni²⁺ are mainly carbonyl O atoms in the polypeptide chain. the Ni²⁺ ion is encapsulated.

(3) Binding of multiple ions

We designed this metal-binding peptide with the hope that they would capture multiple metal ions and increase efficiency. In order to analyse the binding of multiple Ni²⁺ ions, we added 3 Ni²⁺ ions to the molecular model, and to balance their charge we added 4 Cl^- ions for a total system charge of +2. Through structural optimisation, we obtained the following results:

It can be seen that the antagonist ion Cl^- is involved in the coordination, and the antagonist ion also binds to the carbon backbone, which can modulate the charge distribution and regulate the structure of the polypeptide.The 2 Ni²⁺ are connected by O-bridge bonding in the polypeptide wrapping, with the possibility of forming clusters in other cases.

NBP2

(1) Peptide structure optimization

Here we performed machine learning quantum chemical calculations for the sequence NBP2:HAVSPTLPAYSK. The geometry of the peptide chain was first optimised to obtain the folded conformation as follows.

(2) Structural analysis of individual ion binding

We simulated the binding of a single Ni ion to a folded polypeptide chain. Through structural optimisation, we obtained the following results:

It can be seen that the atoms coordinated to Ni²⁺ are mainly O and N atom atoms in the polypeptide chain.Ni²⁺ ions are encapsulated.

(3) Binding of multiple ions

We designed this metal-binding peptide with the hope that they would capture multiple metal ions and increase efficiency. In order to analyse the binding of multiple Ni²⁺ ions, we added 4 Ni²⁺ ions to the molecular model, and to balance their charge we added 4 Cl^- ions for a total system charge of +4. Through structural optimisation, we obtained the following results:

It can be seen that the High-grade amines' N and carbonyl O atoms are involved in the coordination, and the counterbalance ion Cl^- is involved in the coordination, and the counterbalance ion also interacts with the carbon skeleton, which can regulate the overall charge distribution of the polypeptide chain, regulate the structure of the polypeptide, and stabilise the system.

NBP4

(1) Peptide structure optimization

Here we performed machine learning quantum chemical calculations for the sequence NBP4: CNAKHHPRCGGG. The geometry of the peptide chain was first optimised to obtain the folded conformation as follows.

(2) Structural analysis of individual ion binding

We simulated the binding of a single Ni²⁺ ion to a folded polypeptide chain. Through structural optimisation, we obtained the following results:

It can be seen that the atoms coordinated to Ni²⁺ are mainly O and N atom atoms in the polypeptide chain. Ni²⁺ ions are encapsulated.

(3) Binding of multiple ions

We designed this metal-binding peptide with the hope that they would capture multiple metal ions and increase efficiency. In order to analyse the binding of multiple Ni²⁺ ions, we added 4 Ni²⁺ ions to the molecular model, and to balance their charge we added 4 Cl^- ions for a total system charge of +4. Through structural optimisation, we obtained the following results:

It can be seen that the higher amines' N and carbonyl O are involved in the coordination, the sulfhydryl S atoms are involved in the coordination, and the counterbalance ion Cl^- is involved in the coordination, and the counterbalance ion also interacts with the carbon skeleton, which can regulate the overall charge distribution of the polypeptide chain, regulate the structure of the polypeptide, and stabilise the system.

In the future, our project will go deeper, and we will also further optimize our computational input structure model and analyze a larger number of simulation results to find patterns and predict the optimal conditions for protein-metal ion binding, such as pH, temperature, and ion concentration, to provide a more precise guide for experimental design.

Calculation of the binding energy of metal ion-bound peptides

In order to analyze the binding capacity of the metal-binding peptide LBP3: GPGDPGPGDPGPGDPGPGDPGPGDP to the metal ion Li⁺, we carried out calculations of the binding energies of the ions to the peptide, which were used to analyze the number of bound ions by comparing them with the experimentally measured values. In order to derive the selectivity of our designed metal ion binding peptide, Prof. Shilu Chen suggested that we change the ions bound by LBP3 in situ to Ni²⁺, Co²⁺ , and Mn²⁺ as the input file to compare their binding abilities. Here we used the XACS platform for structure optimization and energy calculations.

Ionic and peptide structure modeling

Since Ni²⁺, Co²⁺, and Mn²⁺, a class of excess metal ions, are usually bound in a 6-coordinated manner, we constructed the pre-binding structure with the 6-coordinated form M²⁺(H₂O)₆ as shown in Fig. 1. Correspondingly, 6 H₂O molecules need to be added to the post-binding structure.

Fig. 2 M²⁺(H₂O)₆ and combined structure

The coordination of Li⁺ atoms to transition metals varies considerably, and here we chose a structure such as Li⁺(H₂O)₂, with which two H₂O molecules need to be added to the combined structure.

Fig. 3 Li⁺(H₂O)₂ and combined structure

2. Combined energy analysis

Through cauculating the reaction energy of M²⁺(H₂O)₆ + P -> P-M²⁺(H₂O)₆, the binding energies obtained are shown in the table below.

According to the data in the table, it can be seen that LBP3 binds more strongly to the three transition metal ions and reduces the energy of the system more compared to binding with Li⁺.

Considering that the calculated results for Li may have the effect of hydration, we used the XEDA program to calculate the interaction energy for Li⁺, which yielded an energy of -334.16 kJ/mol, which is closer to the binding energies of the three transition metal ions; therefore, analyzing only a single protein is not sufficient to elucidate the metal ion binding selectivity. Next, we chose the Ni²⁺ ion with the strongest binding energy in Table 1 and analyzed the binding energies of its binding peptide, NBP1, for the four ions, comparing them with those of LBP3 to derive the selectivity.

As can be seen from Table 2, only comparing the binding energy of Ni²⁺ and Li²⁺, it can be seen that the binding energy of NBP1 to Ni²⁺ ions is -508.58 kJ/mol, which is much greater than that of LBP3 to Ni²⁺ ions -370.40 kJ/mol. According to the theory of statistical physics, the Ni iron will combine with NBP1 first and almost all Ni²⁺ may combined with NBP1. Although the binding energy of NBP1 to Li⁺ is stronger than that of LBP3, the ability of NBP1 to rebind to Li⁺ after adsorption of Ni²⁺ is weakened, and LBP3 mainly binds to Li⁺ ions, so it is selective.

As can be seen from Table 3, only comparing the binding energy of Ni²⁺ and Li⁺, it can be seen that the binding energy of NBP4 to Ni²⁺ ions is -566.08 kJ/mol, which is much greater than that of LBP3 to Ni²⁺ ions -370.40 kJ/mol. According to the theory of statistical physics, the Ni iron will combine with NBP4 first and almost all Ni²⁺ may combined with NBP4. Although the binding energy of NBP4 to Li⁺ is stronger than that of LBP3, the ability of NBP1 to rebind to Li⁺ after adsorption of Ni²⁺ is weakened, and LBP3 mainly binds to Li⁺ ions, so it is selective.

As can be seen from Table 4, comparing the binding energy of Co²⁺ and Li⁺, it can be seen that the binding energy of CBP2 to Co²⁺ ions is -381.12 kJ/mol, which is much greater than that of LBP3 to Co²⁺ ions -329.92 kJ/mol. Like NBP1, the Co iron will combine with CBP2 first and almost all Co²⁺ may combined with CBP2, CBP2 has a weakened ability to rebind to Li⁺ after adsorbing Co²⁺, and LBP3 mainly binds Li⁺ ions in the future, so it is selective.

Comparing the binding energy of Co²⁺ and Ni²⁺, although CPB2 has a stronger binding ability to Ni²⁺, the binding energy is -422.10 kJ/mol, but it is much smaller than the binding energy of NBP1 to Ni²⁺ -566.08 kJ/mol.Comparing the binding energy of Co²⁺ and Ni²⁺, although CPB2 has a stronger binding ability to Ni²⁺, the binding energy is -422.10 kJ/mol, but it is much smaller than the binding energy of NBP1 to Ni²⁺ -566.08 kJ/mol.

Since the computational power required to calculate the binding peptide for Mn ions was out of range, we performed a segmentation operation, and obtained that the binding energy of this binding peptide for metal ions was about 1000 kJ/mol, and it had the strongest binding ability for Mn²⁺ ions, which was much larger than that of the previous calculations, and the effect of the statistical theory's assignment of the number of states to the bouyant on the selectivity was much more pronounced. However, the error of the results introduced by the segmentation operation is too large to be used as a reference, so it is not discussed here.

Since the energy obtained from finding a single peptide conformation is not accurate enough, Prof. Chen also suggested that we try more peptide conformation calculations, change the spin multiplicity, and select the one with the lowest energy for further analysis.

Structural simulation of the peptides polymers with the help of AhphaFold3

To identify metal ion-binding peptides with potential for self-assembly, we have decided to utilize AlphaFold3 for structural prediction of these peptides.

Fig. 4 Prediction of metal ion binding peptide monomers

To evaluate its self-assembly potential, we gradually increase the number of its oligomers, where the color intensity represents the confidence of the predicted structure (with deeper blue indicating higher confidence and yellow indicating lower confidence). The self-assembly potential of a peptide will be assessed from multiple perspectives, including its extendibility, symmetry, and structural stability.

Fig. 4 different polymers of N3&N2

Clearly, we observed that for certain peptides, the feasibility of their structures rapidly declines with an increasing number of oligomers, indicating a low likelihood of forming stable oligomers and suggesting a lack of self-assembly potential. Additionally, for some peptides, although their oligomer stability is acceptable, their shapes are either barrel-like or non-extendable tetrahedral forms. Such peptides are also considered to lack self-assembly potential.

Fig. 5 potential of different peptide for self-assembly

However, to our surprise, among the numerous metal ion-binding peptides, we identified one peptide—N4—that exhibits favorable properties in terms of extendibility, stability, and symmetry. Subsequently, we conducted a series of tests based on this peptide. Experimental results indicated that our predictions were indeed effective, as N4 demonstrated the capability to self-assemble into nanospherical structures.

Fig. 6 self-assembly capability of N4

Based on the aforementioned simulation and experimental results, given the self-assembly ability of N4, we hypothesize that it can enhance its metal adsorption capacity by increasing metal binding sites through the self-assembly process.

Reference