MAGen NOFO Webinar FAQ

Last Updated: July 18th, 2024

Questions and answers to help applicants responding to the ML/AI Tools to Advance Genomic Translational Research (MAGen) notice of funding opportunities (NOFOs): RFA-HG-24-004 and RFA-HG-24-005. The list will be continually updated to reflect new questions received.

1. Can foreign institutions participate in the MAGen Sites NOFO RFA-HG-24-004?

   Yes, foreign institutions can participate as components of an application in which the primary organization is a US institution. However, a foreign institution cannot be the primary applicants.
    

2. Can foreign institutions participate in the MAGen Coordinating Center NOFO RFA-HG-24-005?

   No
    

3. Can one Institution apply for both the MAGen Sites and the Coordinating Center NOFOs?

   Yes, applicant organizations may submit more than one application, provided that each application is scientifically distinct. However, we encourage you to carefully consider the roles and responsibilities of the investigators in the context of the review criteria in each such application. Please note the criteria outlined in the “Review and Selection Process” section of the NOFOs.
    

4. Does pathogenic variant refer only to germline variant?

   Yes, pathogenic variants refer to germline variants.
    

5. Can the proposal include development of a ML/AI tools for functional annotation of variants? 

   Applications that solely develop in-silico variant effect predictors, generate a catalog of variant effects, elucidate gene function, biological pathways or networks will be considered not responsive and will result in the application being withdrawn prior to review.
    

6. What is the requirement for describing the datasets in the proposal?

   The proposed multimodal datasets including EHR data should be well described including summary statistics and access to the datasets to allow their evaluation for adequacy and availability to develop and cross-validate tools. The summary statistics should include sufficient details to assess the proposed ELSI projects and the other specific aims. 
    

7. Do you require each applicant team to include expertise to achieve all of the goals of the NOFO? 

   Yes, the applicant teams need to self-reliant in all of the expertise required to achieve the stated aims and thus include multidisciplinary expertise to support the development, validation and cross-validation of the ML/AI tools and the ELSI framework and conduct the ELSI research projects.
    

8. Based on this phrase in the MAGen Sites NOFO (RFA-HG-24-005) 'Funded MAGen Sites will work collaboratively with each other and the MAGen CC (RFA-HG-23-005) within the Consortium to meet the goals of this RFA”, it appears that MAGen NOFOS were first released in 2023. Is this accurate?

   This is the first release of the MAGen NOFOs and therefore no MAGen awards have been made. The noted text has a typographical error. “RFA-HG-23-005” should be RFA-HG-24-005. The text however is linked to the correct MAGen NOFO URL.
    

9. Can you clarify this activity “Selecting the Consortium Gene Set for ML/AI Tool Development"? 

   You should propose gene sets and data that will help you answer the specific aims in your application. During the UG3 Phase, the Consortium will collectively select the subset of consensus genes and datasets for tool development and cross validation.
    

10. What is meant by “Guiding the Common Data Model Development”? 

    The Common Data Model is a standard and extensible collection of schemas (entities, attributes, relationships) that represents the data used by the AI tools with-defined semantics, to facilitate data interoperability. This model should allow for seamless deployment, training, and validation of models across MAGen Sites.
     

11. Can you clarify the requirements for the cross validation plan? 

    The cross validation plan should include a detailed description of how the tools proposed by the applicant MAGen Site will be made portable to allow the other MAGen Sites in the Consortium to cross validate their tools either in a Consortium-selected common platform or at a MAGen member Site. The cross validation platform could be a cloud-based platform, an on-prem computing cluster, or any other HPC system that is adequate for the task.  A description of any anticipated challenges to implementing and cross validating tools developed by other MAGen Sites in platforms they propose for such cross validation should be described.  

    Applicants are welcome to propose their own datasets to cross-validate their models (perhaps data from an alternate medical system with different population characteristics and other features). The application should also include a detailed description of the availability and any anticipated barriers to the use of datasets for cross validating tools developed by other MAGen Sites.  

    Applicants are reminded that ultimately, during the course of the UG3 phase, the Consortium-wide cross validation plan will be developed anew and jointly by the Consortium in concert with NHGRI program staff to best meet the goals of the NOFO. 
     

12. Are M/AI models developed to enhance the prediction of penetrance of pathogenic variants responsive to the NOFO?

    Yes, predicting penetrance of pathogenic variants could enhance the prediction of how individuals with pathogenic genetic variants manifest disease—the goal of this NOFO—but applicants should clearly describe how ML/AI models predicting penetrance relate to how individuals manifest disease and identify genomic and non-genomic factors that affect their penetrance.
     

13. Are M/AI models developed for variant effect prediction responsive to the NOFO? 

    No. However, you may incorporate existing variant effect predictors within your model. 
     

14. How integral is clinical expertise to the activities of the NOFO? 

    Clinical expertise in the proposed disease should be integrated at all stages so the developed tools and other resources will be informed by such expertise and have utility in clinical settings in the future. 
     

15. What are potential outcomes of tools? 

    Outcomes of tools should be useful in Genomic Translational Research as defined in the NOFO, for example, patient stratification to tailor recommendations related to preventative care, treatments, frequency of screening etc.
     

16. The RFA states that applications cannot generate new primary data for this project. However, would collecting data for the ELSI objectives be in scope of the NOFO?

    Yes, collecting data for the ELSI objectives would be in scope of the NOFO. Note, data generation is not within the scope of the NOFO for training and cross validation.
     

17. The RFA suggests we should "Provide verification of accessibility, for example, include evidence of approvals by data access committees, the Head of Research, the legal department and the Chief Information Officer, Chief Security Officer, etc." - some of these may be appropriate as Letters of Support, but others may not (e.g. approvals). Are these allowed in an appendix?

    Yes, please include the designated authority approval(s) for using the data as proposed in your application.
     

18. Are any specific disease groups that are being targeted?

    The NOFO is intended to be disease agnostic and applications can focus on any diseases.  Please see review considerations listed in subsection 2. Review and Selection Process under Section 4 Application review Information. Also, during the UG3 Phase, the Consortium will collectively select the subset of consensus genes and datasets for tool development and cross validation.

    The National Institute of Aging has issued a Notice of Participation in the MAGen NOFO (NOT-AG-24-019) that invites proposed MAGen Site applications that align with NIA's mission.
     

19. How will collaboration within the research consortium be facilitated?

    While each MAGen Site is responsible for working collaboratively with all members in the Consortium to achieve the goals of the NOFO, please see RFA-HG-24-005 ML/AI Tools to Advance Genomic Translational Research (MAGen) - Coordinating Center for how they will support the Site activities.

     

20. The RFA describes a Project Manager (PM) role with effort - can you clarify if this is a distinct role from the PIs, and if so, what are the expected duties for this role?

    The Project Manager (PM) responsibilities include all activities (administrative/day-to-day/strategic)  and would be the single and primary point of contact for other groups. This role can be done by a PI, if they have the expertise and if adequate time is dedicated specifically for such activities.

     

21. To what extent are the ELSI research project aims meant to be fully developed now versus at the end of Y2, where they might better reflect the Consortium’s decisions in Y1-2 (I thought an earlier slide suggested the ELSI research aims would be a Y2 deliverable for proceeding to the UH3 phase)? 

    At least one ELSI research project that is relevant to the proposed ML/AI aims should be described in the application.  Applicants are referred to details in the funding opportunity under “Research Plan” which provide instructions. 

     

22. Should applicants prepare specific aims for each phase? 

    Applicants must provide the overall goals for the entire application and indicate separately, Specific Aims to be accomplished in the UG3 Planning-Exploratory Phase, and those to be accomplished in the UH3 Implementation Phase. Applicants are referred to details in the funding opportunity under “Research Plan” which provide instructions regarding the “Specific Aims,” “Research Strategy,” etc. 

     

23. Who will select the members of the External Scientific Panel (ESP)?  NHGRI or the MAGen Steering Committee? 

    NHGRI/NIH. We will ensure selection to avoid conflicts. 

     

24. Will the consortium require use of a single IRB? If so, is there an expectation of which site will support the sIRB?

    This is unlikely, but not inconceivable since we are unlikely to have a single protocol that will be used by all of the Consortium members.  

     

25. Can the researchers share the original data or the embedding of them? 

    If the institution and IRB permits the sharing of the original data, then we welcome that. If embeddings will be publicly released is a question around identifiability and privacy in the ELSI research that will be addressed during the course of this project. 

     

26. Is there any desire to connect these projects to public benchmarking challenges, such as the DREAM challenges? 

    This option should not be included in the application but could potentially be undertaken if that is jointly decided as appropriate by the consortium and NHGRI. 

     

27. Would you accept genes with pathogenetic variants that are predictive of disease without an available cure, but for which there are available symptoms and supportive therapies where early intervention could be helpful, or are you mainly looking for genes for which there are consensus clinical guidelines for screening, prevention, etc. 

    Such an application will be responsive, assuming it fulfills the other criteria laid out in the NOFOs. 

     

28. Will the coordinating and contributing sites be independently responsible for budgeting for AnVIL cloud computing costs?   

    Yes, all compute costs proposed in the application should be budgeted by the application. The coordinating center's responsibilities will be leading the development of the data model which may entail the use of the cloud for which costs need to be included in the proposal.  

     

29. There aren’t any good genomics representations in the OMOP common data model. Do you except the consortium to figure this out? 

    The data model will be developed with collective input from the consortium members to best meet the goals of the NOFO. 

     

30. Will AnVIL be making any plans to support collaboration with MAGen?  

    The NOFO says, “Enabling cross validation in the NHGRI Analysis Visualization and Informatics Lab-space (AnVIL) or any other NHGRI-approved computational platform proposed by the Consortium.”  The AnVIL Clinical Environment for Innovation and Translation (ACE-IT) project is developing the AnVIL Clinical Resource (ACR) to support the analysis of clinical genomic data. The Sites supported by the CC will be responsible for activities required to use the approved computational platform, which could be this platform or other approved platform. 

     

31. Given that details regarding data being used in the study won’t be available until after the final selection of the Dev Sites is announced, can you confirm that the most appropriate Human Subjects form for the CC to use in the application is the delayed onset option?  

    Human Subjects designation is based on what you propose to do in your application, it does not have anything to do with which applications is finally funded.