Mr. Aarush Mohit Mittal’s interview with Bio Patrika hosting “Vigyan Patrika”, a series of author interviews. Mr. Mittal is currently a PhD student in the lab of Prof. Nitin Gupta in the department of Biological Sciences & Bioengineering at Indian Institute of Technology, Kanpur. Mr. Mittal completed his B.Tech from IIT Kanpur. He published a paper titled “Multiple network properties overcome random connectivity to enable stereotypic sensory responses” as the first author in Nature Communications journal (2020).
How would you explain your paper’s key results to the non-scientific community?
All organisms show typical behaviors in response to certain stimuli. It could be salivating in response to food, approaching a sweet-smelling flower, or fleeing from a mortal enemy. The brain controls these and all other activities that we perform. Inside our brain are millions of cells, called neurons, that connect and form neural networks, like an electrical circuit wherein different elements connect through wires. In the case of the electrical circuit, we know precisely which part connects to which other part and how they work together to, say, light a bulb; in the case of our brains, we do not. Another difference is that the neural connections that one individual has between parts might be very different from those of another individual. However, both individuals can still show the same behavior in response to the same input. How does this consistency arise despite the randomness inherent in the system? This puzzle inspired us to dig deeper into such neural networks. We used the insect olfactory circuit (the circuit that controls our sense of smell) as a model to try and answer what controls the level of similarity between individuals.
Insects have two antennae which have tiny neurons that sense the odors present in the environment. These neurons all go inside the brain to an area called antennal lobe; let’s call it layer 1.
Neurons from layer 1 take the odor information, process it, and then relay their information to layer 2, which then conveys it to layer 3, and beyond. Layer 1 to layer 2 connections are random, i.e., two individuals will have very different neural connections. Layer 2 neurons, thus, respond very differently in different individuals. Despite this randomness, layer 3 neurons show very similar responses across individuals. We found that even when individual neurons of layer 2 are not consistent if we sum up the activity of this layer and compare it across individuals, it is very similar. As hundreds of layer 2 neurons connect to one layer 3 neuron, layer 3 neurons also become consistent. How?
We found that the more neurons that a neuron takes input from, i.e., the more the convergence, the better the neuron will be at differentiating between odors. Layer 3 neurons take advantage of the convergence from layer 2 neurons and thus can consistently identify and differentiate between odors. But how does the layer 2 neuron population become consistent? For two different odors, the number of active layer 1 neurons, the range of activity of each neuron, or the total activity of layer 1 is very different. Layer 2 neurons identify these differences as a whole and become consistent across individuals. Also, the more the number of layer 2 neurons that respond to odors, the more consistency there is in their responses.
Most organisms have some capacity for learning from their experiences. These experiences change the connections that we have in our brains so that we can all, say, ride a bicycle subconsciously. Insects also have some capacity for learning from their environment. An earlier study done by a group from researchers from Columbia University had shown that the consistency seen in the olfactory circuit required two different individuals to have a similar learning experience. We disprove this result as, in our model, we did not incorporate any form of learning mechanism but still got consistent responses. In fact, we show that learning makes individuals less consistent.
“[…] there is no need to for identical connectivity between layers; randomness followed by a convergence of neurons makes responses consistent across individuals.”
What are the possible consequences of these findings for your research area?
The multi-layered olfactory circuit seen in insects is very similar to the circuit that is present in mammals, including humans. Thus, it provides an excellent model not only to understand insect behaviors but also to learn about human behavior by proxy. Our results can be extrapolated to understand why everyone likes the smell of a rose or dislikes bad breath. Additionally, all neural circuits built in a similar way will follow the same principles. Random connections between neural layers are commonly seen in many species’ brains and are more prevalent the more the number of neurons there are. We show in our study that there is no need to for identical connectivity between layers; randomness followed by a convergence of neurons makes responses consistent across individuals.
What was the exciting moment (eureka moment) during your research?
One of the exciting moments was when we discovered that even if individual layer 2 neurons were not consistent, their total response was highly consistent. This discovery formed the focal point of all our subsequent insights into the system. The other exciting moment was when our theoretical computations complemented the analytical results. It strengthened the argument that our conclusions were valid for not only the insect olfactory system but also for all circuits with similar architecture.
What do you hope to do next?
Currently, we are working on expanding our knowledge of the insect olfactory circuit. By focusing on each different layer of neurons involved in odor detection, we hope to build a comprehensive map of this system’s capabilities. As the lab studies mosquitoes, we hope to use this map to engineer solutions that will help reduce the spread of diseases like dengue, malaria, etc. I am particularly interested in figuring out the basic principles of computation that drive the brain by studying how it handles multiple sensory inputs like light, sound, odors, and combines them together.
Where do you seek scientific inspiration?
In nature itself. Our brains are excellent at what they do, but how they do it is still a big mystery. We have barely scratched the surface into understanding our own minds. However, we still hope to create artificial intelligence to mimic and surpass us in the future. I believe that goal will only come to fruition once we learn what we, ourselves, are capable of. Researching the intricacies of natural neural networks and learning their secrets inspires me to keep persevering towards my goals.
How do you intend to help Indian science improve?
The work discussed above was done in collaboration with a group that had expertise in working with the fly model system, so I have first-hand experienced the importance of collaborations in the scientific process. And I feel there could be more collaborations across laboratories and institutions in India. Collaborative research facilitates faster discovery and efficient use of resources. It also promotes an environment of healthy discussion amongst peers. I would like to help set up such collaborative spaces to allow the Indian scientific community to be far more productive than we currently are.
Mittal A M, Gupta D, Singh A, Lin A C, Gupta N. Multiple network properties overcome random connectivity to enable stereotypic sensory responses. Nat Commun 2020, 11: 1023.