Scientists at the Laboratory of Neural Intelligence at the Lomonosov Moscow State University Institute for Advanced Brain Research have developed DRIADA, a Python library for analyzing neural data, which the laboratory has been working on for the past few years. The library combines different analysis methods to simplify working with data and analyze research results more accurately. The framework is suitable for studying different types of neural activity, including for analyzing artificial neural network data.
Nikita Pospelov, a research engineer at the Laboratory of Neural Intelligence at IPIM Moscow State University, spoke about the work of DRIADA and its benefits for the scientific community.
The library is already publicly available. We invite you to familiarize yourself with its functionality by following the link: https://github.com/iabs-neuro/driada
– Nikita, tell us, what is the main practical advantage of DRIADA over existing solutions for analyzing neural data? What tasks will be more convenient and efficient to solve using your library?
– The key thing that we have built into DRIADA is the ability to reuse the same results in another method of analyzing neural activity. In neuroscience, brain activity can be analyzed in different ways: you can pay attention to individual neurons and see what each cell encodes, you can study the entire population, or you can build networks and explore the functional connections between neurons.
We are eliminating the usual approach where, when investigating the activity of a system, we analyze the same population of neurons using a variety of unrelated tools.
For example, you found out that specific neurons are responsible for encoding space. With DRIADA, you can create a network of interactions between these particular neurons in a few steps. It will be much easier and more convenient to do this, because this data already exists in the same structure and format.
In our research, we very often record neurons from the hippocampus of mice. The hippocampus is an area of the brain that is responsible, among other things, for short–term memory. There are cells in this zone that actively encode space.
When we use DRIADA to analyze hippocampal activity data, we can reconstruct the geometry of the space that the mouse explored, without looking at the mouse, the video recording, or other parameters, just from neural activity. Moreover, we immediately get a unified view.: As each neuron worked during her movement, we see that the virtual space in her head is encoded by certain neurons, while other neurons were doing other work at that time.
And then you can fantasize: for example, remove some of the neurons from the model and see what information remains in it. And doing all this with DRIADA becomes easier: you can immediately reuse the results of one analysis in another – it's convenient.
In general, convenience became the "engine" of our work: previously, we used several unrelated developments and obtained results describing neural code "in different languages." Largely thanks to the support of the Intellect Foundation, the amount of data we process has increased enormously in recent years, and there is a need to systematize the tools we work with.
– Who do you think will become the main user of DRIADA: novice researchers, experienced neuroscientists, or maybe specialists from related fields?
– DRIADA will be useful globally for researchers with programming skills. In other words, these are people who have data, but there are no good tools for analysis. We provide a tool that you can start using in a few clicks.
First of all, our tool will help computational neuroscientists: we hope that they will take advantage of the integration of different types of analysis. For example, if they used to look for individual selective neurons, now they can easily see what the whole population encodes.
But I wouldn't say that only strong programmers can use DRIADA.: we offered to use the library to students and people who are just starting to program. They coped with the task quite well, periodically consulting with the developers.
It seems to me that if you sit down and figure it out, even a person without programming knowledge will be able to use DRIADA and reproduce the actions that we show in the examples already on their data. But to solve more complex tasks, you need to have more serious training, of course.
– Tell us about the feedback: is there technical support or a community where researchers can share their experiences, ask questions and share the results of their work?
– Yes, there is a community, but it is not very big. They are mostly people who have faced the same problems as us, with the same experimental questions, and have asked us for analysis tools. So we gave them the tools and packed them into one library.
– Share your future plans for the development of the library. Are you thinking about creating a graphical interface or simplified versions for students to use in class?
– We are not planning a graphical interface yet: returning to the question of skills, in order to use DRIADA, you need to be able to work with python libraries. The immediate goals are to bring DRIADA in line with the standards accepted in global neuroscience and publish it in an international journal. We hope that DRIADA will be of interest to the international community, and people from all over the world will use the results of our work.
It should be noted that thanks to the systematic support of the Intellect Foundation and the Oleg Deripaska Volnoe Delo Foundation, specialized experimental systems, equipment for training neural networks, components for miniature microscopes are being purchased at the IPIM Laboratory of Moscow State University, as well as the necessary conditions for keeping laboratory animals and conducting research.
This makes it possible to significantly expand the experimental capabilities of the laboratory and improve the quality of scientific research, making a significant contribution to the study of the basic principles of the organization of intelligence in biological neural systems and creating a scientific foundation for the development of artificial intelligence technologies.
