 Good morning everyone. Thank you for the introduction. So I'm a senior reviewer in the division of molecular genetics and pathology in OIR at FDA. So normally I review genetic tests, such as the chromosomal microarrays for the postnatal diagnosis, companion diagnostics, and next generation sequencing based test, and also some of the non-invasive prenatal testing. I also serve on the steering committee for the precision medicine initiative at FDA, and I oversight the precision FDA projects. So this is my disclaimer slide. For today's talk, I'm going to briefly go over the basic components of IDE submission, and then I will focus on the analytical validation studies that are required to be provided in IDE submission. So Jonathan has gone over the both analytical and clinical part, but for my talk I will focus on the analytical part, and then I will use one example to demonstrate the analytical studies submitted to the IDE submission. So this slide lists basic components of IDE submission. So basically we like to see what your device is intended to be used, and then how your device looks, what components you have in your device. So Jonathan asked questions earlier how you define the device. So it's your choice based on your purpose in the clinical trial, so you can define what components you have. You just let us know clearly that how you intended to use it, and then what component it has, and also how it works to generate the test results. And then in the clinical protocol, we like to see what's the targeted population, how many clinical sites needs to be used to enroll patients, and then how many patients will be tested using the device. And then this, how the device is planned to be used in the clinical trial will help us to understand your device and then understand your risk of this, the uses of your device in the clinical trial. And then the other part is the abbreviated analytical data. I will show you, we don't ask much in the analytical validation in IDE, and then there are also other some administrative elements. You can find them in the IDE checklist. Those are the elements that we normally follow when we do the review. So for this talk, I will focus on the analytical part only. So when you consider how you designed the analytical validation study for your IDE submission, you first need to think how you can demonstrate that your device can accurately and reliably detect and analyze, so by the intended user in different situations. So it's exactly like what Jonathan just mentioned. So the accuracy, precision, and the sensitivity models. So in this case, when you conducted the analytical studies, the intended type of clinical specimens are preferred. So this will get into the pre-analytical studies. So you need to consider how the specimens are handled, because different specimen types will require different studies. For example, the serum plasma or urine, they will have different metrics comparison studies. And the preservatives or stabilizers in the collection devices may impact your device performance, so you need to consider those factors in your study design. And also, if archived samples are used for the clinical studies, then it's very important to study the sample stability. So these are the factors you need to consider when you design the analytical studies. So in IDE, we really don't ask a lot of analytical validation studies. We only ask for critical components, the analytical accuracy, sensitivity, or we also call it limit of detection study, and then the precision or reproducibility study, and then some analytical studies. And if applicable, analytical specificity study need to be conducted. So on the right here, I'm showing what the studies are required in other submissions, in 510K, in PMA submissions. You can see we ask a lot more studies in those submissions. But in IDE, we only ask these studies. And also, I like to point out that the purpose for the IDE submission is to demonstrate that your device is plausibly effective. So we don't ask a large study that we can draw the statistical inference from the study. So the sample size of the IDE study can be small. So both quality and quantity of the studies between the IDE and 510K PMA submissions are different. And of course, when you conduct the IDE studies, there will be a lot of challenges. Like what Jonathan mentioned, in the accuracy studies, there may be a lack of comparative method or reference method. But normally we accept analytically validated comparative method to be used in the accuracy study. So if you can justify why the reference method is not available and why you think this comparative method is a validated and appropriate comparative method, then you can come to us and discuss through the pre-submission process, then we will give you our comments and the suggestions. Then you can use those comparative methods in your study. And also I want to emphasize for the precision study, it's important to use samples around the cutoff because we want to see the performance for those difficult to test samples or samples near the cutoff regions. So those will be the important samples to be included. So next I'm going to use the NGS-based ANCLE panel as an example to show what analytical studies we ask for. So the NGS-based ANCLE panel is a single test that can detect multiple biomarkers. Sometimes it can be used for multiple indications. So this creates the challenges for the traditional regulatory paradigm, such as in terms of clinical validity and analytical validity and all those robustness across different tissue types. And also for the next generation sequencing technology itself, it has challenges and unique aspects, such as it has potential to detect novel and rare variants. So the technology is evolving rapidly. As Dr. Kingsmore just mentioned, there will be new aligner coming out, there will be new algorithm coming out, there will be new chemistry coming out. Then how do we deal with those changes? This challenge. And then in terms of the analytical validation, we are facing a lot of challenges, as Jonathan mentioned as well. So because the panel can detect millions of variants, then how do we validate those variants? We cannot validate them one by one. Then what's the unit of validation? Should we do it by variant types, by different genomic contexts? Or how do we identify different specimen sources? If the platform claims that they can detect both FFP samples and the blood, then how do we do the different specimen types? And also there's lack of comparative method and reference materials. So currently the only available reference material is the genome bottle project reference material, the A12878. And then there's a new reference material just released but not been widely used in the community. And also for some rare variants, it's hard to find clinical specimens. So it creates another challenge for the analytical validation. So with all those challenges and issues, we acknowledge all these challenges. And then we are thinking what kind of validation study we like to see. So we like to see the basic key components of the analytical performance of a device. So in terms of the analytical accuracy, when the sponsor or investigator is designing study to select a sample panel, we like to see a panel to represent the variants that your device is claimed to be able to detect. So for example, even though you have a whole axon sequencing platform, or you have a big target panel next generation sequencing platform, but in your clinical study that you are coming into FDA, you only detect let's say 10 variants in two genes, then you only need to first specify these are the variants and genes that you are going to use in this clinical trial. Then for the IDE submission, you only need to show the data for those variants. And then when you select the sample panels, the clinical samples are preferred. But for the rare variants, if you cannot find sufficient sample numbers, then cell lines may be acceptable with sufficient justifications. And then for the reference method, as I mentioned, we do accept analytically validated comparator method. So if you have the reference method for the results that as Jonathan showed, you will summarize your result as sensitivity, specificity. But if you don't have the reference method, and you use the comparator method, then you will show the result as the percent agreement. So it's showing the similar metrics, the terminologies are different. And then it's important to know that the reference methods need to be pre-specified with the exception criteria and how it's going to be used. And then the results need to be summarized by sample and by variant types in terms of the percent agreement, the overall percent agreement, negative agreement, and positive agreement. For the precision and reproducibility study, it's similar in the sample panel selection to cover the variants that you want to detect or you claim to be able to detect. And then for the specimens, it's better to include different tumor types. If you have pan-cancer claim, so for example, in the clinical trial, you will say you want to enroll different cancer patients, then you will need to have different sample tumor types included in your analytical study. This is to show that you can do what you claim you are able to do. And also different tumor contents, like if you say I can detect tumor content with 20 percent, 30 percent, or 50 percent, then you need to have representative coverage of these tumor contents in your sample panel for the analytical study. And then as far as you have sufficient number of samples to cover different variant types across different genomic regions, then it's acceptable. And of course, so these like a tumor, content tumor types, it seems like it's oncology-oriented, but it can be generalized to other applications. So if you have questions, you can talk to us through the pre-submission process. Then we can give you our comments or suggestions to your questions and proposals. And for the precision and repro studies, you want to include different runs or different instruments if applicable, and then same methods to summarize the results. Another perspective in terms of the analytical study for the NGS test in ID is that we want to see the limit of detection of your device. So you need to show what's the minimum DN input or genomic DN input you need to use for your device. And then for the tumor content, can you go down to 5 percent of tumor out of this sample? And also another one is the mutant allele frequency of variants. For example, whether your device can detect down to 1 percent of the SMVs or 5 percent of SMVs. So these need to be specified by different variant types. So I think these studies are the performance that you also, so you as an investigator also wants to see because you want to know whether your device is accurate, whether it's reliable by different users. And then another one is the pre-analytical study. So this will include the specimen types, how the specimens are collected, whether it's corneal biopsy or other biopsy procedures, and whether you can handle the FFP samples or fresh frozen samples or only blood. So it all depends on your device description. And one point I want to emphasize is that, I said earlier, cell line may be acceptable to be included in the analytical study. However, if your platform claims to detect FFP samples, then when you use cell lines as part of your analytical study, then you also need to treat the cell lines by the FFP procedure, formerly fixed and paraphernalia embedded, the cell line, and then look at the performance. So we acknowledge all the challenges that the analytical validation is facing, such as what I have mentioned, so specimens can be handled differently, and it's difficult to obtain some clinical samples for rare alleles. And also there's lack of a comparator method or reference method. And also for the, when we get to the whole genome technology, it increase another complicated area. But with what I have, hopefully with what I have showed you, I'm telling you that for the analytical validation in the IDE, you only need to show the performance for the variants that you claim to be used in the clinical trial. And also a representative of variants on samples can be used in the analytical studies. Also I listed the analytical studies that we will require in the IDE submission. And as Paula also emphasized, it's useful if you go through the pre-submission process with early conversation with us, then you can find more information or guidance in this link. Yeah, with that, thank you. Next we have Dr. Haja El Mubarak, who's a master reviewer within the Office of Invitro Diagnostics at FDA as well.