 So, there's going to be a little tiny bit of my commentary in here, but I think what I heard over the presentations and the discussion was, again, iterating back to the last session of knowledge generation through a rapid learning system. I mean, we need to be able to learn more about variants from their phenotype, from the clinical outcomes. I mean, so this feeds into the interpretation of variants and the actionability of those variants. So that's a broad overarching comment that I heard. Guidelines, and then there are various areas where I've heard recommendations. In guidelines, we need to feedback information on the use of the guidelines to modify or refine those variant interpretation guidelines, like almost a closed-loop decision system, so that you know where the guidelines are falling short. And then also look at the guidelines that they're not addressing certain issues, like low penetrance or mild alleles or risk alleles, and how do you start thinking about those other kinds of guidelines? We need database improvements, reporting recommendations that include the evidence used for variant classification, as well as the annotation justification. And how we move to databases with case-level data is a very important next step. I think that's in ClinGen, but CSER also could be very important in this because you're looking at the total picture of the patient, which is very clinical, in a database rather than just the variant and the interpretation of that information. And then how will sites be supported in updating information in databases? I mean, if you're creating these databases for interpretation, what incentives do providers have to keep that data updated? On phenotyping, that was another large area that I heard points being made. We need to develop processes to re-phenotype based on genotype. And as our knowledge develops, we need guidelines for reinterpretation of sequencing data and how that will be supported. We have models for initial interpretation in guidelines, but not this reinterpretation aspect. So, I think that's something NHGRI may want to consider for CSER or CSER may want to consider. And then there were a range of clinical needs. One was whether or not there should be a comprehensive gene list. I know the U.K. has created a medically actionable genome that it uses, at least in one of the centralized laboratories in the U.K. And so I think CSER are looking at a more comprehensive gene list than disease-specific so it can be more global in what we're sequencing if we are going to use a gene list. And I also heard arguments about exome and genome sequencing being a much better approach for learning. And so some studies around that might be helpful. And then holes need somatic variant interpretation and database. That's not being addressed as well. Physician education, clearly. And then integration of genome-level data into the EHR. So those were some clinical needs. Thank you all. And we'll turn it over to the next group. Great. Thanks to those on the interpreting variance panel. So next we have assessing clinical utility, moderated by Mary Relling and Aril Chenayan with talks given by Robert Green and you and Ashley. And if I could just make a comment to those who want to ask questions, please make sure to introduce yourself so we know who you are and can capture it on the webcast. Thank you.