Applying swarm intelligence to study molecules

Applying swarm intelligence to study molecules

Can that line of ants on the wall, or the flock of flying birds help us decipher the properties of a very large molecule? Yes, say scientists from the Indian Institute of Technology (IIT) Bombay, IIT Kanpur, IIT Guwahati and the Indian Association for the Cultivation of Science, Kolkata. In a recent study, they have described how the concept of swarm intelligence can be used to determine the most stable configuration of a molecule and its electronic structure. With this knowledge, scientists can design molecules with targeted properties for application in drugs, vaccines and polymers. This study was published in the International Journal of Quantum Chemistry.

Swarm intelligence is a behaviour exhibited by simple ‘agents’ that interact with one another and with their environment by following a set of simple rules. These interactions, over time, lead to the emergence of an ‘intelligent’ global behaviour, unknown to the individual agents. For example, if you look at a single ant, it may not come across as smart enough or powerful enough to carry a big insect to its nest. However, when many ants from the colony come, the task seems doable!

Many forms
Now, the researchers of the study, led by Professor Shankar Prasad Bhattacharyya from IIT Bombay, have developed an evolutionary algorithm based on the concepts of swarm intelligence to calculate the energy configuration and electric charge distribution in a molecule of polythiophene, a polymer, on doping it at various levels. These polymers have excellent electrical conductivity and optical properties with dramatic colour shifts in response to changes in solvent, temperature, applied potential, and binding to other molecules. Hence, they are used as sensors in many applications.

Every molecule consists of a group of atoms linked together by chemical bonds. The structure of the molecule is expressed in terms of the length and the angle of these chemical bonds formed between different atoms. A molecule can have various physical forms and properties based on the arrangement of its constituent atoms. For example, the high surface tension of water and the ability to dissolve many particles is because of the physical and electrical properties of the chemical bonds between two molecules of hydrogen and one molecule of oxygen.

Such an arrangement of atoms in a particular configuration also corresponds to different energy levels of the molecule. Lower energy levels imply higher stability of the configuration. These energies, when plotted against geometrical parameters, like bond length and angle, give rise to a potential energy surface in an n-dimensional space. Each point on this surface corresponds to a unique structure having a particular level of energy and electronic structure.

Search efficiency
It is challenging to determine the global minimum potential energy point of a molecule and its electronic charge distribution simultaneously due to presence of multiple local maximum and minimum potential energy points. To address this challenge, the researchers have applied the technique of swarm intelligence. “To search for a global minimum point on a function, we can have multiple agents, known as swarm particles, exploring the space and communicating with each other to optimise the search efficiency,” says Rishabh Shukla from IIT, Guwahati, and the study’s lead author.

So how does swarm intelligence work in this context? “Let’s say we start with 10 or 20 swarm particles with random positions and random initial velocities that will move in the n-dimensional space of the molecule of interest. While traversing their path, they will remember the point with minimum energy configuration found by them so far, called ‘individual best point’. If they come across a point, which is better than the individual best point, then it is updated. In addition, the swarm remembers its global best point. If an individual best point is found to be better than global best point, then global best point is updated,” explains Rishabh. Furthermore, the velocity of each particle is updated based on current velocity, its distance from the individual best point, and its distance from the global best point. This way, the swarm is continuously going towards better points and finally reaches the true global minimum, he adds.

This study is a unique attempt to apply techniques of swarm intelligence to the study of molecular structures. “The work signifies that collective intelligence displayed by classical swarms can be exploited to search through the space of the nuclear degrees of freedom and bring in simultaneous evolution of the electronic charge distribution. The net outcome is a mixed quantum classical method that smoothly and simultaneously locates the global minimum energy configuration of nuclei and the associated electron density distribution in a large molecule,” concludes Professor Shankar.

(The authors are with Gubbi Labs, a Bengaluru-based research collective)

Comments (+)