LABORATORY: RESEARCH

Research


	Computational chemistry is a research field in which various chemical phenomena are understood and predicted by solving the physical equations that molecules, atoms, or electrons obey. Generally, these physical equations are very complicated, so we use computers to carry out our research. In our laboratory, we hope to contribute to society by researching and solving various problems in the life sciences based on computational chemistry. * Click on an item below to see its description, and click on an image to enlarge it.

▼ Methods of Computational Chemistry in Life Science


	There are various methods in the field of computational chemistry. Here, we explain them using a protein as an example. Proteins are made up of many amino acid residues connected like a chain. In the coarse-grained simulation, a protein model is constructed considering these amino acid residues as a single particle connecting by springs, and the Newton equation of this model is solved to investigate the property or dynamics of proteins. Next, an amino acid residue is composed of several atoms, including hydrogen, carbon, nitrogen, and so on. In molecular dynamics simulation, a protein model is constructed considering these atoms as a single particle connecting by springs, and the Newton equation of this model is solved to investigate the property or dynamics of proteins. Furthermore, an atom is composed of a nucleus and several electrons. The method of solving the Schrodinger equation of these electrons to investigate the property of protein is called quantum chemical simulation or first-principles simulation. As explained here, methods of computational chemistry can be classified according to what is the smallest constituent unit used in the model. In addition, there is a method called docking calculation, which is specialized for screening of compounds in drug discovery. Docking calculation predicts which compounds strongly bind to a target protein out of tens to millions of compounds by using a relatively rough energy function or scoring function. In our laboratory, we are conducting various researches using these calculation methods.

▼ Development of an Original Quantum Chemical Program “PAICS”

In our laboratory, an original quantum chemical program “Parallelized Ab Initio Calculation System Based on FMO (PAICS)” is developed (http://www.paics.net/).This program is available to use the fragment molecular orbital (FMO) method, which enables us to perform quantum chemical calculations of large molecules such as biomolecules. In the FMO method, a target molecule is divided into small fragments, and only calculations of monomers and dimers for each fragment are required [1]. As a result, the computational efforts can be greatly reduced compared to the traditional quantum chemical methods. In addition, these monomer and dimer calculations are completely independent of each other, resulting in high parallel efficiency. In fact, PAICS is parallelized using Message Passing Interface (MPI) and is available to perform parallel computations using hundreds or thousands of CPUs. Fig. 1 shows the results of a benchmark on parallel efficiency. When the number of used CPUs increases by a factor of 8 (from 16 to 128), the computation speed increases by a factor of 7.18, indicating that high parallel efficiency is achieved. PaicsView, which is a user interface of PAICS, is also developed (Fig. 2).

References：
[1] K. Kitaura, et al, Chem. Phys. Lett., 313 (1999) 701.

▼ Original protein-protein interaction analysis method ”VIINEC”

We are developing an original protein-protein interaction (PPI) analysis method, “Visualization of the interfacial electrostatic complementarity (VIINEC).” [1][2] In VIINEC, the partial electron densities of each protein in a complex are calculated by the quantum chemical calculations, and the surface where they are the same value is defined as the contact surface between the proteins. Next, the partial electrostatic potentials of both proteins at the contact surface are calculated by the quantum chemical calculations, and their complementarities are visualized. As an example, Figure 1 shows an application of VIINEC to the PD-1/PD-L1 complex, which is an important molecule in cancer immunotherapy. We can visually see that the positive and negative electrostatic potentials (blue and red) are opposite at the contact surface. Additionally, it can be quantified, i.e., 92.1% of the electrostatic interaction at the contact surface is an attractive interaction.

One of the potential applications of VIINEC is to analyze the antibody-target interaction. Figure 2 shows the results that VIINEC has applied to the receptor-binding domain (RBD) of the spike protein of SARS-CoV-2 complexed with a neutralizing antibody [3]. From this calculation, we can see that 88.1% of the electrostatic interaction at the contact surface is an attractive interaction, but what is most interesting is that the complex pattern of blue and red is reversed and well-matched, resulting in a large attractive interaction. This is similar to complicated patterns of a dimple key and keyhole, illustrating the mechanism of the highly specific nature of the antibody-target interaction. We expect that VIINEC can be used in the design of antibody drugs.

参考文献：
[1] T. Ishikawa, Chem. Phys. Lett., 761 (2020) 138103.
[2] H. Ozono and T. Ishikawa, J. Chem. Theory Comput. 17 (2021) 5600
[3] T. Ishikawa, et al, J. Phys. Chem. Lett., 12 (2021) 11267

▼ In Silico Drug Discovery for Prion Diseases

Prion disease is a fatal disease caused by the structural change of prion protein from the cellular form (PrP^C)to the abnormal scrapie form (PrP^SC). The Researchers of Gifu University have discovered GN8, a compound that binds to the PrPC and inhibits the structural change to the PrP^SC [1]. However, to use this compound as a drug, it must be improved to a compound that binds more strongly. For this purpose, it is important to understand the details of the molecular interaction between GN8 and PrPC. Figure 2 shows the interaction energies of the four sites of GN8 (A, B, C, and D) with the protein, which were obtained from the fragment molecular orbital calculations [2]. It can be seen that the middle sites (B and C) interact strongly with the protein and play an important role in binding, while the sites A and D little contribute to the binding. In this study, we succeeded in developing a compound with a higher affinity to the protein by improving the B and C sites, which were found to be important for binding from our FMO calculations [3].

At Nagasaki University, we also developed an anti-prion disease compound that is different from GN8. A unique feature of this research was that we developed an original docking calculation system that took full advantage of the supercomputer owned by Nagasaki University. Among the 96 compounds selected by the calculation, NPR-53 and NPR-56 (Figure 3) were experimentally confirmed to exhibit high anti-prion activity [4].

Refereces：
[1] K. Kuwata, et al, Proc. Natl. Acad. Sci. USA, 104 (2007) 11921
[2] K. Yamaguchi, et al, Nat. Biomed. Eng., 3 (2019) 206
[3] T. Ishikawa, et al, J. Comput. Chem., 30 (2009) 2594
[4] D. Ishibashi, et al, EBioMedicine, 9 (2016) 238

▼ In Silico Drug Discovery for Infectious Diseases

Due to the recent global warming, changes in biological habitats caused by population growth, and the development of transportation networks such as airplanes, the threat of infectious diseases caused by a variety of pathogenic microorganisms has become a global problem that humanity must tackle together. In our laboratory, we are working on drug discovery for various infectious diseases using computational chemistry methods such as docking calculations, molecular dynamics calculations, and quantum chemical calculations. Fig 1 shows an anti-influenza compound targeting RNA polymerase, which was discovered using docking calculations from about 700,000 compounds [1]. We have also successfully developed an anti-influenza compound targeting NP proteins [2][3]. Other than the influenza virus, we are also developing anti-malarial compounds using quantum chemical calculations [4] and new drugs targeting APS kinase of amoebiasis [5].

As mentioned above, drug discovery for infectious diseases is a very important research issue, but even if new compounds are developed, in many cases pathogenic microorganisms that have acquired resistance emerge. In particular, RNA viruses, such as influenza, mutate very rapidly, and existing drugs may soon become useless. Therefore, we are also working on predicting resistance mutations to specific drugs using computers.

Refereces：
[1] K. Watanabe, et al, Sci. Rep., 7 (2017) 9500
[2] J. N. Makau, et al, PLoS ONE, 12 (2017) e0173582
[3] J. N. Makau, et al, Viruses, 12 (2020) 337
[4] T. Ishikawa, et al, J. Phys. Chem. B, 122 (2018) 7970
[5] F. Mi-ichi, et al, PLOS Neglected Tropical Diseases, 13 (2019) e0007633

▼ Simulation of Chemical Reactions in Condensed Systems

Chemical reaction simulations involving the bond formation and annihilation require quantum chemical calculations that directly solve the equations of motion of electrons rather than classical calculations that use empirical potential functions. However, it is generally difficult to perform such calculations in aggregated systems such as solvents because of the huge computational time required. Therefore, we have developed a simulation method using the fragment molecular orbital method that drastically reduces the calculation time [1]. As a first application study, we performed a simulation of the hydrolysis reaction of methyl diazonium ion in an aqueous solution. In this case, all of the more than 150 water molecules and reacting molecules were treated quantum chemically, and the chemical reaction was successfully reproduced in the computer (see the movie on the right).

参考文献：
[1] M. Sato, et al, J. Am. Chem. Soc. Comm. 130 (2008) 2396

▼ Computers in Our Laboratory

In our laboratory, we address various problems in chemistry and life science by using computer simulations instead of conducting experiments. We are using computers with a total of 584 cores and 2208 GB of memory for our researches (as of Jan 2022).

CPU	Core	Memory	Number
Xeon E5-2695v4×2	36	128GB	1
Xeon E5-2699v4×2	44	128GB	1
Xeon Gold 6240R×2	48	192GB	1
EPYC 7502×2	64	192GB	1
Core i7-6770HQ	4	16GB	24
Core i7-9700K	8	32GB	22
Core i7-10700K	8	32GB	11
Core i7-11700K	8	32GB	5
Core i7-12700K	8	32GB	11
Core i7-13700K	8	32GB	13
合計	584	2208GB