Methods of de novo drug design by computer-aiding (part one)

New drug design, also known as de novo design, three-dimensional structure generation or de novo construction method. According to the requirements of the space structure and properties of the receptor receiving point, the three-dimensional structure of a new ligand molecule with complementary shapes and properties is automatically constructed directly by the computer. Because it can propose a heuristic lead compound with a completely new structure, it is called a new or de novo design. It is a direct drug design method. In theory, this method takes into account the advantages and stability of the interaction between the receptor and the ligand, which is better than the structure searched in the database. The structure of the lead compound obtained may be brand new and undisturbed by anyone's prejudice.

The development process

 In 1989, Lewis and Dean used automatic template positioning for drug design; in 1991, Nishibata and Itai proposed an automatic atom growth method based on the structure of the receptor; in the same year, Moon and Howe proposed the concept of the molecular fragment method based on the structure of the receptor. Since then, these methods have developed rapidly, showing great superiority, and developing many corresponding software tools.

The starting point of the new drug design is also the complementarity between the receptor and the ligand. A basic building block is assigned to the acceptor, and then the database is searched and calculated to place the appropriate building block. Atoms or groups of atoms, resulting in real molecules that are complementary to the nature and shape of the receptor.

According to the different building blocks, the new drug design methods can be divided into: automated molecular template-directed drug design, atom build and molecular fragment approach. The molecular fragment approach can be further divided into fragment connection method and fragment build method. No matter what design method is used, a large number of molecules are obtained. The step of selecting the best ligand from these molecules is called scoring. At present, the method of analyzing the interaction energy between the ligand and the receptor is mostly used to score each molecule, and according to the score, a high-segment molecular structure is selected for further research.

1. Automated molecular template-directed drug design

Automated molecular template-directed drug design was proposed by Lewis in 1989. It uses a template to construct a three-dimensional molecular skeleton with complementary shapes at the receptor's receiving point, and then the molecular skeleton is based on the nature of the receptor into specific molecular structures. The first step of this process is called primary structure generation, and the second step is called secondary structure generation.

In the primary structure generation, a template library must be established first. A template is a three-dimensional molecular figure, with vertices representing atoms and lines representing chemical bonds. Vertices have sp3, sp2, and sp hybrid atoms, and bonds include single, double, triple, and aromatic bonds. The distance between the vertices is the bond length between atoms. The bond length and bond angle can be adjusted appropriately depending on the substituted atom. The vertices of each template can be replaced by any suitable atoms with different hybridization states, and thus many fragments can be generated. Place a template with a suitable shape on the target position, and place one of its vertices in the center of the target. Rotate the template around the vertex to get the best position, and then connect the new template in a certain way toward the remaining targets. Grow the skeleton to obtain a spatial skeleton that meets all the three-dimensional conditions, and remove the vertices that exceed the boundary and collide with the target, to obtain the molecular skeleton.

In the secondary structure generation process, according to other properties of the target, such as electrostatic, hydrogen bonding, and hydrophobic properties, replacing the vertices and lines in the molecular skeleton with appropriate atoms or atomic groups and bonds generates new molecules that are complementary to the properties of the target.

1. Atom build

The atomic build is proposed by Nishibata et al. According to the properties of the target, such as static electricity, hydrogen bonding, and hydrophobicity, the atoms are added one by one, and molecules complementary to the shape and properties of the acceptor point are added. Atoms are the basic element of blocks. The atomic library contains various types of atoms required for atom growth, such as sp3 hybrid carbon, aromatic carbon (sp2), carbon-based oxygen (sp2), light-based oxygen (sp3), amino nitrogen (sp3), etc .; also There are various types of chemical bonds, such as single, double, triple, aromatic, amide, etc.

The basic steps

The first step is to create a starting point. There are two kinds of starting points: one is an atom that is easy to form hydrogen bonds on the surface of the acceptor receiving point, which is called an anchor atom or a seed atom. Generally, it is a more electronegative atom, such as oxygen and nitrogen; the other is ligand or a part of the ligand which is known in advance at the accepting point of the receptor is called a starting structure.

The second step is atom growth. There are two ways: one is the system growth method, which uses atoms to combine to grow all possible molecules. This method has the problem of combined explosion. The number of structures produced is astronomical, even if it is difficult to handle using a computer. The other is the random growth method, which samples different structures and conformations. Random numbers are used in the selection of atom types, the selection of bond types, and the determination of atomic space positions, thereby improving efficiency and avoiding combinatorial explosions. At present, the second method is generally used, but the problem is that the random selection lacks chemical rationality.

The starting point for atomic growth can be entered by the user into the starting structure or automatically generated by the program. In the process of atom growth, the type, bond type, and spatial orientation of the new atom are determined according to the potential energy of each new atom. If the potential energy in 2 or 3 directions is favorable, branches can be generated; if the new atom in closed loops are preferentially looped; if the van der Waals radius between the new atom and the previously generated atom is unreasonable, or the van der Waals interaction is too high, then the new atom is re-designated; if the repeated process is still repeated, the shame process will return Pre-step, that is, retracting the previous Niu Cheng atom and regenerating a new atom. When the atom grows to a dead angle, that is, when all possible growth energy is high, or when the generated structure reaches the specified number of skeleton flexors, the program stops atom growth, the third step is to complete the molecular structure, add carbon atoms that may be missing to form the aromatic ring, and add hydrogen atoms to the vacant valence bonds of the atoms.

The final step is the optimization of the molecular structure. The molecular structure was used to optimize the resulting structure. Molecules were selected based on hydrogen bonding, Fan Lihua's force, and electrostatic and interaction.

Atom growth molecular design programs include Logend and Genstarn.

To be continued in Part Two…

Creative Biolabs provides professional antibody modeling assessment services for antibody modeling. With the company’s most devoted scientists, it is confident to deliver researchers the comprehensive and professional assessment to meet various project development.

Antibody Homology Model

The three-dimensional (3D) structure of antibodies offers an understanding of their function and evolution, and assists in drug design and optimization. When an experimental structure is unavailable, the antibody’s 3D structure is usually obtained through comparative modeling, also known as homology modeling, as well as via de novo computational methods. Commonly, there are four steps to construct a homology model: template selection, template–target sequence alignment, model building, and model evaluation. Besides, many sequence alignment tools and protein structure databases are available to meet the task, such as the Protein Data Bank.