Neuroscientists from St. Petersburg University, led by Professor Allan V. Kalueff, have become the first in the world to use artificial intelligence (AI) algorithms to profile zebrafish psychoactive drug responses in partnership with an international team of IT specialists. They were able to train an AI to determine which psychotropic compounds were utilized in the experiment based on fish responses.
The outcomes of the study have been published in the journals Progress in Neuro-Psychopharmacology and Biological Psychiatry.
The zebrafish (Danio rerio) is a freshwater bony fish that is currently the second-most used model organism in biomedical research (behind mice). There are various advantages to using zebrafish as a model biological system, including inexpensive maintenance costs and strong genetic and physiological similarities to humans. We share 70% of our genes with zebrafish.
“The most challenging part of the study was to find the critical optimal AI training protocols and to confirm the validity of the research methods and the accuracy of the research findings,”
Professor Kalueff
Furthermore, the zebrafish nervous system’s simplicity allows researchers to acquire more explicit and accurate results when compared to experiments with more complicated creatures.
According to Professor Allan V. Kalueff, the study’s Principal Investigator and Head of the Laboratory of Biological Psychiatry at the Institute of Translational Biomedicine at St. Petersburg University, neural networks are gaining ground in biomedicine as a promising, reliable, and efficient research tool.
They allow for the unbiased and objective examination of biological data, which contributes to the discovery of new general patterns that may not be visible at first glance or accessible from the general data set.
Despite the fact that artificial intelligence is rapidly being employed in neuroscience, the researchers at St. Petersburg University were the first to use AI neural network-based algorithms to assess zebrafish locomotor traces (movements).
The researchers compared data acquired earlier in a series of in vivo tests with adult zebrafish exposed to neurotropic medications and control groups that were not treated.
In the study, zebrafish were given acute doses of several psychotropic substances such as nicotine, ethanol, caffeine, and others. Each of these medicines has an effect on the zebrafish’s central nervous system (CNS) and locomotor patterns.
The experimental results have already been published in peer-reviewed journals by scientists at St. Petersburg University.
Differences in zebrafish locomotor activity under the effects of several CNS drugs were used to train AI algorithms on video data collected in prior investigations.
The study employed a convolution neural network (CNN) model, which was inspired by the cortical visual information processing system in human and animal brains and was specifically built to work with visual data (pictures). It captures simple visual elements like gradients and lines and then combines them at the next layer to provide a richer and more sophisticated image representation (shape).
Each CNN layer adds to the complexity of the processed data, making it easier to extract non-trivial drug-specific patterns of animal locomotion.
For example, ketamine causes fish to exhibit conspicuous circular behavior (usually around the water’s surface), but ethanol causes dose-dependent biphasic effects with initial activation and later drowsiness.
“The most difficult element of the study was determining the crucial ideal AI training procedures and confirming the validity of the research techniques and the correctness of the research findings,” Professor Kalueff added.
In the end, everything worked out beautifully. The study outcomes demonstrated the theoretical applicability of our AI approach for analyzing the behavioral impacts of neuroactive medications on zebrafish. From a practical standpoint, this provides us with an abundance of opportunities to explore new psychotropic medications, he explained.
The researchers of this study believe that by improving and changing the neural network models, the functionality of AI applications can be increased even more. Furthermore, its significance is projected to increase as more experimental data becomes available for AI training.
Based on the animal’s behaviour and movement characteristics, researchers trained an AI to distinguish which psychotropic chemical a zebrafish had been exposed to.
St. Petersburg State University is the source.
