Inverse Boltzmann Iterative Multi-Scale Molecular Dynamics Study between Carbon Nanotubes and Amino Acids

Abstract

Our work uses Iterative Boltzmann Inversion (IBI) to study the coarse-grained interaction between 20 amino acids and the representative carbon nanotube CNT55L3. IBI is a multi-scale simulation method that has attracted the attention of many researchers in recent years. It can effectively modify the coarse-grained model derived from the Potential of Mean Force (PMF). IBI is based on the distribution result obtained by All-Atom molecular dynamics simulation; that is, the target distribution function and the PMF potential energy are extracted, and then, the initial potential energy extracted by the PMF is used to perform simulation iterations using IBI. Our research results have been through more than 100 iterations, and finally, the distribution obtained by coarse-grained molecular simulation (CGMD) can effectively overlap with the results of all-atom molecular dynamics simulation (AAMD). In addition, our work lays the foundation for the study of force fields for the simulation of the coarse-graining of super-large proteins and other important nanoparticles.

Keywords: CBNs; IBI; multi-scale molecular dynamics.

Figure 1

(A) Schematic diagram of the 20 amino acids–CNT55L3 simulation system. Amino acids are grouped as: (1) non-polar amino acid group (orange); (2) polar uncharged amino acid group (yellow); (3) aromatic amino acid group (green); (4) positively charged amino acid group (blue); and (5) negatively charged amino acid group (blue–purple part). (B) Simple schematic diagram of IBI process, IBI reverse Boltzmann iteration process to extract coarse-grained force field from all-atom trajectories, and $U_0$ is obtained from the AAMD.

Figure 2

(A) Schematic diagram of the coarse-grained method of the simulation system Mapping: (1) CNT part: 10 AAbeads are coarse-grained into one CGbead, 24 CNT-CGbeads in total; (2) Amino acids Part: An amino acid molecule is coarse-grained into one CGbead. Mapping the main chain and side chain of the amino acid to one CGbead. (B) Linear addition function. (C) Detailed flow chart of IBI-20Aminoacids-CNT55L3, including the interaction process of each file. (D) Table: list of the average field potential energy of each initial interaction relationship.

Figure 3

(A) The radial distribution functions between different types of amino acids and carbon nanotubes and their potential energy functions extracted by full-atom simulation: non-polar amino acid group RDF (a1) and its U_PMF(r) (a2); the RDF (b1) of the polar uncharged amino acid group and its U_PMF(r) (b2); aromatic amino acid group RDF (c1) and its U_PMF(r) (c2); the positively charged amino acid group RDF (d1) and its U_PMF(r) (d2); the negatively charged amino acid group RDF (e1) and its U_PMF(r) (e2). (B) The bond distribution functions extracted by the all-atom simulation and its potential energy function (take CNT55L3-15-Trp as an example): NH2-amino acids, (a1) P_Bond(r), (a2) U_Bond(r); amino acids-ACE, (b1) P_Bond(r), (b2) U_Bond(r); CNT-CNT, (c1) P_Bond(r), (c2) U_Bond(r). (C) The distribution of the angle distribution function extracted by the all-atom simulation and its potential energy function (take CNT55L3-15-Trp as an example): NH2-amino acids-ACE, (a1) P_Angle($\theta$), (a2) U_Angle($\theta$); CNT-CNT-CNT, (b1) P_Angle($\theta$), (b2) U_Angle($\theta$).

Figure 4

Representative amino acid studies: (1) short-chain non-polar amino acid 1-Gly, (a1) RDF iteration, (a2) IBI potential energy iteration; (2) long-chain non-polar amino acid 7-Met, (b1) RDF iteration, (b2) IBI potential energy iteration; (3) Aromatic amino acid 15-Trp, (c1) RDF iteration, (c2) IBI potential energy iteration.

Figure 5

The radial distribution function and target radial distribution function between 20 amino acids and CNT55L3 obtained after IBI iterative optimization. The black lines in each figure are their respective target RDF, and (a1–a7) are the non-polar RDF of the amino acid group, the line color is orange; (b1–b5) is the RDF of the polar uncharged amino acid group, the line color is yellow; (c1–c3) is the RDF of the aromatic amino acid group, the line color is green; (d1–d3) is the RDF of the positively charged amino acid group, the line color is blue; (e1,e2) is the RDF of the negatively charged amino acid group, the line color is purple.

Figure 6

The non-bond potential energy function and initial non-bond potential energy function (PMF) between 20 amino acids and CNT55L3 obtained after IBI iterative optimization. The black dotted lines in each figure are their respective initial non-bond PMF, and (a1–a7) are the non-polar PMF of the amino acid group, the line color is orange; (b1–b5) is the non-bonding PMF of the polar uncharged amino acid group, the line color is yellow; (c1–c3) is the non-bond PMF of the aromatic amino acid group, the line color is green; (d1–d3) is the non-bonding PMF of the positively charged amino acid group, and the line color is blue; (e1,e2) is the non-bond PMF of the negatively charged amino acid group, and the line color is purple.

Figure 7

(A) Initialized by PMF, it then obtains a proven force field with IBI; (B) IBI through CGMD to generate a new potential energy function every iteration; (C) The accuracy of the decision function between the initial target radial distribution function of amino acids and CNT55L3 after IBI iterative optimization and the radial distribution function obtained by the best iteration.