Fujitsu has created a new molecular modeling technology.

The development of new drugs is associated with large financial investments and can last more than a decade. That is why scientists are constantly looking for promising methods for creating medicines. One of such methods is computer modeling technology. IT drug development has great potential, since this approach allows you to create virtual chemicals and evaluate their effectiveness on a computer model. Fujitsu has developed a molecular modeling technology that reduces the probability of errors 10 times during the modeling phase of new chemicals. In this article we will talk about it in detail.


The effect of a chemical as a drug manifests itself when it binds to the target protein. In this case, the substance changes its shape in accordance with the shape of the target protein. The level of change is directly related to the binding affinity of the substance and protein and provides a general idea of ​​the effectiveness of the drug. In the process of research scientists are very important to accurately predict this figure.

Methods based on the principles of quantum or Newtonian mechanics are used to calculate the level of change of a chemical. Ab initio calculations based on quantum mechanics provide the highest accuracy, representing an analysis of the state of electrons based on the type and position of atoms. However, the implementation of such studies takes a very long time. Exact modeling of the level of change of chemicals with this method takes years, which makes it unsuitable for practical use.

On the contrary, approximate calculations based on molecular modeling are performed very quickly and are widely distributed in science. They use the principles of Newtonian mechanics to calculate the interaction force between atoms and can be used to identify the state of large molecules, including protein.

From the point of view of Newtonian mechanics, the forces arising between atoms are expressed as follows:

1. As a force that depends on the distance between two bound atoms;
2. Like a force that depends on the angles between three bound atoms;
3. As a force that depends on the level of torsion in the bundle;
4. Like a force that depends on the distance between unbound atoms.

When a chemical substance is bound to a target protein, the level of twist of the ligament reflects the degree of deformation. However, when using existing technologies, the accuracy of determining the dihedral angle, the value of which is necessary for calculating the torsion of the ligament, is quite low, which causes a problem of low accuracy of binding affinity.


Dihedral angle (angle formed by the plane of atoms A, B and C and the plane of atoms B, C and D)

What is the essence of the development of Fujitsu?

Fujitsu has been developing molecular modeling for over a decade. Using the accumulated experience, scientists of the company have created a technology that takes into account not only the junction points, in which torsion occurs, but also the influence of adjacent atoms on each other.

The existing technology estimates the value of dihedral angles based on the position of 4 to 2 atoms in the bundle and other atoms to which these atoms are attached. However, depending on the structure of the molecule, there are cases when the atoms that are not included in these 4 atoms taken into account have a great influence on the neighboring ones, and in such cases the probability of errors in the calculations is high.

Fujitsu specialists have created a database of evaluation formulas for those cases where the influence of atoms located far from the bundle is crucial. Using the appropriate evaluation formula, it is now possible to carry out an accurate molecular torsion estimate, which was previously impossible.


Molecular structure example: 3- (methylamino) pyrazole

After Fujitsu integrated this technology with specialized software, the accuracy of the new development was studied under laboratory conditions. Fujitsu conducted an assessment of this technology for 190 types of chemicals, comparing the results obtained with the correct, obtained during non-empirical calculations, and calculating the coefficient of error. The study made it possible to prove that the likelihood of errors in estimating the level of twist in a new development is, on average, 10 times lower compared to the technology used previously.

Practical use

Estimation of the dihedral angle parameter using 190 types of chemical structures

A new method for determining the binding affinity of target proteins and chemicals demonstrates much more accurate measurement results compared to the previously used General AMBER Force Field 1.8 (GAFF 1.8) technology. Scientists suggest that its practical implementation will create fundamentally new drugs. Fujitsu also plans to incorporate this technology into its own drug development service.

* thermodynamic characteristic, quantitatively describing the strength of the interaction of substances