Awesome. Thanks very much, everybody. It's my pleasure to be moderating a panel with Amit Etkin, founder and CEO of Alto Neuroscience. I'm sure folks listening in probably have heard the Alto story a bit, but just to orient things in a minute, maybe you can give us a quick sort of background on the company, you know, your novel development approach, and maybe orient things with ALTO-100 and ALTO-300, and then I can get into more specifics. So thank you again.
Perfect. Happy to. And thank you for having us on. So the core approach with Alto is around precision psychiatry. It's basically the idea that we have these very large, heterogeneous groups clinically or biologically in any psychiatric disorder, and never to date have we actually used any biology for understanding who it is that actually benefits from a drug, even though we know that heterogeneity is there. And so our approach is all about identifying a biomarker, and I can say more in a second about what a biomarker here means, that selects patients who respond better to one of our drugs, which attack various types of subgroups through a series of different novel mechanisms. So you mentioned ALTO-100 and ALTO-300. They touch on two of our biomarkers.
One of them is a cognitive test, an objective test of memory, which identifies patients with low hippocampal plasticity as part of their depression. That drives cognitive problems and mood problems, and the drug enhances hippocampal plasticity, and it's currently in a Phase 2b. ALTO-300, which is a drug that is actually a known antidepressant and approved in Europe and in Australia, but an NCE here in the United States. There we're using an EEG biomarker that identifies who responds best. On top of those two classes of biomarkers, we use wearables as well. So really anything objective about the brain, about people that's scalable, that we can take to large commercial scale, like the 20 million people in the United States alone who have depression, are how we kind of formulate our biomarker platform.
Excellent. Okay. So as you built out your pipeline, right, you've identified assets either via, like, a preexisting hypothesis or, you know, maybe you just kind of run it through the platform, so to speak. And so the lead here is ALTO-100. That's got some data second half of this year. So do you want to talk about the backstory of this drug? You acquired it from Neuralstem. You know, their work in depression was unsuccessful. What stood out about their data and the biological hypothesis that emboldened you to move on this drug?
Yeah. So if you kind of go back to the beginning of why even go after this drug, why go after this population in the first place, we've been working in my lab at Stanford. I was on faculty for a decade prior to founding Alto and then at Alto on a group of patients with depression who have an objective cognitive deficit. So things like memory that's different from a subjective deficit, which is somebody complaining of cognitive problems. Those two things actually align quite poorly. But the folks with objective memory have worse outcome to standard of care treatment. So it's a higher need population, greater chronicity, greater disability.
And what has been thought to drive that, that sort of mechanism, that picture, is a reduction, as I mentioned, in neuroplasticity within the hippocampus, which is important for that function is important for the brain to adapt to changes in, in internal or external stimuli. And so, that thesis has been out there in the field, and Neuralstem designed a molecule that enhances a form of hippocampal plasticity called neurogenesis that takes several weeks to play out. That drug also was found to enhance synaptic plasticity even more acutely. So multiple different time scales of a proplasticity effect fits our bill in terms of the mechanism that we're looking for. Now, you mentioned that they went through a Phase 2 study. That Phase 2 study, unfortunately, was fated for failure. It was powered at a Cohen's d of 0.5.
You'd expect 0.2-0.3 in all-comer populations based on everything we've seen for the history. That's exactly what they saw. Unfortunately, they underpowered it. There were some other factors in the study that didn't help. They did see around a 0.2 effect size. What motivated us in bringing that particular drug in is even just during diligence, we found in a more minimal cognitive battery than ours that the patients with poor cognition, exactly the group we had set out to treat by bringing this mechanism through, in fact, benefited statistically significantly, benefited over placebo and showed a dose-response relationship.
So that's actually a lot of really good information for bringing into our prospective study where we finalize the biomarker as verbal memory, a direct test of hippocampal plasticity, so lining up MOA with how we characterize the patient, and then prospectively replicating to show that, in fact, in yet another population, those patients with, poor memory, this poor cognition phenotype, in fact, respond better and that patients receiving placebo do not respond better. So there's specificity, in this targeted group that lines up with the MOA.
Right. Okay. Makes sense. So as it relates to the cognition biomarker, this verbal memory test, like, what are you getting at exactly? And I think the one interesting element of this that you correct me if I'm wrong, right, but that, you know, when we talk about MDD with poor cognition, this at least is from how you've described it as I understand it, right? It's not a discernible phenotype. It's not like a patient walks into the office and they say, you know, I can't remember this or that. So, you know, maybe talk about the biomarker, but then what are you sort of actually, for lack of a better word, just like getting at in terms of who these patients are and, you know, what the kind of what this cohort really, like, looks like in the real world? Does that make sense?
Yeah, absolutely. So memory is one of these things that if you spend enough time as a clinician, you can pick it up. I mean, it's not amnestic, folks, right? It's not like we're talking about Alzheimer's. But most clinicians don't really have that amount of time. And at more moderate levels of severity, they're going to have a hard time picking it up anyway. I think we have to understand globally that we have approached depression, the field, not Alto, from the lens of symptoms alone. And symptoms actually don't have very much direct relevance to the biology. So somebody could tell you that they have poor cognition or even poor memory. You measure cognition or memory, and those two things are totally uncorrelated. The simplest way to think about it is actually, what about sleep, right?
There's no way somebody can tell you about the structure of their sleep. It's only how they feel about their sleep. That same concept rides all the way through any symptom in depression. So you have to get out of that echo chamber, which has not done us well in the past, and get to an objective biomarker.
Yeah. Okay. And so as it relates to the verbal memory test specifically, you know, what work have you done or have others done to kind of validate that, again, this is a deficiency driven by depression and not a deficiency driven by educational status, IQ, or some other kind of performance-oriented element that would have less validity to the drug hypothesis?
Yeah. So there's a boatload of data starting from the fact that even risk for depression, where there's, you know, you can look at people who have no symptoms, that's associated with lower verbal memory. You can control for all the premorbid, you know, aspects of cognitive functioning, and you still have that. And verbal memory is one of the areas of greatest deficit in depression relative to, you know, other areas. It's something that has also been assessed in a known way for a long time. So we digitized the test, kind of adapted for self-administered testing, a test that's been around for over 80 years and has incredibly high test retest reliability. And you often see that deficit also when the person is well as well as when they're ill. So it's really part and parcel of the disease itself.
Okay. Okay. Got it. And so as it relates to the data so far that you've generated for ALTO- 100, I think there's just some confusion in the investor community about what is or isn't prospective. And, you know, given that these studies are open label, right, like, how much does Alto have a window into how the data are accruing over time and patient cutoffs and things like that? So do you want to just, like, walk through the Alto hypothesis more specifically? Sorry, Alto process. Excuse me. More specifically, right, like, you have a hypothesis. You do this initial test data set. What do you do then in terms of figuring out the right measure, the cutoffs?
And then, you know, to what degree is that truly prospectively tested, you know, with you having no insight into it, you know, in this kind of Phase 2a work before you actually go into a larger study?
Yeah. So let's start with what the study design is. So as you mentioned, it's a single-arm trial. The goal here is to identify who responds better and show and replicate prospectively, truly prospectively, that that's, in fact, the case. And that's best suited to be done most quickly, most directly with a single-arm design. And we have and we can talk about placebo. We have other data sets, plenty of other data sets on cognition and placebo. But what we do is take that Phase 2A study over 200 people, take a portion of it that we call the discovery set to finalize and optimize a biomarker, things like what is the measure, what is the cutoff, how is it being processed?
The rest of the data that is the prospective replication data are locked and blinded and put away, sometimes hasn't even accrued at that point, and nobody has access to it. And so everything that's being done is in that discovery set, blinded to the replication data set. But moreover, you then require a single biomarker to be nominated to then be tested in the prospective test sets. So you have to have a lot of confidence that you're willing to suffer, in fact, a false negative, that you incorrectly reject a potential biomarker in the service of truly having confidence when you replicate prospectively at a sufficient enough effect size for us. So it was a Cohen's d of 0.5 for the magnitude of the enrichment that you've indeed found something.
So, you know, we try to put as many hurdles as we can in front of us for all of our data science and make sure that whatever we rely on is replicated. We don't make any conclusions about either that retrospective Neuralstem data or the discovery data. It's entirely about the prospective data.
Right. Yeah. Okay. Makes sense. So, you know, I think to round out the discussion on 100, I think it's, it's relevant because it becomes a broader sort of platform question. And that is, you know, obviously it'll depend on the Phase 2b data later this year, right, in the placebo-controlled study and how those data look. But, you know, assuming they are positive, how do you think about the regulatory path for what you're working on? Like, to what degree do you have buy-in from the CDER division around this companion diagnostic approach? And, you know, would you actually need to get, like, a diagnostic that is formally reviewed, like a device is reviewed, or, you know, how, how does that kind of work?
Yeah. So to start with, the FDA itself has issued enrichment guidelines in 2019, which we're following. And so there's a lot, you know, fewer uncertainties when you're just following what they actually said. We've also, in a different program as part of the Type C meeting, put forward our approach, exactly what I described, which is find a biomarker, prospectively replicate, run a Phase 2b in that targeted population against placebo. And they were in line with that approach with any particular biomarker. Of course, there's the vagaries of that biomarker. And so for this one, which more closely resembles measures that they already understand, traditional neurocognitive measures, it'll be a different path, most likely than EEG, which most certainly will require a device clearance software as a medical device. But none of those facets actually worry us particularly.
That, that will be work and that will be validation and, and so forth. But most of that data comes as part of your Phase 3 program anyway. And we've already been doing a huge amount of validation on our cognitive battery that people using it in their homes can understand because it's self-administered how to do it. And that we get very reliable results looking at reliability over two months, an ICC of 0.8, what the norms are, the control for age and gender, education, all of these sorts of things. So a lot of that work's already done. And we're going to have a discussion with the FDA as a Type C meeting to further refine what needs to be done in advance of a Phase 2 readout and end of Phase 2 meeting to plan the Phase 3.
Okay. Great. Maybe one last question before we move on to 300 or agomelatine. So on 100, you know, the Phase 2 study, it's placebo-controlled. And you've said that it's powered conservatively for an effect size that is considerably smaller than implied by the open label, which is great. The one nuance of the study, though, I wanted you to kind of expound upon, Amit, is just that you're including some biomarker negative patients in this study, on drug and placebo. Like, why are you doing that? Why is that important? And what are the implications of, you know, showing no benefit there versus showing some benefit or actually showing a similar effect size? Like, how does that kind of play into the overall development and regulatory strategy?
Yeah. So again, that's following the enrichment guidelines. You need to know about risk-benefit considerations for the folks who are not on label, the on-label population here being those with the biomarker, poor cognition patients. That's what's actually powered, as you mentioned, at a Cohen's d of 0.4, which would put us above standard of care but be conservative relative to what we saw in the Phase 2A. But the biomarker negative folks do a lot of good for you in running the study even alone. So we're trying to keep everybody blinded about their biomarker status. Nobody knows their biomarker status, including our staff. We run our own trials. Our staff working with sites, the patient and the clinician likewise don't know. They don't, the sites don't know what proportion of the study is going to be biomarker positive or negative.
We don't want expectations built by patients as being as part of being in an Alto study. And then we'll ultimately, in a not sufficiently powered population, be able to tell qualitatively what the effect size is in biomarker negative patients, which again is in line with the enrichment guidelines. And there's not really a bar for how much enrichment you need to achieve. You need to achieve some. It doesn't have to be statistical from everything that's out there. But it'll give us a flavor for that. And ultimately, it'll be Phase 3 where we do more thorough risk-benefit assessments for those off-label population. If they don't drive much benefit and they don't have many adverse events, then, you know, if some of them get it, it's not a big deal. But our focus is really on advancing the on-label biomarker positive population.
That will ultimately be what payers, I'd imagine, focus on as well.
Yeah. Okay. Makes sense. I actually got one question on 100 from an investor who submitted it, and that is trying to clarify a couple of things related to the analyses you've done related to the Neuralstem data. So the question here is, you've shown MADRS data for two dose cohorts of poor cognition patients compared to placebo. The question is, in that data cut, does that placebo comparator only contain poor cognition patients? You want to clarify that quickly? There's a second question too.
So that's apples to apples.
Apples to apples.
Poor cognition, same cutoff for everybody. Then just look at the drug and placebo. Again, it's, it's the prospective data that we put weight on, right? I, like the others, might be totally comfortable going with a retrospective analysis to run the next trial.
Yeah, yeah.
We've set a higher bar for ourselves.
Yep. And I think you've talked about data from, I think it's from the Trintellix program looking at, you know, poor cognition, non-poor cognition. So as a validating fact that, like, the placebo shouldn't have any impact on placebo. But is there any insights you can glean from the Neuralstem data on that? That was the other question here. What was the placebo response of poor cognition patients versus biomarker negative patients within the Neuralstem data set? Any insight there?
Yeah. So the Trintellix data that you're referencing are two Phase 3 programs where they used exactly the same memory measure so we can calculate the same biomarker. And we saw across those two different studies, similar or slightly worse response. We have six additional data sets. We have some element of cognition that correlates with memory that we have looked at. And again, across those, you now see a slightly lower response to placebo in those poor cognition patients. And that's exactly what you actually see in the Neuralstem data as well, is a slightly lower placebo response. But just to be, like, super crystal clear, what we have, set as our kind of powering and our expectation of effect sizes is purely from the enrichment of drug alone, assuming zero effect on placebo. Anything better than that, you know, would be great.
But we've tried to be as conservative as possible in our powering.
Yeah. Okay. Sounds good. You want to switch gears to 300?
Let's do it.
So as it relates to 300 or agomelatine, do you want to just briefly orient us with the history and, you know, how that drug ended up getting approved in Europe but not the U.S.?
Yeah. It's a drug with a unique mechanism of action. It stimulates melatonin receptors and blocks 5-HT2C receptors, which leads to an increase in, in dopamine. Unique relative to everything else that's out there, either in development or approved. It has similar all-comer efficacy as all other antidepressants, but better tolerability than all other antidepressants that, that are approved. What ended up happening is Servier had basically mixed results in their Phase 3, much like pretty much every program using all-comers. I mean, recall that, you know, Prozac needed something like 7 trials to get to two positive, right? This is very, very common. That was enough to eventually get them approval in Europe and Australia, but not, you know, they didn't develop it for the United States. They licensed that to Novartis, who got the same mixed results.
In the Novartis US studies, they split the doses in two Phase 3s, and one dose was positive in one and one dose was positive in another. It's literally the poster child for why you need precision, because we know there is efficacy there across multiple studies. Then that's where we step in, is a drug with known efficacy, excellent tolerability. Then we found and prospectively replicated an EEG biomarker that we are now selecting on in that Phase 2B, which is on track to readout first half of next year.
Got it. Okay. Great. From a safety perspective, you know, what gets you comfortable with the, with the margin here related to liver?
So the liver issue is actually a bit of a, you know, a historical almost do-it-yourself issue, according to some, which is that, frankly, if you measure LFTs for most psychiatric drugs like antipsychotics, you'll get some bump in LFTs. And that's what they found, but only at the 50 mg dose. It's always been reversible. It's been on the market since 2009, never led to liver failure. So on that side, it's not really a big deal, period, to do some monitoring. But what we have seen consistently in prior studies is that at the 25 mg dose, there's not an increase in LFTs, for example, in Novartis studies, the same level as placebo. And the all-comer efficacy was similar. Again, Novartis hit on one of those Phase 3s at 25 mg.
So we thought, why not just take that issue off the table and only develop 25 milligrams? Again, we replicated the enhancement of clinical response. But then we also found out of 239 patients that we've dosed that none had an elevation of LFTs greater than a three times upper limit of normal, which matched what we saw in the even larger Novartis studies. So that all says that we're probably, you know, in the right ballpark for efficacy, tolerability, and safety.
Got it. Okay. Great. What can you tell us about the EEG biomarker and why it makes sense?
Yeah. So this is, by contrast to 100, more of an empirical observation that's driven by our machine learning approach, which we've done and validated now in multiple, you know, studies and populations and interventions over the years. So it pulled out a signal related to neural variability. The more variable that neural signal is with EEG, which you can measure even in patients' homes now, the better the response to agomelatine, to ALTO-300, no difference in response, by the way, to placebo, and no difference in response to standard of care treatment. We've looked at all of those. Now, that's an empirical observation, so it helps us advance the drug. And what FDA has told us in a prior Type C meeting is, while biological rationale is helpful, it's not required. So things continue to move forward.
That said, we'd like to understand more about why it is, and we're doing the work now. One hint for that is that if you dopamine deplete a human being, you get a signal that looks more like that high variability signal. And recall that ALTO-300 will increase dopamine, so the opposite of a dopamine depletion scenario. So that could be one reason, why. But there's a lot more work to be done that's ongoing as we speak, and we'll report on more of that as it starts to yield fruit.
Yeah. Okay. Great. From a commercial perspective, you know, I think you've said the verbal memory test, right, is something you could easily do from the home on your computer, 20 minutes in the real-world version. Is EEG trickier? Like, how do you think that would play out?
Yeah. So we've done now so many thousands of EEGs that we have really figured out how to do them consistently. We've developed our own software for doing real-time QC of EEG data. QC being done in real time is happening at sites. It's also happening in patients' homes. A very low-cost, $200 to manufacture EEG system can get this signal in all likelihood and just be mailed to the patient, or the clinician does it. But in either case, the expertise for doing it is taken out by the software, and the expertise for interpreting is taken out by the software. And so that really makes it easier to embed a new tool in an area where psychiatrists have not historically had experience with EEG. If the patient can do it, most certainly a psychiatrist can do it.
Okay. Okay. Great. I know there's a whole pipeline here we could talk about to wrap things up. Amit, I'm going to put you on the spot, and it's okay if you don't answer this question, but I'm going to try. Which of 100 or 300 do you think has a higher POS?
You know, it's something that I'm going to let the folks, you know, in the peanut gallery come up with their own ideas. Both of them, I think, have the same logic from a data science perspective behind them. One of them will just come first. So, you know, that will teach us about the other, to a certain degree. But we're enthusiastic about both. That prospective replication, specificity versus placebo and standard of care, is very similar in both. So, you know, I have three children. I can't choose between the three of my kids, and choosing between our assets is likewise a little hard to do.
Okay. Fair enough. From a cash runway perspective, how far does that take you?
Well, into 2027. As you mentioned, there's a pipeline here. We have two proof of concept trials that will kick off, the first half of this year to readout next year. And that's already well covered, you know, cash well beyond the final readouts, which will happen in 2025. So excited for where that puts us as a company.
Okay. Excellent. All right. Thank you for taking the time. It was great seeing you.
It was a pleasure.