In an interview with Quanta Magazine, Vijay Balasubramanian discusses his work as a theoretical physicist, noting his study of the foundations of physics and the fundamentals of space and time. He speaks of the importance of interdisciplinary study and about how literature and the humanities can contextualize scientific exploration in the study of physics, computer science, and neuroscience.
Balasubramanian is Cathy and Marc Lasry Professor in the Department of Physics and Astronomy in the Penn School of Arts and Sciences and a member of the Penn Bioengineering Graduate Group.
The human brain uses more energy than any other organ in the body, requiring as much as 20% of the body’s total energy. While this may sound like a lot, the amount of energy would be even higher if the brain were not equipped with an efficient way to represent only the most essential information within the vast, constant stream of stimuli taken in by the five senses. The hypothesis for how this works, known as efficient coding, was first proposed in the 1960s by vision scientist Horace Barlow.
Now, new research from the Scuola Internazionale Superiore di Studi Avanzati (SISSA) and the University of Pennsylvania provides evidence of efficient visual information coding in the rodent brain, adding support to this theory and its role in sensory perception. Published in eLife, these results also pave the way for experiments that can help understand how the brain works and can aid in developing novel artificial intelligence (AI) systems based on similar principles.
According to information theory—the study of how information is quantified, stored, and communicated—an efficient sensory system should only allocate resources to how it represents, or encodes, the features of the environment that are the most informative. For visual information, this means encoding only the most useful features that our eyes detect while surveying the world around us.
Vijay Balasubramanian, a computational neuroscientist at Penn, has been working on this topic for the past decade. “We analyzed thousands of images of natural landscapes by transforming them into binary images, made up of black and white pixels, and decomposing them into different textures defined by specific statistics,” he says. “We noticed that different kinds of textures have different variability in nature, and human subjects are better at recognizing those which vary the most. It is as if our brains assign resources where they are most necessary.”
A study published in PLOS Computational Biology describes a new model for how the olfactory system discerns unique odors. Researchers from the University of Pennsylvania found that a simplified, statistics-based model can explain how individual odors can be perceived as more or less similar from others depending on the context. This model provides a starting point for generating new hypotheses and conducting experiments that can help researchers better understand the olfactory system, a complex, crucial part of the brain.
The sense of smell, while crucial for things like taste and hazard avoidance, is not as well studied as other senses. Study co-author Vijay Balasubramanian, a theoretical physicist with an interest in how living systems process information, says that olfaction is a prime example of a complex information-processing system found in nature, as there are far more types of volatile molecules—on the scale of tens or hundreds of thousands—than there are receptor types in the nose to detect them, on the scale of tens to hundreds depending on the species.
“Every molecule can bind to many receptors, and every receptor can bind to many molecules, so you get this combinatorial mishmash, with the nose encoding smells in a way that involves many receptor types to collectively tell you what a smell is,” says Balasubramanian. “And because there are many fewer receptor types than molecular species, you basically have to compress a very high dimensional olfactory space into a much lower dimensional space of neural responses.”
Understanding how the brain works is a major research challenge, with many theories and models developed to explain how complex information is processed and represented. One area of particular interest is vision, a major component of neural activity. In humans, for example, there is evidence that around half of the neurons in the cortex are related to vision.
Researchers are eager to understand how the visual cortex can process and retain information about objects in motion in a way that allows people to take in dynamic scenes while still retaining information about and recognizing the objects around them.
“One of the biggest challenges of all the sensory systems is to maintain a consistent representation of our surroundings, despite the constant changes taking place around us. The same holds true for the visual system,” says Davide Zoccolan, director of SISSA’s Visual Neuroscience Laboratory. “Just look around us: objects, animals, people, all on the move. We ourselves are moving. This triggers rapid fluctuations in the signals acquired by the retina, and until now it was unclear whether the same type of variations apply to the deeper layers of the visual cortex, where information is integrated and processed. If this was the case, we would live in tremendous confusion.”
Experiments using static stimuli, such as photographs, have found that information from the sensory periphery are processed in the visual cortex according to a finely tuned hierarchy. Deeper regions of the brain then translate this information about visual scenes into more complex shapes, objects, and concepts. But how this process works in more dynamic, real-world settings is not well understood.
To shed light on this, the researchers analyzed neural activity patterns in multiple visual cortical areas in rodents while they were being shown dynamic visual stimuli. “We used three distinct datasets: one from SISSA, one from a group in KU Leuven led by Hans Op de Beeck and one from the Allen Institute for Brain Science in Seattle,” says Zoccolan. “The visual stimuli used in each were of different types. In SISSA, we created dedicated video clips showing objects moving at different speeds. The other datasets were acquired using various kinds of clips, including from films.”
Next, the researchers analyzed the signals registered in different areas of the visual cortex through a combination of sophisticated algorithms and models developed by Penn’s Eugenio Pasini and Vijay Balasubramanian. To do this, the researchers developed a theoretical framework to help connect the images in the movies to the activity of specific neurons in order to determine how neural signals evolve over different time scales.
“The art in this science was figuring out an analysis method to show that the processing of visual images is getting slower as you go deeper and deeper in the brain,” says Balasubramanian. “Different levels of the brain process information over different time scales; some things could be more stable, some quicker. It’s very hard to tell if the time scales across the brain are changing, so our contribution was to devise a method for doing this.”
Before CRISPR became a household name as a tool for gene editing, researchers had been studying this unique family of DNA sequences and its role in the bacterial immune response to viruses. The region of the bacterial genome known as the CRISPR cassette contains pieces of viral genomes, a genomic “memory” of previous infections. But what was surprising to researchers is that rather than storing remnants of every single virus encountered, bacteria only keep a small portion of what they could hold within their relatively large genomes.
In recent years, CRISPR has become the go-to biotechnology platform, with the potential to transform medicine and bioengineering. In bacteria, CRISPR is a heritable and adaptive immune system that allows cells to fight viral infections: As bacteria come into contact with viruses, they acquire chunks of viral DNA called spacers that are incorporated into the bacteria’s genome. When the bacteria are attacked by a new virus, spacers are copied from the genome and linked onto molecular machines known as Cas proteins. If the attached sequence matches that of the viral invader, the Cas proteins will destroy the virus.
Bacteria have a different type of immune system than vertebrates, explains senior author Vijay Balasubramanian, but studying bacteria is an opportunity for researchers to learn more about the fundamentals of adaptive immunity. “Bacteria are simpler, so if you want to understand the logic of immune systems, the way to do that would be in bacteria,” he says. “We may be able to understand the statistical principles of effective immunity within the broader question of how to organize an immune system.”