Earnings Call: Q1 2021

May 25, 2021

Good day and thank you for standing by. Welcome to Burning Rock's 2021 First Quarter Earnings Conference Call. At this time, all participants are in a listen only mode. After speakers' presentation, there will be a question and session. Please be advised that today's conference is being recorded. Before we begin, I would like to remind you that this conference call contains forward looking statements within the meaning of Section 21E of the Securities Exchange Act of 1934 as amended and as defined in the U. S. Private Securities Litigation Reform Act of 1995. These forward looking statements can be identified by terminology such as will, expects, anticipates, future, intends, plans, believes, estimates, target, confident and similar statements. Statements that are not historical facts, including statements about Burning Rock's beliefs and expectations are forward looking statements. Such statements are based upon management's current expectations and current market and operating conditions and relate to events that involve known or unknown risks, uncertainties and other factors, all of which are difficult to predict and many of which are beyond Burning Rock's control. Forward looking statements involve risks, uncertainties and other factors that could cause results to differ materially from those contained in any such statements. Burning Rock does not undertake any obligation to update any forward looking statement as a result of new information, future events or otherwise except as required under applicable law. And now I'd like to hand the conference over to the management team of Burning Rock. Thank you. Please go ahead. Thank you. Welcome to Burning Rock's earnings call. I'm Yishun Han, the CEO and Founder of Burning Rock. And today, we also have our CEO, Shannon Cai our CTO, Zhou Zhang and our CFO, Leo Li in this call. So, Burning Rock is a Chinese molecular diagnositor for precision oncology. There are 2 parts of our business. The first one is early detection using liquid biopsy for cancer and the second is for severe selection and MRG. So, Bruce will turn to Page 4. So today, we're going to recap the recent progress for both early detection and therapy selection. So for early detection, we are very excited to launch the multiomics 22 cancer test, which is called present. And the second thing is that the PREDICT trial for mind cancer test is going well mostly. And our CEO, O Sheng Zhai, will talk about these two trials in detail. And in the meantime, the preparation of commercialization of the Assist Cancer task is ongoing. And we are continuously building our commercial and operation operating team and optimizing the SOPs. Nothing significantly important to report for the sick cancer so far. But if you have any question, welcome to ask at the Q and A session. And for the therapy selection, we have great news that finally the results of our weekly biopsy part of FE2C2 has published in Nature Biotechnology, which proved that starting wealth quality is a top tier level in the world. And our CTO, Joe, will talk about that in detail. And after that, our CFO, Liu, will talk about the financial numbers. So let's turn to Shenmeng first about the early detection part. Shenmeng? All right. Thanks, Yixin. So if we go to Page 6, this is to recap the product development roadmap for our early detection programs. So we started with the proof of concept on lung cancer and the study and the methodology has been actually most recently accepted for publication and the manuscript is pending publication right now. And then we moved on to a 3 cancer test, which we presented the data last year in January in the AACR special conference on nuclear biopsy. We were able to achieve a 95% specificity and 81% sensitivity. And then most recently, I think a lot of you are familiar with our 6 cancer test results that we released last year in November on the Eastern Asia. The 6 cancer tests include all covers lung cancer, colorectal cancer, liver, ovarian, pancreatic and esophageal. And the data showed that we were able to maintain our 81% sensitivity while improving the specificity improving the specificity to 80 about 98%. And we also were able to achieve a reasonably good accuracy for TLO analysis, T Shock Origination analysis from this FIS about 81% accuracy from the fixed cancer test in the results we released last year. And then for the fixed cancer test, after the Thunder study, which was a case control study to validate this specificity and sensitivity of the product. We also are planning to move on to a prospective interventional study for asymptomatic population. So that study is currently under planning and I'm going to show you an overview in later pages. And today, we wanted to focus on the most recent very exciting progresses that we are making on the 9 cancer test and looking forward to the 22 cancer test. As Yixuan has mentioned, for the 9 cancer test, our current status is that the PREDICT study that we started earlier is going on very smoothly. The progress is as expected. And also for the 22 cancer test, first of all, the 22 cancer will eventually cover about 88% of China's cancer incidents altogether. And then for this 22 cancer test, we have just kicked off a large clinical development study a few weeks ago in May 2021, and it's called Prussian. So in the later pages, I'm going to tell you about a little bit about the design and timeline for both the PREDICT study and the PREDICTION study. So if we go to Page 7, this is sort of an overview of the clinical programs that we are looking at for our early detection program. So for each product or each version or each generation of our product, there are 4 roughly 4 phases for the product development or development. The first is the SAA development in which we do the panel design, the marker selection and also the assay chemistry finalization. And then we do the analytical validation to validate the performance analytically, including using reference materials or clinical samples to validate the performance specifications. And then we move on to the case control study. And for this one, you can sort of think of the CCDA studies from GREL. It's similar to what we're laying out here as the case control study. In these studies, they are real clinical cases and controls or healthy controls. And in these studies, we will be able to validate the sensitivity, specificity and PO accuracy, the key statistical performance features for the early detection products. And then after that, eventually, we'd want to move on to the asymptomatic population in which we will be able to validate again the sensitivity to our accuracy in these eventually intended to use intend to use population. And for this one, you can sort of compare that to the pathfinder trial from GRILL as well in which it's testing VALORI on the asymptomatic population in a prospectively reputed cohort. So for our pre cancer tests, we after we finished the assay development and analytical validation, we didn't move on to the clinical development. So we moved forward to the 6 cancer test. And for the 6 cancer test, we have so far completed the assay development and analytics validation as well as the case control study, which was what we mentioned in the last page as the Thunder study. The Thunder study we're able to show in a training and a pre specified validation cohort the 81% of sensitivity and 98% of specificity. And then again, the prospective interventional study on the asymptomatic population for the sick cancer test is currently under planning. And we hope that we will be able to disclose more details about that study once it's finalized and released or kicked off in the near future. Again, we in the meanwhile, in parallel, we are also moving on to more cancer to cover more cancer tests, of course. So for the night cancer test, we have already finished the assay development, which means that we have finalized the chemistry and also the market selection, the panel design, etcetera. And the analytical validation for that product is currently ongoing. And in the meanwhile, we were able to start the enrollment on the case control, the PREDICT study in parallel to sort of shorten the development time overall. And the PREDICT study, if you might recall, it actually contains 2 phases. And in Phase II, it does contain a factor of testing or validating among a small asymptomatic cohort or healthy controls. So that's why over here, we are sort of spanning that a little bit beyond the post control study even though it's not fully powered to test on the healthy or asymptomatic population. In the later page, I will be able to give you more details of how the two phases of this study will work out. And then for the 22 cancer test, we are actually working on developing this next generation of the product in the meanwhile and it's currently under the assay development stage. However, because we collected a lot of information, our preliminary results from the previous versions of the product, we were able to finalize the design of that patient study as the case control study that we are planning for the 22 cancer tests. So that's why we have kicked off the enrollment of the study. So we will be able to recruit samples, clinical samples for future testing for validation for the TEN2 cancer test in parallel with the Florida development efforts. So if we move on to Page 8, this is this outlines the study design of the PREDICT study. Again, it covers 9 cancer types, which are listed out on the upper right here. And overall, this study contains more than 14,000 participants, about 55% of them coming from cancer, about 10%, I think 10% coming from benign diseases and the rest from healthy controls. And one thing we wanted to point out is that more than 75% at least 70 5% of the cancer participants will be from Stage 1 to 3. So most of the cases we will focus primarily on the early stage patients. We wanted to know or be able to assess our sensitivity among the early stage patients because that's what really matters. And then in terms of the study design, as I mentioned, there will be 2 phases. The Phase I will be an open label design, which means that for Phase I, we will divide the Phase 1 samples into the pre specified training and testing steps. And within the training, we will be able to customize or tune our models and cut offs and then to be able to report the results or performances on the validation set within the Phase I cohort. And then after that, the model, those CSA and the model will be locked and then we move on to Phase 2 sample processing. So in Phase 2, we will have a total independent set of data to test or to validate the performance of the locked model or master from Phase 1 and to be able to have a rather accurate assessment of the model performance for the night cancer test. And then one thing we also wanted to point out is that for PRODEX study participants, we are planning for a 12 month follow-up, especially on the healthy controls, which will have a positive testing results so that we will be able to have a positive predictive value assessment among these healthy control cohorts over the follow-up. So on the next page, we listed our objectives and timeline for PREDICT. So of course, the primary objective will be to test the to train and validate the sensitivity, the specificity NTL analysis of our CFD and immunoassulation based model for these 9 cancers, 9 types of cancers. And then for the secondary objectives, we of course wanted to learn about the performance among different types of cancers and also among different stages of cancers. And also we wanted to know whether other biomarkers, including protein biomarkers that we are processing in parallel in these predict samples, whether they will help in any way, which we do expect they might help in at least some of the cancer types. And how do they help and how do we combine them with the methylation markers? That's something we will explore from the PREDICT data. And then last but not least, we will have a chance to evaluate the PPV in this cohort as well after the 12 month follow-up period. And then in terms of the timeline, we expect the Phase I enrollment to finish, to complete by 2022. And then we will be able to have a readout for Phase 1 data by the end of 2022. And then by the end of 2023, we will have the readout for Phase II data. And by 2024, we will be able to finish the follow-up and have the complete data set for the PREDICT study. On page 10, we wanted to briefly mention the kind of attention that the PREDICT study has attracted among the Chinese oncology community. And this is the picture of the principal investigator, Doctor. Jeff Fan, when he presented at the keynote speech on the National Oncology Conference on standardized diagnosis and treatment conference in that was held a couple of weeks ago in Beijing. And the PREDICT study design and also the news was announced during that conference by Doctor. Dazan and it has attracted a lot of interest and attention from the community so far. And then moving on to the next page, Page 11 lays out the study design for the Prussian study. So compared to the PREDICT study, Prussian study actually has 2 dimensions of expansion. The first is, obviously, extending from the 9 cancer types to the 22 cancer types, which are listed on the upper right again. And as it has similar design, but not divided into 2 phases. However, it has another dimension of extensiveness that beyond methylation and protein markers, we will profile other omics of biomarkers in the patient samples as well. We won't be able to disclose too much detail on what exactly are we testing there, but the extension will be the exploration from the collection study will be focused not just on the cancer type expansion, but also on the omics combination. And then on Page 12, again, the objective for depression study is to we will be able to validate the methylation plus protein markers among the 22 cancer types. And then we will be able to assess different the performance among different stages and different types of the cancer. And then very importantly, we will be able to evaluate the potential combination, including methylation protein and other genetic, epigenetic biomarkers from the patient study. So in terms of the timeline, we expect to complete enrollment for patient by 2023. And then for the patient study, we will divide it again into pre specified training and validation sets. So by the end of 2023, we expect to be able to lock the model for the 22 cancer types from the training set. And then in another year, we will have this study readout for the validation set. So on Page 13, this is again to introduce you to you the proof of investigators for predict and questions. We are very proud that we have successfully attracted the top tier oncologists in China to lead the PREDICT and PRASHANT trials. So on top, Doctor. Jia Shen is the leading PI for the PREDICT trial. He is the fellow of the Chinese Academy of Sciences and also the President of the Shanghai Zhongshan Hospital. So for those of you who are not very familiar with the Chinese hospitals, Shanghai Zhongshan Hospital is one of the China's largest comprehensive academic hospital. And it performs more than 100,000 operations each year and serves about 100 and 69,000 in patients per year. And in 2019, it was ranked top 5 among China's general hospitals. So on the bottom, Doctor. Jie He, he is the leading PI of our patient study and he is also a fellow of the Chinese Academy of Sciences and also President of the hospital called Cancer Hospital of the Chinese Academy of Medical Sciences. This hospital is arguably the top cancer specialist hospital in China. So we are very proud that these top oncologists are leading our private and patient studies and it actually reflects the sharply growing interest and acknowledgement from oncologist community in China for cancer early detection, especially in the past year or so, it has attracted it did attract a lot of attention in this field. And also we think high quality of cohorts and data will ensure timely recruitment and of course successful operation of the study, which was served as a key in establishing a leadership position and maintaining our leadership position in the development of our cancer early detection products. So we are excited to share with you PIs and their level in among the oncologist community in China. So with that, I think I'll pass to our CTO, Doctor. Joe Zhang to tell you more about recent results released from the FQC2 study. Joe? Thanks, Shannon. So I'm going to cover a little bit about therapy selection part. So for the Slide 15, which is a highlight, what's the strength of Benelux in terms of the therapy selection business. So with regards to superior products as well as the NDA approval process for the different identity kit in the pipeline, but also the commercial penetration. But today, I'm going to focus on the first bullet point, which is the superior product, which one of the evidence shown is a paper published last month in Nature Biotechnology. This is in the Slide 16. This is basically a leading effort led by consortium committed effort FDA and which they call MAQC Consortium and focusing on the quality control of the sequencing business. So what you can see here, so the paper has been published. We participated in both the liquid biopsy part as well as the brain cancer, which is tissue based. So the liquid biopsy study is being published last month. So on Page 17 basically highlighted what's the participating assay and the study design. So there are 5 different companies participate. They are all kit vendor, which means all being capable to produce the liquid biopsy panel and as well as sell as a kit format and let the customer to use them. So Bunny Rock is the one of the only Chinese vendor who participated in this study and each vendor will distribute their kit to different labs. Also, the lab will receive the FDA distributed reference material and perform the assay based on the vendors' kit guidance and trying to generate the library and the sequence and also using the CATE vendors bioinformatics pipeline to perform analysis, then all the result will be submitted to FDA and for the principal investigator of this study, look at the data and based on the peer project also led by SEQC2 effort and trying to know which is a positive, what's the ground truth for this data and trying to evaluate sensitivity and specifically we call it positive rate as well as evaluate the reproducibility within that also across labs. So Benirock using the non plus non V4 panel, which we currently call onco COMPASS target panel for the liquid biopsy study and it covered 168 genes in this in the top table here. So for the next slide, Slide 18 basically highlights several key performance comparison across different kind of panel and product. So, Benirox has been brought in the green color. And as you can see here the top part, we see the fragment depth, which means based on the sequencing, how many unique fragments that we can collect or recovery recovered from 25 nanogram reference material. As you can see here, the data is showing the higher the better, which means with limited amount of DNA input, how many real fragments you can collect from the model. And the bottom panel compared coverage uniformly, which means like across this panel, what's the average coverage and how uniform this panel will be. And as you can see here, the Bunyong also showed good performance compared to other vendors. It's that all the coverages are close to similar to each other. So for the next slide, Slide 19, which compares 4 different hybridization capture plan that look at a different sensitivity, like say, how what's the calling probability for different kind of your frequency in building? And since each different panel has different kind of true positives, since the panel are different for the PI, they basically compare based on their ground choose and then look at different variant allele frequency being and whether this panel can be able to call it and each column represent one sample, one replicates in one site, in one lab. So, Brandy Rock is on the right most Basically, if it's been colored, which means this variance has been called, if it's blaring, which means this variance has been missed and we can't fall negative here. You can see here, basically, almost all the color have been filled for Bunin Rock, even if it's like 0.1% to 0.2% of your frequency being and the calling percentage is higher than some other panel. This is just to give us a lot of confidence showing like our panel as well as our biosimilaris pipeline showing pretty top performance compared to these other vendors, especially this kind of classic molecular biology vendors. So for the reproducibility slide, next slide, Slide 20, we just compare across different kind of panel look at how reproducibility it is across panel, across lab or within lab. So each lab process same sample 4 times, also there is a multiple labs performance. As you can see here, the reproducibility also, Brandy Rock Stadium is showing pretty good performance compared to other vendors kit. For the Slide 21, just briefly compare different kind of input of DNA amount and compare the sensitivity as we know like the more DNA put in there and within the higher sensitivity there, each panel showing this kind of trend. For the green line, basically represented Bernie Ross assay and for both sensitivity as well as reproducibility showing the Benu Ross panel as a pretty good performance and very stable on that. And also even at a very high frequency of 0.1%, 2.5% showing high performance. For the Slide 22, very briefly, just to compare analytical accuracy based on the sensitivity and precision curve. And as you can this is based on 25 nanogram of reference material input and across compared for different panel and the precision representing the positive predict value PPD and the sensitivity here means recall, which means how sensitive how many the true positive rate it is. So the closer to the top right corner of this graph means the better performance. As you can see here, the overall analytical accuracy and the specificity of GreenRock showing the best compared to other panels. So all this information just give us from the paper published and then this is basically a relatively, I'd say, fair comparison across different panel and using same reference material and give us a lot of confidence showing that our product in the surface selection zone, showing the top performance, not only in China, but also compared to what a lot of famous kickstanders and we've been doing pretty good on that. So here basically, I just conclude the service selection part highlight and I'll hand over to Leo talking about financial. Thanks. Thank you, Joe. Our financials are shown on Page 25 of our presentation. And for this call, we'll focus mostly on our top line numbers. And first, we recap that all our revenues are generated from our therapy selection business. So there is no contribution from anti detection yet, which is still under R and D and clinical development. In the Q1, we are happy with the year over year growth that we've been able to achieve. We grew our revenues by 58% on a year over year basis. We grew our gross profits by 72% on a year over year basis. By channel, our central lab revenue grew 72% Our in hospital revenues grew 70%. In our observation of some anecdotal industry data points, this is above industry growth rate, indicating that we've been able to gain some share in this period. Within the Q1, talking about the monthly and the sequential trends, January was impacted by COVID resurgence in Beijing, Shanghai and a few other key cities in China. So that did have a negative impact on testing volumes for some of our key customers. February was a quiet month due to Chinese New Year and March was an opening month. Because of the negative drag from January February, the sequential growth rate was negative for the Q1 at minus 14% Q on Q. Now looking at the rest of the year, we have a guidance of RMB610 1,000,000 for the 2021 full year, which is unchanged from our previous earnings release. We have not hit the monthly run rate yet to achieve that full year target. So there is certainly more work that we need to do. And in terms of what we're doing in terms of driving additional NGS penetration, we are doing, number 1, executing our in hospital strategy, putting our test available at more hospitals, which we think is important for building NGS penetration because this is the most typical format of testing in China. And second will be the continued execution of our multiyear NNPA registration pipeline process, which will be key in terms of competitive differentiation. So we remain focused on these initiatives for driving the long term success of our Therapy Selection business. And with that, we conclude our prepared remarks and we open up for questions, please. Thank Your first question comes from the line of Doug Schenkel from Cowen. Please go ahead. Hi, good day, and thank you for taking my questions. Starting on the topic of asymptomatic screening, I appreciate all the detail you provided today. Regarding the 6 cancer, 9 cancer and 22 cancer asymptomatic screening programs, Three things that are pretty important remain unclear to me. 1, do you believe studies like THUNDER, PREDICT and PRECIENT will be sufficient to allow for product launch from a regulatory standpoint and reimbursement? 2nd, if not, how big a study will be required to allow for regulatory approval and reimbursement? And third, what is the acceptable target from this perspective when it comes to sensitivity and specificity? I want to go back to these questions because you referenced CCGA and Pathfinder in your prepared remarks as good precedents or at least comparable studies. Neither of these studies are sufficient in the United States to support FDA approval for CMS reimbursement. Most companies, in fact, that are based in the West have indicated FDA approval and reimbursement would require large, but So I understand your programs are not targeted at the U. S. Or Western markets. They're targeted at China. So it just would be helpful to different. Okay. Hi, Dot. Thanks for your question. I'll take your question. I'll try. So first of all, a very straightforward answer for your first question. No, we don't think the predict or impressions will be enough for testing the asymptomatic population because they are apparently not powered enough to have a precise enough assessment on the sensitivity, especially the sensitivity for the asymptomatic population. And also the recruitment strategy actually naturally conflicts with post facto asymptomatic validation study because in this study, the participants, the control arm actually, we define them as healthy quote unquote because they need to go through a health checkup or physical examination once at the recruiter point. But actually for a purely asymptomatic study, they don't necessarily have to go through that. It's just symptom free and relying on whatever health checkup habits they are going through in their real life. So predict and question are not designed they're not designed to give us answers for the asymptomatic population performances. They are, on the other hand, powered or designed to give us answers for the case control cohorts, which will help us to define or to design the future asymptomatic, prospective or even interventional study. However, for the fixed cancer test, the study that we I just mentioned that's under planning, that one will be designed and powered to give us a definite answer for the asymptomatic population for the sick cancer test. So that one, we do think or we are designing it for the purpose potentially down the road for registration. Of course, the registration pathway for early detection products in China is not critical here or anywhere near critical here at this point. We're having a conversation with an MPA. So we don't have a 100% answer that it's going to be enough, but at least it's powered to answer the question about an individual level, whether the benefit would be enough to pass the product through the registration at least as a pay out of pocket product. However, for reimbursement, I agree with you that it's a completely different story because for reimbursement, you don't have you not only have to establish an individual level benefit, you have to establish a population level benefit. So in order to do that, there are a lot more you need to evaluate beyond just sensitivity specificity on that individual level for different cancer types, etcetera. You also have to establish benefits in terms of health economics, etcetera. So in China, I think we also talked about this a few times before. In China, the market is an out of pocket market. So we do believe that in China, it's possible to have the registration and reimbursement. There are sort of 2 things and we won't be able to have the registration by just showing the individual level of benefit. But of course, again, this is preliminary thoughts and things might change and we might have new information as time goes on. Thanks so much for that, Shannon. That was really helpful. I think my other topics are probably more for Leo. So Leo, just in terms of the quarter and specific to the central lab, volume dropped relative to Q4. Maybe that wouldn't have been shocking regardless of how January March went given Lunar New Year was in the quarter and wasn't in Q4. That being said, volume was also lower than Q3. Additionally, revenue dropped back to levels not seen since the Q2 of last year in this channel. And that was largely a function of both the volume dynamics and maybe just as if not more importantly, ASPs dropping a bit. So on the topic of ASPs, because it sounds like you don't think there's any competitive pressure on volume, it sounds like you think that's just the market. So when it comes to ASPs, were there market pricing pressures in the quarter? Or was that a function of product mix? And then kind of building off of that, how are you thinking about volume and pricing in the central lab channel over the balance of the year? Essentially, what's built into guidance? Yes. So for the central lab channel, we did see ASP fluctuations quarter over quarter, and that was more to do with product mix. So we have not made any pricing changes for that channel during the Q1. So pretty much driven by product mix shifts and some seasonalities. So that was for the Q1. For the remainder of the year, for the central ramp, we think there are structural challenges that we'll we do need to work through. And if we look at as we mentioned earlier, if we look at building NGS penetration, we are putting efforts into the in hospital channel. We think this is will be very important for the future growth of NGS penetration. That's the most typical format of testing. The central lab channel is a more fragmented channel with lower entry barriers, whereas the in hospital channel is a more institutionalized channel where our product strength will be able to compete better, we believe, versus other non products and some aggressive commercial factors in the central lab channel. So we think looking for the rest of the year, in hospital channel will be important in terms of driving growth. For the in hospital for the central lab channel, we have been building our sales team and headcounts. So we have seen sales and marketing expenses increasing over time, and that's mostly due to headcount increases. So we are putting more manpower on the ground, speaking to more physicians to build up this channel, but we think that this will take time. So Leo, also keeping in mind that in hospital revenue dropped below levels generated in both the 3rd Q4 of last year. Would you attribute the performance in Q1 largely to normal seasonality and thus you feel pretty confident about a more pronounced ramp in the in hospital channel versus the central lab over the coming quarters? The 1st Q drop of the in hospital channel was expected as we were expecting Chinese New Year. And typically, there was not a lot of ordering during that month. Then the January COVID resurgence was unexpected, so that did hit us. Without that, we would have been better. Looking at volume trends, we were happy about the in hospital volumes for the month of March, which grew double digits. And we are keeping a watch on the second quarter as we have not closed the second quarter yet. Okay. And that's a perfect segue to my last question, which again is on guidance. Obviously, you knew Lunar New Year was in Q1 as it always is. It sounds like what surprised you was the COVID impact on January and probably more the central lab performance in March versus the in hospital performance in March. What is it that you saw coming out of the quarter and over the early part of Q2, which made you confident in reaffirming guidance in spite of the fact that it does seem like there were more headwinds in the Q1 than you might have anticipated? Yes. As we were building the guidance, we were expecting a second half heavy versus first half light of the year, and it played out that way. And we did leave some buffer for COVID fluctuations, and we did get hit by that in January. So not a lot of surprises in terms of looking at our guidance. Looking forward for the rest of the year, we do need to ramp up our monthly revenue run rates, which we haven't hit the run rate yet to be able to achieve that full year guidance. So we need to go back and work hard, and we look forward to update you guys in the next earnings call. Okay. Thanks to all of you. I appreciate all the details. Thanks, Tom. Thank you. Our next question comes from Ethan Choi from Bank of America. Please ask your question. Thank you for taking my question. I'm Yi Sun from Bank of America, and I will ask 2 questions on behalf of analyst, David Lee. The first is that, can you help some updated information about our fixed cancer test? Is there any changes for the timetable guidance? You mean for early detection commercialization? Yes. And the approval and the like our discussion with NIMP, is there any update? No. Last time, we talked about working within the EAP and also the prospective clinical trial. And in terms of the commercialization timeline, we are building our team for commercialization and operation. So we think that the key point for the early Tanger's early detection is for the consumer side. So we recently are recruiting our team from our consumer industry and also the Internet industry. And we believe that we have already found the right way to commercialize that. But at the same time, that's a totally new thing in the market. So how to build up SOP in this time and when you're trying to optimize the whole process? So, so far, we're seeing that everything is on the right track. As we said, the commercialization will start early next year. And in terms of the clinical trial or settlement, I thought about that. Great. I don't think we ever gave any guidance on registration because we honestly are ongoing having an ongoing conversation with NMPA. So as I said, there's nothing unsure at this point. And also we are, of course, for the whole field, the early detection products, the registration pathway for that is not clear yet. So I think it's a dynamic process or discussion with the NMPA. So we don't have a specific timetable that we could give out yet. But as you said for all the progresses or efforts on getting everything is going as what we planned or expected, including the study that we are planning for among the asymptomatic population. That's the ongoing as expected and also as what we have released the last time. Thank you. Very clear. And the second question is about the participants seeing our PREDICT and the patient study. Do you think that for the PREDICT study, there are around 14,000 participants, while for the patients, they are around 12,000. And can you help us to illustrate more about how this number has been confirmed and why there is a difference and why the patients has fewer number of participants? Thank you. Thanks. It's a very good question. Thank you for noticing that. Actually for PREDICT, because we have quite rich preliminary results for those 9 cancer, at least 6 out of the 9 cancers and there's still a little bit data on the other three cancer types. So we were able to design the study and planning for the central size a stage specific estimate for the sensitivity and PO accuracy. So that's why for PRODICT, even with fewer cancer types, we will we are planning for a larger sample size because the sample size was calculated so that on each stage, each cancer type, each stage, we will have a precise enough estimate for the sensitivity. However, when we are planning for the question set because it's longer down the road and also because apparently we don't have as much preliminary data or knowledge about the other 13 cancer tests as for these 9. So that's why when we designed the PRASHAN study, it's more on a cancer specific estimate for the sensitivity instead of cancer and stage specific estimate. So that's why for question for each cancer type, we actually have a smaller sample size. And also even among predict and question, each cancer type actually has different sample size planned in terms of or depending on our estimated sensitivity that we will be able to reach or achieve. So especially for the PRASHAN study, we actually have fewer samples planned for the 9 cancer types that were already covered in the PRYTICS and we allocated more samples for the other 13 cancer types. But all in all, the design, the sample size calculation was based on different objectives. That's why you see different sample sizes for each cancer type. Okay. Thank you. Very clear. That's all my questions. Thank you. Our next question comes from Sean Wu from Morgan Stanley. And I actually am also quite curious about the full number you used for the 2 studies. You have exactly 14,026 for 56 and 11,879. So how did you come out of those numbers? That's for my curiosity. So for one study, you have to be benign, can't be higher than the other. I mean, in some sense, why don't you just combine them together? So no tool, sorry, to totally clearly are designed for different purpose, I suppose, for 91. Should we expect that you will get a more come out of this. We're effectively designed a product for one type of cancer detection like liver cancer or prostate cancer. What's the difference in the advantage of both hypertrophic like single one? For liver cancer, clearly, if people drink a lot of it, there's one type of it, it's a possibility and time to put them. And then finally, I think you have signed up some very good oncologists at the top hospitals, on top hospitals who do the finished trials. How Okay. Well, first of all, for the study sample size, as I just previously explained, predict and questions are designed based on different objectives. For predict, we are aiming to estimate stage and cancer test specific sensitivity. So for each cancer test, we allocate it more simplified. But for patients, because we didn't have as much previous knowledge to support safe and cancer testing design, that's why we actually will only assess the cancer type or cancer specific sensitivity. So roughly, that's why for Prussian, we have a little bit fewer sample size planned for the Prussian study. And also for your question, why don't we just combine the 2? Because they are for 2 different products. Because for the PREDICT study, we're using our 9 cancer test product. And then for the question that we will use our next generation of 22 cancer test product. They are not just it's not an add on relationship between the two products. Actually, the chemistry and also the marker selection and the model were all changed and hopefully will all improve between the 2 generations. That's why for the new generation, we will have to retest its performance to see whether it holds or it improves on the existing lung cancer that were already tested in PREDICT. And for your last question about the principal investigators, thank you. For the comment, we are also very proud. And as I said, it reflects actually the strong interest and attention that early detection has drawn among the oncologist community. I would say about 3 years ago, none of them really looked in the new technology is getting close to real clinical application or to make real contribution to cancer early detection. But nowadays, a lot of them they live in that and they think the new technology, especially the epigenetic based biomarkers plus machine learning and next generation sequencing, finally, it's bringing into reality that early detection can be realized in a large scale, especially detection can be realized in a large scale, especially on a multi cancer application. That's for 1. And for 2, there are actually very few hospitals in China that has the capability and capacity to be able to host studies like PREDICT or PRASHAN or lead studies like these. It actually requires a lot of organization power and also the impact from the clinical investigators. So that's why actually only the top clinicians or oncologists in China have the capability and impact to be able to operate these really large cohort studies. I think that's also why they have the passion and the ambition as well to fulfill these very innovative studies. Does that answer your question? All right. Thank you. So ladies and gentlemen, we have reached