The smell of coffee is the same whether it is smelled in a coffee shop or grocery shop (different backgrounds), on a hot day or a cold day (different ambient conditions), after lunch or dinner (different temporal contexts), or using a deep inhalation or normal inhalation (different stimulus dynamics). This feat of pattern recognition that is still difficult to accomplish in artificial chemical sensing systems is routinely performed by most biological sensory systems. How is this capability achieved? We systematically investigated this issue using the relatively simpler invertebrate olfactory system.

To achieve our goals, we developed an integrative strategy that took advantage of multi-unit electrophysiological approaches that are well established in locusts (Schistocerca americana) and the rich repertoire of genetic and imaging tools available in fruit fly (Drosophila melanogaster). In addition, we developed quantitative behavioral assays to constrain our interpretation of the relevance of neural response features identified in our studies. I have summarized a few key findings from our investigations below.

Sensing and un-sensing a stimulus: It is well known that all sensory stimuli evoke spiking responses that are patterned across neurons and time. We found that responses elicited by odorants did not terminate after their cessation. Notably, we found two distinct sets of neurons were activated, one during the stimulus presentation (i.e., ON response) and the other after stimulus cessation (i.e., OFF response). Both ON and OFF neural responses contained information about the stimulus identity and intensity. Furthermore, the ON and OFF responses also differed in their ability to recruit recurrent inhibition, entrain field-potential oscillations, and more importantly in their relevance to behavior. While ON responses were good indicators of behavioral response initiation, the OFF responses were better indicators of when that behavioral response ended. Notably, we found that a strikingly similar information representation strategy was also used in the fruit fly olfactory system and the marmoset auditory cortex.

A computational logic for olfaction:  We found that the neural responses evoked by an odorant were not stable, but varied depending on what other stimuli were encountered in the recent past. We found that the perturbations to odor-evoked neural responses were not random but allowed the olfactory neural network to enhance the uniqueness or novelty of the stimulus. Nevertheless, this adaptive strategy to encode information about the odorant identity could lead to challenges in robustly recognizing that odorant. We found that there existed a simple approach to decode information distributed in varying subsets of neurons (i.e. flexible decoding). An example of the flexible decoding approach is a very simple linear logical classifier: OR-of-ANDs (disjunction of conjunctions). Our results indeed confirm that the predictions made using an OR-of-ANDs logical classifier do precisely match with the quantitative behavioral responses evoked by odorants.

We found that adapting the simple logical decoding approach by incorporating activity of ON neurons (i.e., ‘evidence/case for’)  and suppression response in OFF neurons (i.e., ‘evidence/case against’) enhanced the recognition performance. Surprisingly, this integrative scheme was found to be sufficient to allow odorant recognition independent of several perturbations that are likely to be encountered while sensing in natural conditions. 

Bioinspired methods for artificial olfaction: In artificial olfaction, our earlier work in this area focused on sensor material development and characterization (electronic, optical and, biological transducers), designing and optimizing array-based solutions for chemical sensing problems including homeland security and medical diagnostics. Using an approach inspired by how odorants are robustly recognized in the olfactory system (i.e., the OR-of-ANDs approach), we identified biomarkers in exhaled breath that could help diagnose malaria in children infected with those parasites. Furthermore, borrowing approaches used to characterize neural responses in sensory neuroscience, we developed an approach to characterize a chemical sensor’s response that allowed robust and invariant recognition of an odorant for extended periods of operation. More recently, we developed methods to monitor neural responses in moving and behaving insects that could be exploited to develop bio-hybrid sensing technologies that potentially can exploit the sophisticated sense of smell in insects.  

Dissemination of Scientific Results: Our results have been continually disseminated to the scientific community at large through publications in top-tier neuroscience or inter-disciplinary journals. Accompanying News releases associated with each publication provided a layman description of the results to ensure communication of our results to the public audience.

Education and Outreach Activities: Our research efforts allowed us to develop interdisciplinary educational opportunities that integrated basic concepts in mathematics, biology, and engineering. Opportunities at all educational levels, graduate, undergraduate and, K-12 level were created: five Ph.D. dissertations, in-class training to hundreds of graduate and undergraduate students, and hands-on lab activities for nearly 75 middle school girls and kids from under-represented minorities to encourage the pursuit of higher education and careers in STEM fields were all made possible through the support provided by this NSF CAREER grant.

Last Modified: 12/24/2021
Modified by: Baranidharan Raman