Matt Angle:
Hello and welcome to Neurotech Pub. In today’s episode, we’ll be discussing the science of smell and the potential commercial implications of olfactory BCI. Our guests today include my own thesis advisor and Paradromics scientific co-founder, Andreas Schaefer. He’s the group leader at the Francis Crick Institute and a professor at UCL. Also, joining us today is Gabriel Lavella, the founder and CEO of Canaery, a company that’s developing cyborg animals whose noses can be read out electronically. And Dima Rinberg, Canaery co-founder, and professor at NYU. I hope you enjoy the discussion.

Matt Angle:
I wanted to start off by giving everyone a chance to introduce themselves and then we always like to have an icebreaker question in the beginning, so today’s is what did you want to be when you were eight years old? So if you could tell us a little bit about what you’re doing right now, what your affiliation is and what you wanted to do when you were eight years old, that would be great.

Gabriel Lavella:
Sure. My name is Gabriel Lavella, I’m the CEO of Commonsense and what we’re doing is we’re building neural interfaces to the olfactory bulb. What led me here? I did my PhD in the Maharbiz lab. I never worked in neuro technology before going to DARPA about two years ago. So I’d never done… I did probably six months of benchwork in the area in the beginning of my PhD. I put it down and I did my PhD in molecular nanotechnology doing self assembled structures for diagnostics, and then I worked in Silicon Valley for a while designing the MacBook, the iPhone, a bunch of other tech products. Went to DARPA, was an integral part of the NESD program, N3, as well as a few others in the neural interface space.

Gabriel Lavella:
Then I got a call through Phillip Alvelda who is one of the program managers at DARPA. Phillip had taken a role as Chief Technical Officer of a moonshot factory in Barcelona. So, the idea was this company would provide huge amounts of capital and launch one moonshot technology per year. The idea was the technology had to generate a billion euros in revenue and have a positive societal impact. So I went out there to work with Philip and started putting together this company. We got to a very late stage, they closed down alpha and came back to the States. John Viventi, who everybody probably knows, he introduced me along the way to Dima, and he said, “Dima’s been working on this.” And he has some fantastic, pioneering work in the area. So we came back to the States, incorporated, and here we are. So we’re launching now with the support of IndieBio in San Francisco and we’re building the company from the ground up.

Matt Angle:
And is that what you wanted to do when you were eight years old?

Gabriel Lavella:
You know, when I was eight years old I wanted to work at Disney World in any capacity. So, I could have been a character, I could have been operating a ride there. But I wanted to work at Disney World.

Matt Angle:
Thanks Gab.

Gabriel Lavella:
Sure.

Andreas Schaefer:
I’m Andreas Schaefer, I’m a neuroscientist I work at the Francis Crick Institute in London, at the University College London. I’m heading a research group that’s called Sensory Circuits and Neurotechnology Lab. It’s a small lab, we’re largely working on researching how the olfactory system works, working on mice. I’m using a variety of techniques from X-ray imaging, electro-microscopy, in vivo imaging the physiology, more and more computational work, behavior, and a part of my labs has been developing neurotechnology over the last years, that’s where Matt and I have been working together, quite a long time ago now. I’m originally a physicist by training and I kind of, I guess, lost all my physics knowledge and kept all the arrogance, so whether that’s helpful or not I leave for others to judge. But you know, kind of a broad neuroscience lab.

Andreas Schaefer:
What did I want to do when I was eight years old? I think generally I’ve been relatively boring and straight forward. I wanted to be a scientist or anything, Something I want to remember I wanted to… I wrote in my first yearbook I want to be an archeologist, not really knowing what that was. And then I’ve been always been between physics and neuroscience. I studied physics because I thought that would be really exciting to learn more and more about, and then parallel drifted into neuroscience in my PhD and post-doctoral work.

Matt Angle:
Ever look back at your decision not to be an archeologist?

Andreas Schaefer:
No, no. I don’t like Harrison Ford enough I think.

Dima Rinberg:
I’m Dima Rinberg. I’m a professor of Neuroscience and Physics at New York University. My lab is studying the sense of smell, olfaction, doing a lot of research in animals, trying to understand how and what the nose tells the brain and how the olfactory code is organized. We understand very basic how it is working. In parallel, for a few years we tried to do some applied work in the lab, trying to use all our knowledge about how olfaction works, to make the best chemical detector in the world. This project has been worked on for a few years. It was up and down until I get a call from Gabe and he said, “Hey, how about joining forces?” After the first conversation, I understand that we’re talking the same language and it’s super cool that this actually can be done on a much larger scale.

Dima Rinberg:
Talking to Gabe kind of the trick was to learning about the much broader opportunity in neurotechnology. I was always attentive to neurotechnology, but honestly building the first bioelectronic nose, I kind of did not see the path and Gabe showed me the path and we joined forces. Actually, we are part of the big olfactory consortium, we are eight labs trying to crack olfactory code, the name of the consortium is Osmonauts and it’s fascinating work and both the Osmonauts and the applied stuff with the company we founded. I’m a physicist by education and once a physicist always the physicist.

Dima Rinberg:
I grew up in Russia, I did get my physics education first in Russia undergrad then came to Weizmann Institute of Science to get my PhD in physics in fluid dynamics. And then I switched to neuroscience and it was a torturous path and all in all, then my second post-doc I started doing olfaction since 2002. When I was a kid, as many kids like in Russia there, I was interested in science and at dreamt to become an astronaut of course. I wanted to fly. And it was a big disappointment because I learned that having glasses and being an astronaut is incompatible well, that was my first depression.

Matt Angle:
It feels a little antiquated now, doesn’t it. I mean, I know they still have some restrictions, I think in the air force that you have to have.

Dima Rinberg:
Yes. So where they get my glasses in third grade, it was a big disappointment, but, but I started reading many books about flights, about space that brought me into physics, and then I become a physicist. So that’s my story.

Matt Angle:
And Dima, the last time we talked, you were looking to fill a postdoc position in your lab. Do you still have that posted?

Dima Rinberg:
Yes. For the project related to bio electronic nose, I’m actively looking for people and in general, both for the electronic nose and for olfactory coding, we have a few positions in the lab. Some of them funded by NIH Osmonauts Consortium, some of them for the new adventure with bioelectronic nose. So thank you for reminding this.

Matt Angle:
We’ll place the link to the job advert on the podcast.

Andreas Schaefer:
It was Osmonaut your idea to call it Osmonaut considering you wanted to be a Cosmonaut to start with.

Dima Rinberg:
Yeah. Haha.

Gabriel Lavella:
Dima, we’re on the same page after I got over the idea of working at Disney, my next dream was to be an astronaut as well.

Dima Rinberg:
Yes. That’s the conection.

Gabriel Lavella:
For the next six years that consumed me.

Dima Rinberg:
Yeah.

Andreas Schaefer:
And then you turned 35.

Matt Angle:
I’m going to ask a provocative question. Let’s assume that the listeners don’t care about how we smell. Why should they still be interested either from an intellectual standpoint or a functional standpoint in the way that olfaction works? Do people who study olfaction because they want to understand smell or are there other things that are interesting about it?

Andreas Schaefer:
I think that varies a lot. I think some do understand it simply because they’re fascinated by the smell itself. And for me, certainly the attraction of studying smell is that if you… We all want to assume understand how the human brain works. So if you want to do and if understand means sort of a mechanistic understanding, like I want to generally understand things you want to manipulate, you want to measure in all detail, the sector anatomy, physiology activity structure. And then the mouse or the rat is a very good model system. Now, if you look at mice and rats, they’re sort of primary way, they’re exploring the world around them is the sense of smell. So it’s a very good gateway into figuring out how their brain works. If you then look into the brain structure you have an anatomy that allows you to really understand better what’s going on.

Andreas Schaefer:
There are much fewer brain regions that interact to process smells. They are anatomically compact. So I think for me, it’s the intellectual appeal is that we have a quite confined structure where we have a realistic path towards understanding how the brain computes, how the representation of the external word gets into the brain. So I don’t care about smells that much.

Dima Rinberg:
I actually, I would say I have three reasons. The first is actually the same like this, because for me, the sense of smell is [inaudible 00:11:03] to the cords of sense to the brain. And it’s a very transparent being doable. Basically we are very close. Now, a system is much less computation and much less processing before getting to the cortex and objects. We manipulate these objects off of the objects and it’s easy to work with the system. Is it too? I don’t want to repeat what are these sad, but at the same time, I have two more reasons. First of all, it’s fascinating system. It is lets say lost tracheal wall senses.

Dima Rinberg:
And we know how color vision work. We invested so much in descending vision and hearing and unfortunately the big mystery and it is, I do care about olfaction deeply. I do care how first the world is build, how we smell, what happens to the brain when we smell the rose or when we smell the coffee, how we discriminate others. What actually the other spaces can be manipulated, can power they’re mixing. And it’s a lot of unsolved problems. It’s a huge intellectual exchange and I’m happy to be part of it. And the fun that isn’t actually to come to me, honestly, speaking, I didn’t start setting up for them, but more think about it is that we moving to a very new era of this style because we all are surrounded by organic follicles.

Dima Rinberg:
We’re all smelling all the time. We don’t have sensors, we don’t have receptors, we don’t take advantage of that. And all the evolution, this is probably unique system that allows us to have a look at this world of molecules or chemistry that it’s not yet here at our… We don’t have access. We do the very complicated devices like gas spectrometry, mass spectrometry, gas chromatography, mass spectrometry but the nose does it for us. And we kind of the whole animal world, all life in it fully integrated with the chemistry. We know much more about genetics know much more about zebra crossing, but we don’t know much about the others around us and nature does know takes care of all that. And we actually have now accessed, we actually understand how that the nose the device didn’t know that and this is an incident [inaudible 00:13:39] start treating how can smell disease, how we can smell other stuff. What actually, because animal can do this and that’s fascinate me to the same level as big objects of the grapes.

Andreas Schaefer:
So what you’re saying essentially the most that we have microphones that are significantly better than anything we can say, or hear, we have speakers that can reproduce the same, much better.

Andreas Schaefer:
We have cameras that are much better than our eyes but there’s no chemical detectors that are anywhere near the complexity, even of our human quite under utilized and under-trained knows it swept and if you go through more trained and [crosstalk 00:14:18].

Dima Rinberg:
That is untapped role. This is completely untapped role and I think that it’s a million discoveries will be… We have a right detection rules.

Gabriel Lavella:
For somebody who is a systems engineer like my job and coming from the Berkeley Sensor and Actuator Center, the BSAC I don’t know who came up with that acronym, but that’s the name of the center. It’s one of the most phenomenal sensors out there in terms of it’s the versatility of things that it can pick up. You have a natural world and one of the primary communication networks in that natural world are molecular signals, whether they’re airborne or they’re soluble, and olfaction is one piece of that.

Gabriel Lavella:
So for me, coming from a detection and diagnostics backgrounds and somebody who’s innately interested in those things, olfaction is just a fantastic world. That’s really interesting. So many things are communicated across that channel and it’s hidden from everybody and to me that makes it more interesting.

Andreas Schaefer:
And what makes it kind of intriguing in some ways it’s that, we have relatively poor language, relatively poor intuition for what one can do with smells. But if you look at what obviously many animals are doing within terms of navigating, detecting, figuring out where they are and where they have to go, what the olfactory environment is seems like that complexity that they can extract from the environment is very similar to the complexity that we can extract from our environment, but vision let’s say. And I think probably humans are much better with the olfaction than they like to admit it to themselves. I think this discrepancy between our intuition and what actually what the processing power the system has is quite sometimes difficult actually for studying, but quite intriguing.

Matt Angle:
Dima, you brought up the idea of odor space. I think people would be interested to hear about how you think of odor space because it’s not a three-dimensional space. Particularly you and Andreas, can you kind of unpack that for us a little bit?

Dima Rinberg:
I love this Christian, thank you to bring it up. Let’s start with that kind of color space because the receptors and we know that hollow space can be presented by two dimensional diagram. And you guys saw that the color diagram that you can draw all the colors and show how each different colors close to each other. You can draw the pictures of that if you can kind of mix colors on this diagram. If you have to source that everything in between can be mixed and created colors in between. What about odors? With all those let’s just do the very simple mental experiment. I give you to smell an orange and I give you to smell of apple and there’s a two independent objects you kind of hold slowly kind of guys well we know and now we’ll give you to smell and imagine you never smelled a lemon.

Dima Rinberg:
And I give you a first time in your life to smell a lemon. You immediately tell me that lemon will be closer to the orange than to the air. So what does it tell us that we have internal metric between, oops, we can show the similarity with distances between odors. Now, we can add other odors there. We can add grapefruit or banana or Corfield or rose and [inaudible 00:17:59] stop measuring distances are the two dimensional, three dimensional. So initial idea when odor receptors had been discovered that it is a huge number of dimensions, and it basically the analogy was taken that if color space have three receptors which at this [inaudible 00:18:24] humans have all the receptors that maybe did [inaudible 00:18:24] but I believe, and there’s a lot of here that the result of the space is much smaller. I don’t want to say how small it is, it’s five, 10 or 15, but definitely not the habit, but why this is important because the space of extraordinary stimuli exists.

Dima Rinberg:
You can create any odor by chemistry, but that space is projected to the cortex to the upper end. And when you smell something, you actually immediately kind of can establish the distances. So this is a kind of you can make these specimen and you can kind of perceptual judgment. And that means that, that projection, the understanding how this projected Hellman will make this judgment is super important for neuroscience or for computation and knowing the space of this image analogy and stuck to this space is important for in your own presentation. And that part of the big, big, big effort of my lab to try to achieve.

Matt Angle:
I’m going to press you on this for a second Dima because we just had a podcast on dimensionality reduction. And what do you think the kind of latent space of olfaction is? Do you think it’s 299 dimensions, or do you think it’s five?

Dima Rinberg:
I think it’s 10, maybe 10 something, I don’t know. I don’t want to give exact numbers, but it definitely way, way smaller than the total number of receptors, but it probably more than two. And if it would be two, we would already have been discovered this and I would probably all be doing something else. If it will be 300, I also start doing something else.

Matt Angle:
What do we know from psychophysics? When the wine Caisters and the perfume smellers try to establish categorical labels, do we have a sense of what do people think the dimensionality is?

Dima Rinberg:
So probably asking why smellers about the dimensionality of the space it’s wouldn’t lead us too far, but there was a few studies about that. And trying to kind of present the dimension reduction in odor space, I ask the subjects to characterize the odor of multiple distributors, and then being new odors with many distributors and many subjects. And then try to find the low dimensionality representation of the space is now what will the limited style is sexually turns out actually this work has been done. My cooperator, Alex Corvette, I would speak to them this work, this work shows that the perceptual space can be very well off. It’s limited by some a non-linear two dimensional manifold. It’s the chip shape apart. I think that the dimensionality is higher and the study was definitely a little bit with a small number of odors.

Dima Rinberg:
But what we learned from there that it’s small. We don’t know if it will be two or three, maybe it will be seven, but it’s much smaller than what is expected from the ambulatory receptors. And also Professor Weitzman, the students a lot of effort. There’s no students exact numbers, but he can collusion to say my ideas. I wanted to kind of talk about mentioned this one of the remarkable work has been published in nature about is you were doing the set of the dimensionality of odor space. We make, create the position that we may come to the same spot, but if by different mixtures, by creating different parts, it’s the same thing. In color, we can play the green odor by mixing blue and yellow, so that never have been done enough option in systematic way. And there was a recent paper by non-slip a lab where he was able to create absolutely undescribable odor presets by mixing very different mixtures.

Dima Rinberg:
That pals as the dimensionality probably small, otherwise it would be very difficult to use.

Gabriel Lavella:
What sort of data would you need to be able to solve that problem? That’s a hard question.

Dima Rinberg:
Yeah, it’s a hard question.

Gabriel Lavella:
You have access to get an infinite amount of data.

Dima Rinberg:
Well, to ask many, many people about many, many odors and ask them to discriminate them or to characterize them-

Andreas Schaefer:
And many people that have good verbal descriptors for their odors, reliable verbal descriptors, which I think is one of the issues with many of those kinds of exceptions, like a physics study, is that people just not very good in finding the right the right words as well.

Matt Angle:
Well, there’ve been a lot of studies in the sensory world where people record brain activity, and then they compare their ability to decode with the animals behavioral output. And sometimes they finally they can do better than the animals. Sometimes the animal appears to be doing better than them, but not as much. I’m curious, do you think if you went into the olfactory bulb and did a large recording and then you trained on that and then you compare that to people’s subjective descriptions, do you think there’d be a discrepancy? Do you think that it gets winnowed down later in the pipeline or do you think you’d find the same thing?

Andreas Schaefer:
I think it’s certainly maybe stepping back to back a bit, there are different ways to look at odor space, right? One is just the chemical space, just the chemicals that are out there, you can find chemical descriptors, and then you can sort of look at the dimensionality of that. The other is the one that maybe is the most relevant one. There’s a perceptual odor space that Dima mentioned. And then there would be the dimensionality of odor representations. So you sort of the all possible neural activities that are evoked by all possible chemicals, and certainly that be more complex than what humans or animals would be reporting behaviorally.

Andreas Schaefer:
Specifically with your question, what would I expect in the olfactory bulb? I mean, the great thing about affection is that essentially all information goes through the olfactory bulb is processed there and then distributed to all kinds of other places in the brain. So any information that human animal receives about the external chemical world will be represented effectively. So if we can record activity from all neurons or all projection neurons, all those neurons that lead from the effective of somewhere else, then we will have a representation that is better than and get higher dimensional than what anyone would report. Practically, I think anyone who has started recording from a few hundred cells in an olfactory bulb would always find decoding that is at least as good as what the animals would at the animal’s behavior reports.

Matt Angle:
Andreas, can you tell us a little bit about the overall architecture of the olfactory system, especially the bulb, and then going to piriform cortex, just so people can think about how does information flow after it hits their nose?

Andreas Schaefer:
Odors, so essentially any chemical that’s volatile, that’s in air, don’t talk [inaudible 00:25:26] about mammals it’s inhaled. It hits receptors in the neurons, that’s your receptor molecule sit on receptor neurons. They are a few hundred different types of those receptors, approximately three, 400 in humans, maybe about 1,000 in mice, but on the odor of a few hundred in most mammals. And then these perceptive neurons, they send their axons. So they have processes that realize information further down the line. They sent this process into the structure we’ve been talking about the olfactory bulb and they are all those receptive neurons that have the same receptor. So you can think of that code for the same kind of chemical feature. If you’re into chemistry, maybe it’s in an aldehyde group or particular aspect of a benzene ring or something like that, or you can think of maybe slack is it [inaudible 00:26:12] they’d code for fruitiness or sweeteners or some ASCO, flora, and [inaudible 00:26:17]. All these receptors, the receptor neurons, they sent their axons into one or a few small distinct structures in the olfactory bulb.

Andreas Schaefer:
And there they make connections. They make synapses with a couple of different types of neurons, but interestingly enough, there are some already make synapses directly with those neurons that send information to other brain regions that have a protection neurons, mitral and tufted cells. And so they sent their sense and non-information to for example, the piriform cortex, which is the structure of the mammalian brain that’s thought to be involved in categorizing or memorizing or recalling odor memories. But they also send their axons into different other structures like part of the cortex amygdala. So the region dealing with emotion, anxiety, things like that, they send direct connections into part of the bootcamp information into rhinal cortex, possibly to do with some kind of some other aspects of memory and categorization. But also in a few other smaller substructures and within the olfactory bulbs.

Andreas Schaefer:
So this is only a sort of two steps down from the nose, but within the olfactory bulb there are approximately a million neurons in a mouse. And I think probably roughly the same in humans and these million neurons do some kind of computation. They implement some kind of algorithms that allows animals to extract specific type of information from this rounding. But I think it’s fair to say we’ve spent probably about 20 years or so working on a variety of those local [inaudible 00:27:40] in the olfactory bulb. And it’s fair to say that we don’t really understand what kind of information they are extracting, how they allow animals to get a better picture of what the world is. But the general architecture, you have the nose receptors in the nose send information to the olfactory bulb, from there it gets processed in some way and sends information to a variety of downstream areas across the brain.

Matt Angle:
Dima, is there anything you’d want to add to that?

Dima Rinberg:
All information obviously is on the receptor type and you ask what we should learn. It very colorful, the receptors. I think it’s a similar again, the analogy we would be able to make a transformation from the receptors to the perception is the same thing. We will be able to predict the colors from there, from the spectral. We will know the spectral property of individual color is helpers and from there we can predict the color, color is the perceptual variable and spectrum of the physical variable. The same thing here, we have the odors that’s a physical variables, different odor concentrations. And if we know the property of the odor receptors, then maybe that will be enough to build the transformation to the perception. In ideal situation, we can build a big table, which is worrying but maybe it will be able to make as early like a color and the color thing that we can predict perception doing having the receptor properly.

Dima Rinberg:
And in this case, recording for many, many odors for many receptors is what actually also tells how actually works. And one of my excitement about the new adventure or my electronic nose and is that the technologies that we’re developing will allow us to record those from just small number of receptors, huge numbers of receptors. I don’t want to get to say that number, but to cover will they probably have machine interface to connect to much larger number of receptors. And that will give us unique information that will help us not only to solve practical question of device specific odors [inaudible 00:29:52].

Matt Angle:
And if I smell like my beer or a lemon or some normal smell, is there usually one receptor in the nose that’s picking that up, or does it activate all of them, or what’s generally happening when you smell a natural odor?

Andreas Schaefer:
I usually give my students kind of an example of one odor that activates one specific receptor and if that receptor in the human exchange, people smell, perceive that odor as different. They understand that is one example for some people it smells beautiful floral, others don’t smell, others think that smells like urine. Beautiful example, really not typical. Most odors engage with most even single types of molecules engage tens to hundreds of receptors to varying degrees. Only the combination of these activity patterns actually encodes for a specific odor, but even more complicated any your beer, I mean, you might have a Belgium beer, I assume that probably has a rich number of different flavors that come in and so hundreds of different [crosstalk 00:30:58].

Matt Angle:
Exactly. Whereas as German beer would be what, I guess what you’d call a mono molecular or [inaudible 00:31:02].

Andreas Schaefer:
Beautiful in its purity. Exactly. So yeah, but even as in any normal natural smell will have tens hundreds, sometimes thousands of different components, which makes actually the job of a perfume is still so very tricky. Do you put together mixtures of odors and they start smelling completely different in a still a very poorly predictable or predictable [inaudible 00:31:23].

Matt Angle:
Gab that sounds very different than a man-made sensor array. Usually an engineered sensor array has individual sensors with high specificity, high affinity. I’m curious here, how do you think about that when you’re using the olfactory system in an engineering context?

Gabriel Lavella:
It’s really a shift in mindset, right? I mean, it’s different from creating an essay type sensor where you’re looking at a specific binding pattern, right? A molecule A binds to molecule B or surface A binds to surface B and you get some sort of signal out, right? It’s a definitive molecule that you’re detecting or different techniques like I’m ability spectroscopy or other forms of spectroscopy. You’re looking at individual molecules, right? And just starting the individual molecule when you migrate to the olfactory bulb and you have this sort of binding affinity where an individual molecule can bind as Andreas said to scores of different receptors with a different binding affinity, with a different pay on K off, right? That really distinguishes what that code is, right? So it’s a huge shift in mindset.

Gabriel Lavella:
And I think there’s something more elegant about it in the sense that with this set of receptors that have been engineered by evolution over the course of tens of millions, hundreds of millions of years, you’re able to detect a wealth of different molecules by not having this high specificity for one molecule. I think that’s really remarkable and I think it’s a huge shift in mindset and it’s been something that’s been extremely difficult to reproduce. So even finding a surface that can possess those different affinities has been an extremely tall odor. They call it the field of machine olfaction and machine olfaction has been around for 30 years or so it’s made progress, but it hasn’t come anywhere close to doing what the nose can do.

Matt Angle:
Andreas and Dima, what’s the most controversial discussion right now going on in olfactory physiology? What’s the most heated topic?

Andreas Schaefer:
By primacy coding is wrong. No.

Matt Angle:
Well, let me say it’s Andreas who said that you need hundreds of receptors to smell the beer. And I’m saying that maybe 20 would be enough.

Andreas Schaefer:
I didn’t say that eight hundreds, I just say beer will activate hundreds, probably three or four receptors are enough to distinct, I mean, to me, I think it’s been certainly my specific view, to me the most controversial thing is that lots of people study kind of what’s happening in cortex. Lots of people study what’s happening in the olfactory bulb but these sort of elephant in the room is that we have a million neurons that do some kind of computation and they must be there for a purpose, but actually all behavioral tasks people give animals and humans can be solved without any of all these computations. So I think we as a field are just tackling how olfactory works without really acknowledging the complexity of the challenges the system is able to solve.

Andreas Schaefer:
I like to think it’s a bit like studying the visual system like you say what the visual system does is it distinguishes colors. Actually, it’s a small aspect of what it is, but it does shapes its depth perception, recognizes faces all that. And I think all of these like how do you navigate? How do you find the distance of an odor source? How do you figure out which parts of the stimulus belong to a given source? All that from my view is that there’s a lot of information beyond just the chemical, just which odor molecule or which composite odor it is. It’s like what’s the temporal structure of the whiffs of an odor that go to, what can you extract from all this information? That’s what actually requires a lot of computation.

Andreas Schaefer:
So I think maybe the controversy is from my perspective is the more an elephant in the room that we have a million neurons and no one knows what they’re doing.

Matt Angle:
Do you agree with that statement Dima?

Dima Rinberg:
I kind of agree with this and I would like to add something else that there’s a cultural study and kind of option that when we absorb let’s say the disposal of the [inaudible 00:35:49] all these receptors all other, everybody thinks that every response is somehow necessary for something, but maybe in reality, a brain needs to read this information, brain needs only a small portion of this information. And how this is sorted it’s actually fascinating question and not clear what is actually useful, what is not useful because when you… Let me give you an again, analogy with their visual, our brain is very good at recognizing faces, but it’s much more important to look at the person to the eyes and see the features near the eyes and how the shape of the eyes of let’s say the ears or the ears case can be changed.

Dima Rinberg:
Somebody don’t have ears or something like that. So different features have different weight and how this is processed, how the elements is processed and what is the relevance enough of this system is unexplored territory, partially because we didn’t have tools, we didn’t have approaches. And maybe the complexity of the computational in the olfactory bulb to the cortex came with the judging and manipulating the cell levels. For example, what animal need to mitigate, it needs to first understand that this is all there, and then a way to use and then act on it. But Andreas may say something else on that regard.

Andreas Schaefer:
No, I fully agree. I mean, we need to understand what is actually, what are the stimuli, what are the tasks that animals are interested in, in odor to figure out then how information is encoded and process. And I think the hints can come from very different.

Andreas Schaefer:
The can come from looking at the pathology of when animals are doing, can be kind of bottom up looking at the circuitry and what the circuitry actually does. And certainly the sort of tools of like Dima’s work, selectively actually manipulating the representation of a stimulus and see how do you change the representation of a stimulus to the point that the animal perceives is that being still the same or different are the kind of tools that allow us to understand what’s represented. Still think the complexity of stimuli we’ve been giving is just completely under challenging the system, just as a small snippet of what animals are normally dealing with.

Matt Angle:
I’m really glad that you talked about tools because that was the next thing I wanted to ask you about. What is the state of the art and recording the neural activity associated with smell and decoding those recordings? What are the modalities for getting the data? How well can we do reconstructing what an animal’s knowing?

Andreas Schaefer:
I’m happy to start and then Dima can consider continue. So the tools that we use to study a neural activity in the mouse brain let’s say, and to study smells, it’s not so different than for other systems neuroscience guess. Maybe we can use imaging tools to look at individual neurons. We can use electrophysiological recordings. So the differences are a little bit probably on the subtle side, imaging we have the advantage that the main structure we’re interested in is actually superficial. We don’t have to kind of go through a large parts of the brain to actually access those neurons. It also means we can not only record, but you can also stimulate selectively individual regions or even down to individual cells quite effectively. On the electrophysiological side, maybe from using traditional electrical recording techniques, part recording.

Andreas Schaefer:
So very targeted recordings from individual cells quite well because the structure is quite superficial. Doing extracellular recordings is maybe a bit more tricky in the olfactory bulb to do on a large scale than it is in other regions because traditional high channel count recording techniques rely on Silicon probes where most of the recording sites sit along one or a few sort of linear axes. Whereas most of our cells, we are interested in studying the population of the output should sit in sort of a sheath. So recording techniques like the ones developing or dynamics or what my lab has been working on for a while, where you have bundles of wires are more effective to capture a large number of those output neurons.

Andreas Schaefer:
I think what we are lacking in terms of ability to understand recordings compared to other brain regions is probably a good understanding of the composition of different neurons in most neo cortical areas, people understand quite well. There are some other sort of static positive inter neurons that [inaudible 00:40:25] neurons, they’re different types of parameter neurons, and we don’t have very good genetic markers or understanding what these different cell types are doing. So we generally tend to record, we tend to lump together different cell types that prompt different cells that probably correspond to different cell types and possibly different roles. At this point, it’s only slowly that we start getting a handle on what types of cells are there, which will help us to mechanistically understand what’s what’s going on.

Dima Rinberg:
I’m very much looking for the big screen of a large number of receptors for many odors. And here I actually would like to let Gab talk about, or is it mainly developed for that because it’s a new alternative typology that is when we do imaging, we limiting ourselves to the official part of the bulb, and we can record the acuity of the top layer that has actually one mirror layer that is a separate layer or the very first processing layer, but Gab is proposing and what we’ve kind of developed it for the company together with developing the alternative way to measure large number of receptors…

Gabriel Lavella:
We’re basically taking an electrode based approach where we’re using a electrode array to measure local field potentials. So we have a grid of electrodes. We’re trying to get the pitch of those electrodes to a point where we’re able to have singular or gamma array resolution and we can watch, but especially in temporarily the signals coming in into the olfactory bulb, right? So we’re intercepting it at a point before a lot of this processing takes place and the signals get filtered and sent to the brain. So it’s a way of looking at the outside of the olfactory bulb. We’re doing it electrically, but there’s optical techniques as well. Well known techniques, calcium imaging which gives you fantastic spatial resolution to record the signals. From our perspective, the two most important tools that we’re using are calcium imaging and doing electronic recording choosing brace, micro ECAP type brace.

Dima Rinberg:
Just to add to Gab, so very interesting aspect of the grids that we almost hoping it’s one moment to wrap almost all the above. I don’t think we can wrap the whole bulb, almost all the bulb in these greets and collect information that will be extremely hard to get to any imaging recording. This is why I am so excited about this technology to do that [inaudible 00:43:09] with you.

Matt Angle:
I wanted to ask in terms of the coverage of the bulb and odor space that we’re talking about there. Do you feel like you can cover all the odor space? And do you think you have to?

Gabriel Lavella:
Dima, we’ve had a lot of conversations about this and it’s very key to how we’re going to develop our technology. Do you want to tackle it?

Dima Rinberg:
So we started with 64 electrodes on the surface of the Bulb and we were able to detect some odors, disagree on some odors. It’s definitely not sufficient to get the comprehensive picture or the odor space. So imagine if you can get an actual surface electrode throughout all the bulb, maybe 60% of the bulb. We can put it on the lateral side or the dorsal side, if you require a plastic surgical skills, but it didn’t the realm of possibilities.

Dima Rinberg:
Now, the technology became available to collect so much electrophysiological data and we probably would reach full coverage, but completely whole bulb, but to access a large portion of the bulb would be white also. Now, if we can do it, we can also start buying different [geeks 00:44:21]. For example, if we missing cell specific receptors and electrode, our genetic tools will allow us to move receptors under the electronics and the work of my colleagues who are both geneticists [inaudible 00:44:40], they’re working together [inaudible 00:44:43]. There is Tombosa who can over express specific receptor on the doors so part of the bulb would move the receptor to position different locals and combination of genetic depletion and [inaudible 00:45:00] ecology is just open amazing [inaudible 00:45:04].

Andreas Schaefer:
It’s not inconceivable to even reach the ventral side of the olfactory bulb, this the output neurons with pre-filter on imaging or a combination of ultrasound imaging, et cetera, not for BCI or BMI, but from a research perspective. It’s the advantage is the structure we’re looking at is indeed really directly under the skull and size is on the odor of maybe two millimeters or so we would have to go through two and a half millimeters. So it’s still so difficult, but the next couple of years might make it possible to access the entire olfactory bulb if there’s a good scientific driving questions to do it.

Gabriel Lavella:
If we had a microwire technique or a nanowire technique where we could use something penetrating, or that was of sufficient scale, we might be able to get access to those globular like complexes on the ventral side, but what the technology that we’re developing, we don’t plan to get access because of that innovation that’s coming through the cribriform plate up into the bottom of the ventral side of the bulb. If we try to cover anything around it, we would separate those connections. So we’re trying to get as much of the bulb covered as possible using an electrical technique.

Matt Angle:
What about expanding the receptor repertoire? Is there any interest in expressing exogenous receptors like I don’t gum receptor or whatever it is that the military is interested in? Have there been examples of dropping in engineered receptors into the olfactory system?

Gabriel Lavella:
Have been early on also in demonstrating that, and he sort of G protein coupled receptors, these sort of general class of receptors called can actually be used to instruct formation of [dymerurile 00:46:51]. I don’t think in a mouse that has been much drive so far to put in different receptors in odor to detect the new things, simply because the rapid while the thousand receptors is already capable of detecting a very large fraction of all chemical bulletiles.

Dima Rinberg:
I give you one interesting example and that’s the work has been done recently preparation with on Basel up. He takes one specific receptor and over express it significantly. Kind of the number of liminal instead of two on the same of bulb was 80. And he measured the behavior of the show to do that specifically legal. This is most [inaudible 00:47:33] special doesn’t change. Adding more granular doesn’t make you more sensitive. It’s kind of cultural integrity, but well, that means that our olfactory system is actually so well debunked, and we’ll keep it at it’s best performance so it’s very hard to improve it [crosstalk 00:47:53].

Matt Angle:
Unfortunately, I can’t smell carbon monoxide. I wish I had a receptor for carbon monoxide, especially now that everyone’s houses are shutting down and we’re we’re heating our houses by the fireplace. Odor space doesn’t encompass every molecule [crosstalk 00:48:13].

Andreas Schaefer:
You’re right. I think it’s well possible that for certain things that indeed do not carry any smell on the other hand, I think would be a stretch to at this point to think of many molecules that are beyond the detectability of a mouse olfactory receptor repertoires, if you think of BCIS from also factually in odor to detect specific patterns of volatile chemicals that might be associated with specific disease or a specific explosive or something. My sense at this point would be that all these carry enough information across these repertoires of a thousand existing receptors, that the challenge will not be engineering a new receptor to plug it in and allow for detection, but actually to stably and robustly acquire the data along the spirit of what Gab and Dima are working on in odor to then do robustly decode information from that.

Andreas Schaefer:
But it’s certainly why it conceivable that the examples that Dima gave, I mean, I think for the last 25 years, people have been working on engineering specific receptors in the mouse, moving Lamar light a little bit, replacing receptors, studying the difference in point mutations and receptors. That is a relatively tangible challenge to then replace individual receptors, create new glomerular to create new receptors. So if one had a good reason, that would probably be a relatively simple problem compared to some others.

Gabriel Lavella:
Yeah. There was a paper out not too long ago the McGahn Paper, he was talking about some of that specific and ask me as carbon monoxide is one oxygen you can’t smell of course. As Andres mentioned, there’s very few of these molecules that are bigger than one or two apps. Once you start crossing that threshold of two atoms, there is usually a receptor that’s capable of picking it up and the olfactory system is capable of recognizing it.

Andreas Schaefer:
But certainly the flip side of the fact that mammals talking about mice again, that mice can detect almost any volatiles that it makes behavioral experiments with these little chaps, extremely challenging. You really have to think of all the possible controls to make sure that they’re not able to detect whatever [inaudible 00:50:32] there might be. Remember, I think one of my first extensive behavioral experiments, I spent a work through the night to learn that a mouse that I was training, it could smell what I had touched the tubing from the outside or not. So the subtle differences of having touched one tubing and not the other was something the animal readily learned within a few tens of minutes.

Andreas Schaefer:
So you always need to really make sure you do the right kind of controls. The great thing about that notion is of course, that any set of volatile chemicals that differs from any other set of volatile chemicals, for example, a cancer, one sort of kinds of a different sword and the breadth of people. In principle, that difference in smells of the difference. If there’s a difference in those volatile chemicals as they will be for many of those disease that is something that mice should be able to learn to detect, or even more simply, if you’re able to record directly, you should be able to decode reliably. So I think that’s quite an extremely exciting perspective.

Matt Angle:
Okay. What would the implications be if we had a high data rate BCI in the olfactory bulb of a mammal that we could take around wherever we want?

Gabriel Lavella:
One of the things that it would enable us to do is to detect the hidden fingerprints of all sorts of things that aren’t apparent with optical systems. So let me give an example. It’s kind of like the advent of machine vision. If you have machine vision and you look in the back of a truck, you could see in the back of that truck, “Hey, there’s 10 boxes back here. There’s three pallets. These are the sizes of the boxes.” If you had a system that was capable of this broad diversity and sensitivity for airborne molecular detection, you would be able to tell things like there’s an orange in that box. You might be able to tell things like it’s not just an orange in that box, but the orange in the box is from Valencia. So different oranges have different sense. You’ll be able to tell, is there an invasive species onboard, are there amphetamines present?

Gabriel Lavella:
Where was the truck previously at? So it’s picking up all these senses, strive it. All of these chemical signals are contained within that sense space. With a sensor like this, it would be phenomenally incredible. It would impact a lot of different industries. It would impact the food and beverage industry certainly, human safety and health. It would impact medical diagnostics. We know that a lot of different diseases have unique sense. We know well, COVID, right? You’ve seen all the papers that have come out on COVID detection with dogs, detecting it with the extremely high fidelity and reliability and third picking up the volatile organic compounds of those diseases. So all of these things in the world, but manmade and natural, a huge number, if not the majority of them have a unique fingerprint and that’s becoming really critical in the world, right?

Gabriel Lavella:
It’s becoming critical because we have the massive movements of people and goods across borders. And as a result of this, that leaves a lot of vulnerability, but we also have material goods that are coming across borders that are dangerous in a lot of ways. We introduced invasive species here in Pennsylvania. We have nightmare scenarios right now because of all the different invasive species that have come over from Asia are destroying the wildlife here and crops and things like that. So there’s really no detector that’s good enough to do that. A system like this could potentially pick up those things and get to have a lot of market done and could impact a lot of industries, kind of a long-winded answer, but I think it can be extremely impactful to the world.

Matt Angle:
I think at this point there are actually a fair number of investors that tune in and I think many of them would be really interested to hear what is the first thing? What’s the killer app? What’s the big market that’s accessible to olfactory B style.

Gabriel Lavella:
I think the first market is one that has low friction, a market where animals are already of use and that’s important border inspections, cargo, container inspections, domestic policing, domestic security. Those are areas where they’re already using a host of different animals. They understand the technology and I think they would be early adopters and willing to use it and greatly benefit from it.

Matt Angle:
Do you have a sense of what the market size for that is?

Gabriel Lavella:
It’s huge and it’s deceptively huge. In the United States there’s about 50,000 or so detection dogs that are deployed that’s doesn’t include the private space for private sector use. Worldwide, It’s probably around 150,000. That’s the number of units that are out there. Animal units that are used in detection. The market size is in the billions easily in the billions.

Matt Angle:
What would it mean for the people who are at the border using dogs right now, if they can read their dog’s mind or rat’s mind or rabbit’s mind or whatever animals?

Gabriel Lavella:
It significantly enhance their ability to prevent illicit materials and contraband from entering the country. And unless if materials could be things of agricultural importance, right? It could be a new invasive species. It could be meat products that should not be entering the country. It could be a new form of explosives in our meetings with different security forces in the EU had informed us that a lot of different terrorist organizations switch up the chemical compounds because they know existing ones might be detected. They’re clever enough to re-engineer the explosives to get them through. It might be narcotics, different forms of drugs, cash, electronics, all sorts of things. [crosstalk 00:56:25].

Matt Angle:
In analogy to a discussion we had on the podcast earlier about speech decoding and some of the work by for instance, Eddie Chang’s Lab. We talked about the difference between kind of an open set and closed set, and a difference between trying to read speech in the most general terms where someone can think any words they want and continuously decode the brain activity associated with thinking about words versus training a classifier to tell the difference between cat and dog. I’m curious, when you think about what the olfactory BCI would look like, do you think that it’s going to be more the case that you’re training on specific compounds or training specific classification tasks? Or do you think you’re going to be able to have a catalog and just go empty? That spoiled me, that’s an invasive beetle. More specifically pick out the underlying odorants.

Gabriel Lavella:
Well, start by answering it. Then I’m going to pass it up to Dima who will probably have a much more precise answer. One of the things that we can do and they do right now with detection dogs is they’ll train on a group of elements, right? And they’ll put them all together. We can do that, right? So with explosives there’s 19,000 different types of known explosives that are used and things that are used to make things worse, that could be separated and then shipped. So it’s an extremely large number being able to detect each one specifically informing a catalog is one of our goals. We want to be able to have that sort of resolution to say specifically what it is. We’re using a technique called federated learning. The idea is that training does not need to take place in one area, animals all over the world can be trained on a scent and then every other animal will be able to then detect that set. That’s one of our goals.

Matt Angle:
Does the area type to the brain maps across animals of the same species

Dima Rinberg:
From the engineering perspective, imagine if you have a set of identical receptors let’s say, and each you detect the signature of one odor and another on one detector you can cause for the signature different detectors and detected with different constraints. Let’s say all guests chromatographer can be synchronized. You can record that the profile of very complex one on a one GC machine. And now that would be more or less the same, with animals it’s harder. But we do believe that by calibrating and kind of synchronizing two different pretty much interfaces or one and another animal using sub sets of odors as the fiduciaries or anchors, we can align each other noses and make them all it’s walking syncs, or when a new odor will be taken by one machine can detect a signature by another machine. By now that animal that’s to be discovered, will be done.

Dima Rinberg:
From a biological perspective, we all was can be trained to discriminate [foreign 00:59:40] versus [inaudible 00:59:41] and we can do it more or less the same way. And we have our [inaudible 00:59:47] slightly different, but overall it’s the same because we grew up in the same environment. We know the same odors, we’ve told that this is that this is apple and this is orange. So your impression when you smell a new odor, positioning this odor in space from other odors will be very more or less the same like mine, because we’ve been trained on the same set of odor initiative. And that’s actually give me a lot of hope that we can reproduce with man machine interface to be determined.

Gabriel Lavella:
And I will add this quickly in terms of building that catalog, it’s something that’s continually expanding. Every time we upload the profile for a new detective scent, that’s permanently there, this sensor is continually learning. It’s speculated that animals can detect up to billions of sense. I’m not going to say trillions, but billions of different sense. So given an infinite amount of time that cattle up and keep growing and growing and [crosstalk 01:00:43].

Matt Angle:
You’re referencing a paper by Leslie Vosshall, right?

Gabriel Lavella:
Yes.

Matt Angle:
I think that Rick Gerkin actually wrote a response to that paper pointing out perhaps some deficits with it. Do we think that there are trillions of utterance or do we think that it’s a potentially a smaller number?

Andreas Schaefer:
Matt, I know you like controversy and I’m happy to step in here. I think it’s a brilliant paper from Andreas Keller and Leslie Vosshall and several others at Rockefeller. It’s a beautiful of a paper with a very systematic psychophysics in humans about the human ability to discriminate between different mixtures. On some level that’s sort of maybe on a broader scale of things it’s quite a technicality because what’s I think really also very interesting what the papers of scientific discourse, because they did the amazing thing and actually immediately publishing the entire dataset. They collected every single bit of raw data together with that paper. So everyone could look at it and try to understand exactly how did they come to their conclusion. And they were quite quickly afterwards two very interesting discussions emerging from that paper. One was along the line of discussing the specific statistics in the paper, which was very useful thing.

Andreas Schaefer:
And that was the one Mr. Castro and Gerkin you mentioned, the other paper that discussed the results of this manuscript and the data interpretation has been Markus Meister and they’ve looked at what are the underlying assumptions. And I think that turned out to me it’d being the most dramatic conclusion of that paper and the following discussion that it comes back to do a discussion we had here a little bit ago about the dimensionality of odors space, because the conclusions that also Keller and colleagues drew from that they are sort of very clean dataset. The conclusion they drew was based on the underlying assumption that odor dimensionalities is relatively high, it’s sort of tens of dimensions, which I think everyone probably at that point in time with a study as sure that’s the way have hundreds of thousands of receptors.

Andreas Schaefer:
So it probably will be that without having thought about it in great depth, making this assumption, they concluded that they could derive from the data. Well, that means our data means that humans should be able to discover or to discriminate trillions of odors. It turns out that to date even, and no one has really figured out what that dimensionality of odor space really is. So I think to me, that paper, I think is a particularly good example that transparently publishing work with data allows you to distinguish between the actual data and its conclusions quite well, and maybe derive very new, quite fundamental insights. The fundamental insights coming from that was that we need to understand better dimensionality of odor space. And there’s still ongoing discussions about which are the best experiments to actually tackle. So the conclusion certainly from today’s perspective as well, we still don’t know whether humans can distinguish trillions of odors So that kind of you could say, well, that’s a superficial sense the conclusion of that paper is wrong, but I think it’s much, much deeper and more informative than that.

Matt Angle:
Leslie Vosshall was very active on Twitter. And now you just said, the conclusion was wrong. I’m going to get skewered Andreas.

Andreas Schaefer:
That’s fine. I think she agrees that, that superficially that’s concluded as wrong.

Dima Rinberg:
I would add even a little bit more. I very much agree with Andreas take of this paper because it’s fantastic data set, but this brings us again to the dimensionality and the fact that dimensionality actually affect the exponent there, so their estimate was humans can just smell 10 to the power of 12 odors, but this number 12, depending on the dimensionality can vary from two to 22. So that actually worrisome because you don’t know the dimensionality of this exponent is not very, very precise. So I’m very happy for that paper because it provokes so much thinking, but I’m not very happy about this paper because I started reading in the books already that humans can discriminate even odors and that’s propagating of the meme that probably not yet scientifically established.

Andreas Schaefer:
So meant for the editing. My conclusion is should actually be the conclusion of the paper at this point it’s still unclear whether it’s trillions or not, not necessarily that the conclusion is wrong. I mean, no one has shown that it’s not true.

Dima Rinberg:
It’s like a clock that doesn’t work show twice a day the same time.

Andreas Schaefer:
It’s a funny anecdote with respect to the truly notice two three years ago. So my lab and a couple of neighboring labs went for kind of a small day out. And one thing we did, we went to an indoor mini golf place. And it was one of those where at every stage of the mini golf, if you had the option to answer a question on the screen, if you got the answer right, you could go a simpler course. If you got it wrong, you have to go the more difficult course. So we went and you got all kinds of answers, who was, which actor in EastEnders and what’s the capital of Idaho or whatever. And when we went to one stage, we looked up at the question was humans can discriminate trillion odors, true or false? At the indoor mini golf court someone like what city London. So we said, “Nope, that’s not necessarily true.” So we went there, we were told it was correct. We went to the manager and started having a discussion about the money. It didn’t help us, but it was certainly quite a fun event for everybody.

Dima Rinberg:
But we actually know I deserve a day, the birth of a new meme. This is actually fascinating phenomena.

Matt Angle:
I think some people will be very interested in learning more about olfaction. Are there references? Are there places you’d want to point them like a review or a textbook or a blog? Where should they go if they want to learn more about olfaction?

Dima Rinberg:
It’s been sort of a recent issue of cell tissue research general with the dozens of different reviews of the state of the art of olfaction. It’s very up to date. It’s sort of came out, I think January, February 2021. So just now, or a few weeks ago, which might be a good starting point, but I would say, is it a big good and fun introduction into a little bit. Some people that do olfaction and the history of olfactory research has this book from Ann-Sophie Barwich about the Philosophy or History of Olfaction Smellosophy, what the nose tells the mind.

Gabriel Lavella:
If you’re looking for an extremely comprehensive book on odors, The Springer Handbook of Odors 2017 edition, if you want a meaty 1500 H paper that categorizes every type of odor out there, it’s a preference, I want to read it like a novel that for sure.

Matt Angle:
Okay.

Dima Rinberg:
Olfaction is exciting, very exciting area and often underappreciated, very sickly people much more attractive to vision to audition in neuroscience. I think we can address the fundamental question in neuroscience and relatively simple neurologists here. We can ID everything from stimulus to behavior. It’s a fascinating system and really under appreciated. So expect more people to join the pools. The more the merrier. It’s very interesting [inaudible 01:07:56].

Andreas Schaefer:
Because I think it’s fair to say that if someone’s really keen on understanding, how will the mammalian brain does something and wants to get a really mechanistic understanding, figure out how cells direct, how some things encoded and not predominantly describe that? I think olfaction gives you that compactness, gives you accessibility, small volumes, small number of neurons, yet doing the most complex tasks.

Dima Rinberg:
I would even kind of challenge people in the following ways, even understand how brain works can really produce the percept. Can we be one step closer to the metrics movie? Can we create a percept in our brain? And if you ask me in which system you can recreate the percept, I will bet on olfaction, but maybe we’re wrong. So this is things to test. So we’re working on this, it’s exciting field.

Matt Angle:
Gab, if there are any investors on the call who are really interested in this idea of getting a digital fingerprint of olfaction, how would they get in touch with you?

Gabriel Lavella:
They can reach out to me through my LinkedIn, or they could get in touch with me at gabriel@commonsense.co.

Matt Angle:
Perfect.

Gabriel Lavella:
Or they could reach out to the Indie Bio Program or for me directly. We’re probably going to be hiring later this year. People who are interested in this technology and interested in some of the impacts that the scientific side of things, the technological development of things, feel free to reach out early and get in touch and learn more about us. There’s a lot of interesting problems to be solved here because this technology has never been released in the world. Nobody’s ever had a real time, portable sensor out there, pulling down this massive quantity of molecular data. And I think we’re going to start to uncover some really interesting things about the world from that across different industries, from pollutants in the environment and how they affect all sorts of different things. We feel it’s a very meaningful endeavor and feel it’s going to be a profitable endeavor too.

Dima Rinberg:
I would add something that we’re planning to do some scientific development in the very basic staff at NYU in my lab, and then expand knowledge to the company outside of the lab. So if people are interested in the first stage in front of this technology, just contact me, look at my webpage rinberg.com. This specifically is let’s try this project as via electronic.

Gabriel Lavella:
And if you want one controversial statement, Matt.

Matt Angle:
Which I would love, yes sir.

Gabriel Lavella:
My wager is going to be that in the longterm using a BMI, we’re going to be able to detect things better than a trained animal. The reason for this doesn’t have to do with the filtering new olfactory bulb, it has to do with the kinetics of gas flow. Detect things that are great distance by having an a priori understanding of the speed at which those molecules arrived. So the pattern might not exist there in the dog’s mind to be able to pick up. It might not have a holistic pattern that it’s detecting in one sniff, but that signature might be stretched out in time and we might be able to pick that up.

Matt Angle:
That’s interesting. Andreas, I think you’ve done a little bit about odor localization in your lab.

Andreas Schaefer:
We’re looking at odor plumes. So not only the chemical structure of odors, but actually they’re the temples of just how concentrations change rapidly over time. And we find that mice can actually detect these very rapid fluctuations at timescales of tens of milliseconds or faster. And there’s a lot of information that, which we know from the difficult study of turbulent airflow. But if we then do some kind of population recordings and olfactory bulb be it the electrophysiological means or imaging, so we can quite easily pick up signatures about the temporal structures of odors with very, very high precision, much more easily than we have been training animals and performing the same kind of discrimination. So my sense from all those combination of complex behavioral experiments and different physiologically experiments and the variety of linear nonlinear classifiers thrown a data, certainly that’s as soon as you start having tens or at least hundreds of recording sites, almost inevitably are much better in decoding the specific type of olfactory stimulus then animals show in a trained behavior if we train them over periods of days a week. So I’d be very much with you guys and saying that you can expect much more from the actual activity pattern that you’ll be able to record, then you can train animals to report behavior.

Matt Angle:
Gab, that must feel good because your bet turned out to be right within about five minutes of making it.

Andreas Schaefer:
Yeah. I’m glad you consider my statement to be a final judge on Gab’s bet.

Gabriel Lavella:
I was really kind of referring to things that and I think that definitely adds to it, but things that are very far off and adjustments like if you have a dog that’s trying to detect the scat of a whale two kilometers off shore, and the identifying pattern which represents that scat is coming in very slowly, right? Like two minutes later, 50 seconds later, right? Not something that’s in a small temporal fluctuation, but something that’s [crosstalk 01:13:21] time.

Matt Angle:
So gas chromatography, where the medium for chromatography is the air and the distance is very long.

Gabriel Lavella:
Yeah.

Andreas Schaefer:
And you can computationally integrate over much longer time scales that animals would typically integrate. [crosstalk 01:13:34]

Gabriel Lavella:
Exactly.

Andreas Schaefer:
We would expect that for normal behavior, you need to integrate over maybe a few seconds, but not tens of seconds a minute.

Dima Rinberg:
I would say that the challenge here, to my opinion, I think that you can train animal specifically to detect one specific feature with signal. And if the task will be to integrate or just opposite to take away brief flu, you can train an animal there. The advantage of having brain machine interface that we don’t need to train. We can collect all information and then os facto extract what we want whenever we want and as much as physical limits allows.

Matt Angle:
And those models are transferable between animals too as long as you have… As you were saying the fiduciary markers.

Dima Rinberg:
Yes. The physical limit probably any specific detection task can be achieved by the animal as good as with BMI, but flexibility between the tasks and flexibility between the targets is what is very hard, but you will be [inaudible 01:14:37] and get multidimensional report.

Gabriel Lavella:
One very, very interesting aspect of the digitized sense where else is that once you have it digitized, you have the ability to go back in time and parse that data for signatures, which you didn’t currently have at the time of recording the data. So if you’re looking for, when did a new virus cross through this airport and you’re looking for the fingerprint of that virus, you can go back, parse the data and see if there’s any recording of that virus. And when it happened and where it happened, that’s a really, really useful feature of digitizing the sense effects.

Matt Angle:
Well, thank you for taking the time with me today. I think people are going to really find this interesting.

Gabriel Lavella:
Thanks. [crosstalk 01:15:22].

Dima Rinberg:
Thank you Matt. Thanks [inaudible 01:15:22].

‍