Hello, and thank you for joining us to discuss Belite Bio's Second Quarter 2024 Financial Results. Joining the call today are Dr. Tom Lin, Chairman and CEO of Belite Bio; Dr. Nathan Mata, Chief Scientific Officer; and Hao-Yuan Chuang, Chief Financial Officer. Before we begin, let me point out that we will be making forward-looking statements that are based on our current expectations and beliefs. These statements are subject to certain risks and uncertainties, and actual results may differ materially. We encourage you to consult the risk factors discussed in our SEC filings for additional detail. Please note that you can submit questions throughout the call by clicking on the Q&A box at the bottom of your screen, and we will respond to questions following our prepared remarks. Now I'll turn the call over to Dr. Lin.

Thanks, Julie. Thank you, everyone, for joining our second quarter 2024 earnings call. We had another strong quarter advancing our Tinlarebant trials, and I'm pleased with our progress year-to-date. Let me start off with our accomplishments during the second quarter by providing an overview of our programs. We had a productive quarter marked by several advancements in our programs. We have initiated the Phase Ib and Phase 2/3 trial of Tinlarebant in adolescent Stargardt patients, also known as the DRAGON II study. We have already completed enrollment for the Phase Ib portion with six subjects in Japan. We've also received Sakigake designation in Japan, which translates to Pioneer Drug Designation in English. This designation has only been granted to 27 drugs since its initiation in 2015, and Tinlarebant is the first ophthalmic drug to receive Sakigake designation, which is a testament to the groundbreaking potential of this drug and the unmet need it aims to address for people living with Stargardt disease. We've made progress in our pivotal global Phase 3 trial of Tinlarebant in geographic atrophy subjects, now known as the Phoenix study, and have already enrolled nearly 200 subjects until now. In addition, we've also raised $25 million from a registered direct offering in April to strengthen our balance sheet. The Phase III DRAGON I study is fully enrolled, with estimated interim readouts by Q4 2024 or early 2025 at the latest. As we enter the second half of the year, we are well-positioned to execute on key milestones, and we look forward to sharing initial interim analysis from our pivotal Phase III DRAGON study in the fourth quarter. Next slide, please. For those of you who are new to the story, let me explain it briefly. Tinlarebant is a novel once-a-day oral tablet designed to bind to serum retinol binding protein 4 to specifically reduce retinol delivery to the eye. This approach is intended to slow or stop the formation of toxic retinol-derived by-products generated in the visual cycle, which are implicated in the progression of Stargardt disease and geographic atrophy. We believe that early intervention directed at emerging retinal pathology, which is not mediated by inflammation, is the best approach to potentially slow the disease progression in both Stargardt disease and geographic atrophy. Importantly, there is still a significant unmet need for both indications, as currently there is no approved treatment for Stargardt disease and no approved oral treatments for geographic atrophy, and we are already in global Phase III trials for both indications. To give you an indication of the importance and potential for our oral therapy, we have been granted fast track designation, rare pediatric disease designation, and orphan drug designation in the U.S., EU, and Japan, as well as Pioneer Drug designation in Japan, as I just mentioned. Finally, we have strong patent protection with about 14 patent families, most of which are composition of matter patents. The latest composition of matter patents is expected to last until 2040, with additional patent term extension potential with new patents to be filed, ensuring we have patent protection beyond 2040. All of this highlights the potential of Tinlarebant to treat people living with these debilitating conditions and capitalize on large market opportunities. We are very excited about the path ahead, and our team is focused on our mission to leverage Tinlarebant to address the unmet needs of patients suffering from blindness. I would like to pass this down to Nathan, our CSO, to give you a clinical and scientific update. Nathan?

Yes. Thank you very much, Tom. So what I'd like to share with you is the data from our 2-year open-label Phase II study in adolescent Stargardt subjects. This study was, as I mentioned, a 2-year study with 13 enrolled subjects from Taiwan and Australia. What a lot of people don't understand about Stargardt disease is that there are over 1,500 known mutations that are associated with the disease. Not all of them are known to be pathogenic. In fact, many are mild or benign. One of the analyses we did initially was to determine the genetic composition in our cohort. We provided the genetic data to one of the premier genetics experts in Stargardt disease globally, Dr. Rando Allikmets at Columbia University. He evaluated our genetic data and determined that 11 of the 13 subjects in our cohort had severe biallelic mutations, which we predict to be pathogenic. The two where there was a moderate allele in these subjects, in vitro testing showed that these were indeed pathogenic alleles. Thus, our entire cohort has severe pathogenic mutations that are predicted to progress very rapidly through the disease course. An independent assessment of the genetic severity is provided by the CAD score, which stands for combined annotation dependent depletion score. It indicates the degree of severity of a particular genetic variant or mutation. Scores above 20 are predicted to be among the 1% most deleterious. Every single one of our subjects, with the exception of 3 and 5, had these CAD scores above 20. So we have two independent confirmations on the severity of the genotypes of these subjects. Despite the severity of these genotypes, we had five subjects, which represent 42% of the cohort, who never developed atrophic lesions. I should have mentioned that in this study, these adolescent subjects began with an early form of disease, where they only had a type of lesion known as a questionably decreased autofluorescent lesion. Over time, these autofluorescent lesions convert to atrophic lesions, and that's one of the parameters we're observing. We see that, in 42% of subjects, this conversion never occurred. Another interesting outcome from the genetic data was that we found two pairs of siblings with identical mutations. This is significant because there are competing companies that are using the premise that identical mutations predict an identical disease course for their therapeutic approach. This gives us an opportunity to evaluate that premise and determine whether there is any validity to it. Because this is an open-label study, one of the metrics we want to track to see if we have an effect on patient well-being is visual acuity. We looked at visual acuity in subjects prior to enrollment, specifically looking for subjects who were losing letters in both eyes, this is called bilateral BCVA loss. We found a subgroup of six subjects within our larger cohort who were losing an average of 10 letters per year prior to enrollment. The natural history in our study suggests that there would be clinically significant vision loss in these subjects over the course of the study. We need to monitor how these subjects fare. Another important finding from the pre-enrollment data is that since patients were already losing vision significantly without atrophic lesions, it suggests that non-atrophic lesions such as these QDF lesions can compromise visual acuity. This is significant because all of these subjects have foveal-involved lesions, meaning their vision will be compromised over time. However, current scientific thinking suggests that you must have atrophic lesions before there is any effect on visual function, and that appears not to be the case. Finally, regarding sibling comparisons, we found that sibling subjects with identical mutations do, in fact, exhibit different levels of BCVA loss. This data can be found in the appendix of this presentation. If we look at the overview of visual acuity in all subjects, as shown on the left-hand side, over the 2-year study, we see a mean loss of about 2.5 letters per year. That is significant because it suggests stabilization. This vision is not really changing in all subjects. However, if we look at those subjects with prior vision loss, specifically those subjects who were losing 10 letters per year before entering the study, that is shown on the right-hand side, now they're only losing about 1.9 letters per year. So we've significantly altered the visual acuity progression in these subjects, stabilizing it. This is very significant. The only reason we could achieve this is if we are having some effect on lesion growth. As mentioned, 5 out of 12 subjects never developed an atrophic lesion, and I want to illustrate what that looks like. The images you see here on the upper right-hand side are representations of what all these subjects essentially look like at baseline. This is subject 10. But all these subjects have types of autofluorescent lesions encroaching the fovea, just of different sizes. We are measuring over time how these autofluorescent lesions convert to atrophic lesions. As stated before, 7 out of 12 actually grew these atrophic lesions. Interestingly, we found that in every case, except one, the increase of the atrophic area was matched by a decrease in the autofluorescent area in every subject. Where you see an orange bar is an increase of atrophy, and where you see a blue bar is a decrease in autofluorescent lesion size. This is significant because the boundary of the lesion is not growing; only the atrophic lesion is expanding within the autofluorescent area. This suggests that the autofluorescent lesion could potentially resolve over time because there is no room for the lesion to grow. These data indicate that there are cells that may be destined to die, which we cannot save with our treatment, but we're certainly preserving the cells around them that would lead to further lesion growth. These findings are quite important. There was only one subject with a lesion that was outside of the initially identified QDAF lesion area. Regarding the genetic mutations, subjects 9 and 10, and subjects 12 and 13 had identical mutations. Yet if you look at 9 and 10, they have similar lesion growth, while 12 and 13 show no lesion growth, despite having identical genotypes and very similar disease duration. These data suggest on a lesion growth metric that identical mutations do not predict an identical disease course. We have one other very important piece of information to share, and that is about how these lesions are graded. We are currently using the routine methodology that everyone else is using. It's basically an autofluorescent camera that takes a picture of the lesion, and then a reader must draw the boundary around the perimeter of that lesion for the computer to calculate the area. Two readers must grade every single image, and if they don't agree, a third must serve as a tiebreaker. This method is subjective and is subject to intra-reader bias. It’s also time-consuming. To address the shortcomings of currently used methods, our reading center has developed a new AI-based method for assessing lesion sizes. This is a mathematical classification that uses the gray level density in healthy tissue relative to the area of diseased tissue. In this case, we're focusing on atrophic lesions. The importance of this new method is that it eliminates the reader's potential subjective bias. When our reading center used this methodology to reassess our images at baseline, they identified 12 eyes of 8 subjects with atrophic lesions in the macula, which previous methodology had not detected. We asked our reading center to reevaluate all these images and assess macular lesion growth in subjects identified with atrophic lesions. This is the data produced: on the left side, we see the growth of lesions into the macular area over time. It appears pretty linear until about month 16, when it completely stops, showing no further lesion encroachment into the macula during the following 18 months. On the right-hand side, we see the same data in terms of percentage change of lesion into the macula over time, where 100% would mean total involvement of the six-millimeter zone of the macula, and our subjects never reach more than single-digit involvement into the macula. This is significant and validates our visual acuity data, as it explains why we observe stabilization of vision; we are halting lesion growth into the macula. Lastly, the safety data for this two-year observation shows no drug-related systemic adverse events whatsoever in these subjects. This speaks to the specificity of this drug, designed by scientists at Columbia University, as it targets only the residues in the binding pocket of retinal binding protein 4, which exist nowhere else biologically. This specificity is shown in the adverse event data, which indicates that it is quite clean. What we do see in terms of adverse events are anticipated ocular events, which we want to see as they indicate our drug is having the intended biological effect in the retina. The other important aspect of these adverse events is that they are completely manageable by adjusting to differences in lighting, as these adverse events are driven by light. The first adverse event to mention is a form of xanthopsia or chromatopsia, specifically xanthopsia. This results when patients transition quickly from dark to bright light, triggering cone photoreceptors in the eye that require vitamin A immediately. Under our treatment regimen, this supply is slow, which leads to a period in which the cone photoreceptors may misfire and produce transient yellow hues in the visual field. Patients report this as mild, and importantly, none left the study due to this adverse event. A second adverse event, delayed dark adaptation, occurs when transitioning from bright to dark environments, leading to a delay in adapting to dim light. This phenomenon is already familiar to patients with Stargardt disease, who experience delays in dark adaptation. That explains why most report this pharmacological delay as mild or transient, and again, no one left the study because of this. The bottom line is that these adverse events can be mitigated by moderating transitions between bright and dark environments, which has been very helpful for our subjects. I can confirm that in over one year of dosing in our Phase 3 study, the dropout rate due to these adverse events is less than 4%. This is significant. Night vision impairment can be a more severe manifestation of delayed dark adaptation, characterized by delays of 20 minutes or more. Increasing error scores on the FM 100 show more severe exacerbations of chromatopsia, which have been observed in one subject. The intermittent headaches may occur when subjects strain to utilize their visual acuity during these adverse events. Moving on, I will overview the trials that Tom discussed: the DRAGON I and DRAGON II Stargardt trials. These studies are essentially identical, with the key difference being the geography. As Tom mentioned, we have Japan included due to the Sakigake designation. The demographics are similar, with the main differences lying in double-blind design. There is also a 2:1 randomization in the DRAGON study and a 1:1 randomization in the DRAGON II, primarily due to the lower number of patients in the DRAGON II study. Despite these differences, all other assessments, including safety and efficacy, are the same, as well as key inclusion criteria for subjects. Due to the similarities in these two studies and their alignment with the Phase 2 study, and considering the Phase 2 study is trending positively, we are optimistic that we will see very promising safety and efficacy data from both DRAGON I and DRAGON II studies. Regarding geographic atrophy, the trial design mirrors that of the Stargardt disease Phase 3 trials with the main distinction being the indication—geographic atrophy—and a higher number of subjects to reflect the disease's higher prevalence. Otherwise, these studies are essentially identical, and we believe that findings in Stargardt may predict similar results in geographic atrophy. This is mainly because we are using the same dose and there is high pathological similarity between Stargardt's disease and geographic atrophy in the enrolled patients. I will now turn it over to Hao-Yuan for the financial results.

Thank you, Nathan. In Q2 2024, we had R&D expenses of $9.1 million compared to $5.5 million for the same period in 2023. The increase was primarily due to higher expenses related to a milestone payment to Columbia for completing the Phase II study and share-based compensation. On G&A expenses, in Q2 2024, G&A expenses were $1.4 million, basically the same as Q2 2023. Our net loss was $9.5 million in Q2 2024 compared to $6.8 million for the same period in 2023. Regarding cash, we have cash deposits in U.S. treasury bills totaling $112 million. We still expect about three years of cash runway ahead. Thank you. Back to you, Tom.

Thanks, Hao-Yuan. To summarize, we had a strong start to the first half of the year and continue to make meaningful strides in advancing Tinlarebant clinical trials for Stargardt disease and geographic atrophy across several countries. We're also proud to have received Sakigake designation in Japan, which we believe underscores the groundbreaking potential of this drug and the unmet need since no treatment currently exists for Stargardt disease. We are in a strong financial position with $110 million in cash and cash equivalents. As we enter the second half of the year, we are well-positioned to execute on key milestones and look forward to the interim analysis from our DRAGON study in the fourth quarter. We also look forward to seeing some of you next week at the H.C. Wainwright 4th Annual Ophthalmology Conference and hope you will join our presentation on August 15. Additionally, in September, we will attend conferences with H.C. Wainwright, Cantor, and Deutsche Bank, and we hope to see some of you there. Thank you again for joining this call, and now we will open the floor for questions.

Our first question comes from Marc Goodman with Leerink Partners. Marc, your line is now open.

Hi, good afternoon. This is Basma on for Marc. Our question is on DRAGON II. Could you provide insight about this trial and what its specific goals are? Does statistical significance need to be achieved with this trial, yes or no? And do you think the powering is sufficient? Also, do you need to run another long-term safety trial in these global sites for submission OUS, or will DRAGON II alone provide enough safety data, and can the safety database in the U.S. suffice? Thank you. That's it.

Nathan, do you want to take this?

Yes, I'm happy to take that, Basma. Thank you for the question. Regarding statistical significance, yes, of course, we will need to power for statistical significance at the 2-year time point; that will have to be achieved in order to fulfill that requirement for demonstrating efficacy. Conditional power will not be an issue because, in this particular study, we are randomizing 1:1 versus 2:1 in our original DRAGON study. So, in the original DRAGON study, there are roughly 35 placebo subjects and the remainder of the 104 are the Tinlarebant-treated subjects, whereas in the DRAGON II study, there is equal distribution in the randomization, which helps with conditional power. I don't believe we will need to conduct an open-label extension study to evaluate safety based on what the PMDA has told us in Japan; they only require the 2-year safety data from the required Japanese subjects enrolled in DRAGON II, which is a minimum of nine Japanese subjects, and we are complying with that. So I hope that addresses all your questions, Basma?

Yes, thank you. That was very helpful.

Our next question comes from Jennifer Kim with Cantor. Jennifer, your line is now open.

Jennifer, I think you are on mute.

Oh, hi. Can you hear me now?

Yes.

Okay, great. Thanks for taking my questions. Maybe to start off with DRAGON I. With the interim analysis later this year, can you remind us what you are thinking in terms of the format of that update and what level of detail we should expect? And then my second question is following up on the last question on DRAGON II. You've completed the Phase Ib portion. Any thoughts on when we could see initial data from this trial? Is it going to be the 2-year data, or can we see an interim update? Thanks.

Thanks, Jen. The FDA guidance states that we are not allowed to review any data; in fact, only the DSMB has that knowledge, given the ongoing treatment. We don't want to bias the data in any form. Nathan, do you want to take that?

Regarding DRAGON II? Yes. Sorry, Jennifer, what was your specific inquiry on DRAGON II?

The interim data as well...

Yes. The Phase Ib—these studies are interlinked since it’s a PK/PD study, originally designed to ensure that the 5 mg dose we're using in all our other trials achieves the same pharmacokinetic and pharmacodynamic response in Japanese subjects, which we have confidence it will. We won’t disclose this data until we finish the efficacy portion of the Phase II/III study, which, of course, is two years later. We will allow all that data to be disseminated once we have all the topline 2-year data, which will include PK/PD findings and dose efficacy or dose finding.

Thanks. That’s helpful. If I could squeeze in one more question, just on PHOENIX: could you confirm the timing of enrollment completion for PHOENIX? And what kind of 1-year interim look should we expect for that trial? Thank you.

Yes. We are targeting an enrollment size of 429 subjects. We're about at the 200 mark right now, which places us roughly halfway. Based on our current run rate, we expect to close enrollment by the end of Q1 of 2025. Regarding data from that study, you asked about the interim data; we have not yet decided what we will be doing in terms of interim analysis. In fact, we haven't completely fleshed out the SAP, the statistical analysis plan for that study. So let's revisit that one later when we have more concrete information for you.

Our third question comes from H.C. Wainwright, Yi Chen. Your line is now open.

Thank you for taking my questions. My first question is to confirm that data from both DRAGON I and DRAGON II trials are required for submission to PMDA under Sakigake designation, correct?

Yes, that's correct.

And the same data set will be submitted to the U.S. and European regulatory agencies as well. Under the Sakigake designation, do you think there’s a chance that Japan could approve the drug faster, or possibly even before the FDA and EMA?

Good question. It seems Japan is eager to be the first country to approve this drug. However, we still need to discuss with the PMDA regarding their timeline for when they require the DRAGON II data.

To clarify, the Japanese regulatory agency does not require us to complete the entire DRAGON II for Japan submission. Their requirement is that we complete DRAGON I and the data from nine Japanese patients enrolled in DRAGON II. We can submit without the additional U.S. and U.K. patient data because they are mainly interested in the performance of the Japanese subjects in this study.

Thank you very much for your clarification. My follow-up question concerns the ABCA4 mutation; is that something being measured in the DRAGON I and DRAGON II trials? How should we interpret the Phase II observation regarding the mutation of those five patients in relation to the potential outcomes for the DRAGON I and DRAGON II readout?

Nathan?

Of course, we do genetic analysis and genotyping on all subjects; participants are required to have both clinical and molecular confirmation of Stargardt disease. So that will be performed in all our Stargardt trials. As for the results observed from the five subjects in the Phase II trial, it is difficult to extrapolate how that will influence the Phase III data because there are over 1,500 known mutations in Stargardt disease, making it challenging to match them accurately. In those five subjects who never developed an atrophic lesion, they had severe genotypes. However, other severe traits might behave similarly but would differ due to varying genetic loci. It is challenging to take those specific five subjects' genetic insights and apply them to a more extensive patient population with different mutations. Our cohort of 13 subjects all had severe pathogenic mutations, suggesting that pathology determines that this will be a rapidly progressing disease. We are reinforcing that validation during the screening as we will gain some confidence about the presence of pathological gene mutations. Thus, we will not primarily match what we see in those five subjects with broader outcomes in DRAGON studies. However, it is significant to note that those pathological mutations which lead to the disease have been reduced with our treatment, and we hope to see that extrapolated across other severe genotypes.

Thank you.

Our next question comes from Bruce Jackson with Benchmark. Bruce, your line is now open.

Hi. Can you hear me okay?

Yes.

Okay. Super. With the presentation of the interim analysis for DRAGON, is that going to coincide with the medical meeting?

I don't believe it will coincide with the medical meeting; are you referring to our presentation at the AAO?

Possibly at AAO, yes.

No, it will not coincide, as the interim analysis is expected around the end of the year. However, the timing during Christmas is tricky, so we must coordinate with the DSMB regarding when they can review the data. It may extend into January 2025, but we expect to have results by the end of the year.

Okay, great. And then regarding the R&D expense, you had the milestone payment this quarter. Can you clarify what the baseline rate of R&D is for the next couple of quarters and how the milestone timing for the remainder of the year and 2025 might play out?

This quarter, the higher expense is attributed to that milestone payment. For the rest of this year, I would expect our R&D expense to remain stable, probably between $7 million to $8 million per quarter. So for the entire 2024, I anticipate it will be approximately $30 to $35 million. Next year, in 2025, it may be slightly higher due to expected milestones achieved from the PHOENIX study, estimating around $35 to $40 million. Essentially, the expenses for these three years will be similar, around $35 million annually, considering we're running the PHOENIX study during this time.

Okay, great, and congratulations on all the progress. Thank you.

Thank you very much, Bruce.

Please let me know if there are any additional questions.

No, I don't have any written questions here.

Great. This concludes our Q&A portion of the call. I will now turn it back to Tom Lin for closing remarks.

Thanks. Thank you, everyone, for attending our earnings call for this quarter. We will certainly provide updates once we have data from the conferences, including those at the AAO. We will keep you informed about upcoming events as they arise. Thank you again, everyone, and we will keep you updated.