AI-Enhanced Human Performance Research

AI-Enhanced Human Performance Research is sponsored by National Institutes of Health (NIH). Encourages research into AI technologies that augment human cognitive and physical performance in healthcare settings.

Expired RFA-MH-26-105: Multimodal Artificial Intelligence to Accelerate HIV Clinical Care (R01 Clinical Trial Optional) This notice has expired. For NIH, in limited situations, applications may be accepted on a case-by-case basis for a short period after expiration to accommodate NIH late or continuous submission policies . Contact the eRA Service Desk for any submission issues.

Check the NIH Guide for active opportunities and notices. Department of Health and Human Services Part 1.

Overview Information Participating Organization(s) National Institutes of Health ( NIH ) Components of Participating Organizations National Institute of Mental Health ( NIMH ) National Institute on Drug Abuse ( NIDA ) Funding Opportunity Title Multimodal Artificial Intelligence to Accelerate HIV Clinical Care (R01 Clinical Trial Optional) R01 Research Project Grant April 4, 2024 - Overview of Grant Application and Review Changes for Due Dates on or after January 25, 2025.

See Notice NOT-OD-24-084 . August 31, 2022 - Implementation Changes for Genomic Data Sharing Plans Included with Applications Due on or after January 25, 2023. See Notice NOT-OD-22-198 .

August 5, 2022 - Implementation Details for the NIH Data Management and Sharing Policy. See Notice NOT-OD-22-189 . Funding Opportunity Number (FON) Companion Funding Opportunity See Part 2, Section III.

3. Additional Information on Eligibility. Assistance Listing Number(s) Funding Opportunity Purpose This notice of funding opportunity (NOFO) seeks to leverage cutting-edge advances in multimodal artificial intelligence (AI) to accelerate HIV diagnosis, prevention, and treatment.

Funding Opportunity Goal(s) The mission of the National Institute of Mental Health (NIMH) is to transform the understanding and treatment of mental illnesses through basic and clinical research, paving the way for prevention, recovery, and cure.

Open Date (Earliest Submission Date) Letter of Intent Due Date(s) Renewal / Resubmission / Revision (as allowed) AIDS - New/Renewal/Resubmission/Revision, as allowed All applications are due by 5:00 PM local time of applicant organization. Applicants are encouraged to apply early to allow adequate time to make any corrections to errors found in the application during the submission process by the due date.

No late applications will be accepted for this Notice of Funding Opportunity (NOFO). Required Application Instructions It is critical that applicants follow the instructions in the Research (R) Instructions in the How to Apply - Application Guide , except where instructed to do otherwise (in this NOFO or in a Notice from NIH Guide for Grants and Contracts ).

Conformance to all requirements (both in the Application Guide and the NOFO) is required and strictly enforced. Applicants must read and follow all application instructions in the Application Guide as well as any program-specific instructions noted in Section IV. When the program-specific instructions deviate from those in the Application Guide, follow the program-specific instructions.

Applications that do not comply with these instructions may be delayed or not accepted for review. Part 1. Overview Information Part 2.

Full Text of Announcement Section I. Notice of Funding Opportunity Description Section II. Award Information Section III.

Eligibility Information Section IV. Application and Submission Information Section V. Application Review Information Section VI.

Award Administration Information Section VII. Agency Contacts Section VIII. Other Information Part 2.

Full Text of Announcement Section I. Notice of Funding Opportunity Description Novel, data-driven approaches will be necessary to fulfill the Ending the HIV Epidemic in the U.S. (EHE) key strategies to diagnose, prevent, and treat, and respond.

The use of transformative technologies in fields such as artificial intelligence (AI) can accelerate scientific discovery, leading to new breakthroughs and innovative strategies for ending the HIV epidemic. Opportunities for using AI in making predictions, real-time monitoring, or to improve decision-making to address HIV prevention, care, and treatment needs abound but have yet to be fully realized.

Recent advances in highly flexible, multimodal AI models have generated much interest and discussion about the promise of AI to transform research, clinical care, and public health. This NOFO seeks to leverage these advances by encouraging the use of cutting-edge multimodal AI models to expand our capacity to address the dynamic, complex, and evolving HIV epidemic.

Multimodal AI is an advanced machine learning model that integrates multiple modalities of data often encountered in clinical practice (e.g., text, images, and audio) to improve predictive accuracy beyond what simpler, unimodal models (i.e., using a single modality of data such as text or images) can achieve.

These multimodal models, however, face several technical challenges including quantification of multimodal learning (i.e., quantifying dimension of heterogeneity, identifying interconnections, and optimizing learning processes).

While multimodal AI has emerged as a technique that shows great promise in transforming healthcare, current challenges will require the integration of innovations for expanding capabilities to enable greater understanding and the model-based and domain-related explainability of the model outputs.

Several methods have been developed to enhance model-based explainability such as Shapley Additive Explanations (SHAP) and Deep Learning Important FeaTures (DeepLIFT) but are limited in domain explainability.

One innovation to significantly increase domain interpretability and explainability is through the use of knowledge graphs – structured data models representing the complex, multi-modal relationships between HIV symptoms, treatments, medications, behaviors, social determinants of health (SDoH), and so forth – and their dimensions – providing a more nuanced understanding of HIV which allows for better targeting of prevention efforts, optimizing treatment decisions, and improving patient experience.

Furthermore, embedding experimentally testable causality principles into knowledge graphs are essential for further enhancing domain explainability for high stakes uses such as in healthcare. To address the evolving complexities in HIV prevention, treatment, and care, integrating advanced technologies such as knowledge graphs into multimodal AI models offers a promising new approach.

Enhancing the explainability of multimodal AI models with causal knowledge graphs has the potential for more accurate, explainable, and interpretable systems that can be used in a wide range of data-driven applications including comprehensive patient histories, early HIV detection, personalized HIV prevention or treatment plans, remote patient monitoring, real-time patient support, epidemic surveillance, and more to address HIV prevention, treatment, and care needs.

Human-centered AI approaches (i.e., designing, developing, and co-creating AI with end-users for a deeper understanding human needs) are also critical for ensuring ethical considerations and stakeholder needs drive the design at all stages of model development: data selection and preparation, model building, training, and assessment, and evaluation and implementation.

As such, this initiative provides an opportunity for, multidisciplinary research teams to develop explainable, transparent, and trustworthy human-centered multimodal AI models to improve HIV research, clinical care, and public health.

Research Objectives and Scope The overarching objective is to support data-driven, technological approaches to accelerate HIV diagnosis, prevention, and treatment efforts, which is closely aligned with the priorities of the NIMH Division of AIDS Research, the NIH Office of AIDS Research, and the U.S. Department of Health and Human Services.

As such, this initiative supports the development of multimodal AI models to expand our capacity to address the dynamic, complex, and evolving HIV epidemic with three primary objectives: (1) development, adaptation, and evaluation of accurate, safe, efficient, and unbiased multimodal AI models to support a diverse range of HIV applications; (2) creation of knowledge graphs to enhance the interpretability and explainability for applications in HIV research, clinical care, and public health; and (3) exploration of synergistic integration of model-based explainability methods, knowledge graphs, and multimodal AI models for more precise model outputs and increased usability in HIV clinical care with a human-centered approach.

Cross-disciplinary teams such as end-users, domain experts, patients/clients, community members, bioethicists, data scientists, software developers, engineers, and policymakers are encouraged. Teams are expected to use best practices and meaningful stakeholder engagement to ensure data privacy, security, and consent.

Inherent in this initiative is the need for the use of a human-centered AI approach in the design, development, and application to ensure that the system is fair, ethical, and unbiased for a more transformative, meaningful, and sustained impact on HIV clinical care.

Development, Adaptation, and Evaluation of Accurate, Safe, Efficient and Unbiased Multimodal AI Models for HIV Applicants are expected to propose innovative projects focused on the development, adaptation, and evaluation of multimodal AI models that can support a wide range of data-driven applications.

These applications may include but are not limited to early HIV detection, personalized HIV prevention or treatment plans, and real-time patient support, epidemic monitoring and surveillance, and clinical trials research with the final model demonstrating the potential for a significant impact for the specified HIV-related application.

Multimodal AI models must integrate at least two different modalities of data (e.g., text, audio, images, videos) and draw from at least two data sources (e.g., medical records, claims data, pharmacy records, wearable devices, patient surveys to identify SDoH needs).

Applications should describe the entire model development process, including training and architectural design (e.g., data modalities and sources, representation learning, methods for model fusion), as well as how stakeholder perspectives will be integrated, and potential challenges (e.g., handling missing or unbalanced data, managing data dimensionality).

Considerations regarding equity and biases must also be included to ensure fair and unbiased model performance. In addition, applicants may consider adapting an existing multimodal AI model when appropriate for the specific HIV application and context.

This can be achieved by connecting existing models through the use of application programming interfaces (APIs), enabling new functionalities (e.g., to improve model reasoning capabilities).

Creation of Knowledge Graphs to Enhance Interpretability and Explainability for Application in HIV Research, Clinical Care, and Public Health Combining multiple modalities increases model complexity and challenges the interpretability and explainability of multimodal AI models, making it difficult for users to understand how or why a model, for example, predicted a cluster of HIV cases or recommended a specific HIV treatment regimen.

A promising novel solution to this issue is the integration of domain-specific knowledge using knowledge graphs. Knowledge graphs are designed to integrate complex, diverse data to capture nodes (representing real-world entities such as HIV symptoms, biomarkers, treatments, behaviors, SDoH) and edges (representing different relationships between nodes such as symptoms-treatment) in a structured manner.

Generating a domain-specific knowledge graph, however, is challenging due to the heterogeneity of health data and the need to ensure that the knowledge graph represents the complex causal relationships and is context-specific, dynamic, and evolving. Applications should describe the process for developing the domain-specific knowledge graph to represent HIV knowledge relevant to the specific HIV application.

Research teams may propose to develop or adapt an existing HIV-specific knowledge graph, but must address how they plan to assess, maintain, update, and expand the knowledge graph through automated or other means.

Exploring the Synergistic Integration of Knowledge Graphs and Multimodal AI Models for More Precise Model Outputs and Increased Usability in HIV Diagnosis, Prevention, and Treatment Frameworks for integrating domain-specific knowledge graphs into large language models (LLMs) have been proposed to enhance model performance and interpretability on a number of tasks (e.g., enhanced understanding and interpretation, encoding data, knowledge representation and reasoning).

Novel methods using synergistic approaches that allow for bidirectional reasoning between a LLM and knowledge graph is a new and exciting area that can lead to diverse applications in multimodal AI reasoning capabilities. Applicants are expected to explore the synergistic integration of the multimodal AI model with an HIV-specific knowledge graph for enhancing knowledge representation or reasoning.

Use of Best Practices in the Design, Development, and Use of AI Models In collaboration with community and implementing partners, several technical, ethical, and regulatory considerations are expected to be addressed in the application in order to build user trust in the multimodal AI model.

These include, but are not limited to, using secure approaches designed to protect data privacy, training models using balanced data, identifying and mitigating model biases, adhering to relevant data privacy regulations, monitoring data security and compliance with regulations, maintaining detailed documentation to enhance model transparency and accountability, such as by using model cards to describe the model, features, and intended use, and establishing a data governance framework for managing and protected the data including model use limitations.

Research teams are also expected to develop clear consent documentation and procedures with their community and implementing partners that will enable transparent, clear, and open communication with users.

Additional considerations when designing, developing, and using automated systems that have the potential to meaningfully impact access to critical resources or services such as healthcare are identified in the Blueprint for an AI Bill of Rights put forward by the U.S. Office of Science and Technology Policy.

The National Institute of Standards and Technology (NIST)’s Artificial Intelligence Risk Management Framework also provides additional guidance to better manage AI risks for individuals, organizations, and societies.

Research applications should demonstrate enhanced explainability of the multimodal AI model infused with HIV-specific knowledge relevant to the selected HIV application with pre-defined metrics and measures to demonstrate success/impact.

Applications for this funding opportunity must address the following: Development, adaptation, and evaluation of an accurate, safe, efficient, and an unbiased multimodal AI model that supports at least one HIV application as a demonstration of proof of concept but with the ability to support a diverse range of HIV applications in the future; Creating a new, or adaptation of an existing knowledge graph to strengthen the interpretability and explainability of the model outputs for application in HIV research, clinical care, and public health; Explore the synergistic integration of knowledge graphs and multimodal AI models for more precise model outputs and increased usability in HIV diagnosis, prevention, and treatment; Details ethical principles to ensure data privacy, security, transparency and consent and incorporates a human-centered approach (i.e., human-in-the-loop, with consideration of factors such as acceptability, appropriateness, feasibility); Meaningful engagement of key stakeholders to enable critical analysis, support fair and ethical use, and inform decisions impacting care; Formation of multi-disciplinary teams including, for example, end-users, data scientists, behavioral and social scientists, domain experts, patients/clients, community members, ethicists, software developers, engineers, policy makers, and so forth as appropriate for the proposed project.

As applicable, applicants are also encouraged to seek feedback and guidance from the Food and Drug Administration on potential future medical device submissions (Pre-Submission through the Q-Submission Program).

For NIMH applications involving clinical trials : Applications proposing to pilot an AI-generated treatment intervention model must comply with the NIMH experimental therapeutics approach which requires studies not only test the intervention effects on outcomes of interest in real world settings but also informs the understanding of the intervention's mechanism of action.

As such, the scope of work must specify the intervention target(s) and mechanism(s) and include the assessment and analysis of intervention-induced changes in the presumed target(s) and mechanism(s), that are hypothesized to explain the intervention outcomes. The NIMH has published updated policies and guidance for investigators regarding human research protection and clinical research data and safety monitoring ( NOT-MH-19-027 ).

The application’s PHS Human Subjects and Clinical Trials Information, including the Data and Safety Monitoring Plan, should reflect the policies and guidance in this notice. Plans for the protection of research participants and data and safety monitoring will be reviewed by the NIMH for consistency with NIMH and NIH policies and federal regulations.

NIMH and participating institutes will host a webinar for all prospective applicants to provide an opportunity to ask questions related to the scientific scope of this NOFO and technical details for applying. Prospective applicants are encouraged to submit their questions regarding the NOFO in advance of the webinar. Further details on where and when to submit the questions will be provided once the webinar has been scheduled.

Participants must register for the event, but participation in the webinar is optional. Please visit the NIMH meeting and events website for pre-application information and further details regarding the webinar for this specific NOFO.

Applications Not Responsive to this NOFO An application will be considered non-responsive to this NOFO and will not be reviewed or considered for funding if the application: Does not propose the development, adaptation, or evaluation of a multimodal AI model using two or more different modalities of data (e.g., text, audio, images, videos) from two or more data sources (e.g., medical records, claims data, wearable devices).

Does not propose the development or adaptation of an HIV-specific knowledge graph. Does not propose enhancing the explainability of a model by synergistically integrating an HIV-specific knowledge graph including appropriate metrics and measures to demonstrate model performance and enhanced explainability. Does not describe methods for ensuring data privacy, security, transparency and consent.

Does not involve a collaborative, multidisciplinary team with relevant scientific or technical expertise. Does not include a plan for meaningfully engaging community stakeholders and/or implementing partners.

Data collection or generation without significant effort toward model development, adaptation, or evaluation to inform HIV diagnosis, prevention, or treatment; Model evaluation without significant efforts toward model enhancement using knowledge graphs; Using model explainability methods in the absence of domain-specific knowledge graphs. For NIMH applications only, does not follow the NIMH Experimental Therapeutic Approach.

In recognition of the many considerations and requirements above, applicants are strongly encouraged to read the Institute-specific research priorities below and consult with the Scientific/Research Contact(s) listed in Section VII: Agency Contacts when developing plans for an application.

National Institute on Drug Abuse Areas of specific interest to NIDA include developing methods for advancing the use of multimodal AI in epidemiology, prevention, and treatment specific to individuals who use drugs and are at risk for or have co-occurring HIV with the goal of improving clinical care.

NIDA will consider applications that include, but are not limited to, the following types of approaches: Leveraging comprehensive data to integrate substance use-related information into knowledge graphs to inform HIV clinical care, using sources such as surveillance data, observational cohort data, clinical databases, data warehouses, clinical trial or observational data, prescription drug monitoring data, and policy databases that offer insights into relations between substance use and HIV.

Integrating experts into multidisciplinary teams described in this NOFO that represent perspectives on substance use disorder epidemiology, prevention, and treatment as well as persons with lived and living experience with substance use and HIV and providers that serve such clients and patients.

Improving the identification of people who are at high risk of SUD and HIV and/or who may be candidates for PrEP and substance use treatment in healthcare settings. Developing methods to better identify clinical opportunities to improve durable viral suppression among persons with SUD who have been prescribed antiretroviral medication.

Developing and evaluating novel approaches to better identify people who appear to be at highest risk for acquiring HIV and suffer the consequences of drug use and the settings or clinical contexts where they can be reached and offered services more optimally.

Developing, validating, and deploying predictive and generative models to better characterize prevention and treatment practice patterns and provide findings to inform clinical decision-making. These models should demonstrate ways to optimize delivery of HIV therapeutic and preventive modalities and associated clinical outcomes among people who use drugs and optimize reduction in drug use and associated harms.

Areas of interest also include, but are not limited to, studies that investigate methods for changing provider behavior, improving patient communication and retention in care, addressing social bias and stigma, and better addressing HIV viral non-suppression among people with OUD.

Investigators proposing NIH-defined clinical trials may refer to the Research Methods Resources website for information about developing statistical methods and study designs. See Section VIII. Other Information for award authorities and regulations.

Section II. Award Information Grant: A financial assistance mechanism providing money, property, or both to an eligible entity to carry out an approved project or activity. Application Types Allowed The OER Glossary and the How to Apply Application Guide provide details on these application types.

Only those application types listed here are allowed for this NOFO. Optional: Accepting applications that either propose or do not propose clinical trial(s). Need help determining whether you are doing a clinical trial?

Funds Available and Anticipated Number of Awards The participating NIH institutes intend to commit the following amounts in FY 2026: NIMH Division of AIDS Research intends to commit $2,000,000 total costs to fund two awards. NIDA intends to commit $2,000,000 total costs to fund 3-4 awards. Application budgets are limited to $750,000 in direct costs per year and need to reflect the actual needs of the proposed project.

The maximum project period is five years but the scope of the proposed project should determine the project period. NIH grants policies as described in the NIH Grants Policy Statement will apply to the applications submitted and awards made from this NOFO. Section III.

Eligibility Information Higher Education Institutions Public/State Controlled Institutions of Higher Education Private Institutions of Higher Education Nonprofits Other Than Institutions of Higher Education Nonprofits with 501(c)(3) IRS Status (Other than Institutions of Higher Education) Nonprofits without 501(c)(3) IRS Status (Other than Institutions of Higher Education) For-Profit Organizations (Other than Small Businesses) City or Township Governments Special District Governments Indian/Native American Tribal Governments (Federally Recognized) Indian/Native American Tribal Governments (Other than Federally Recognized).

Eligible Agencies of the Federal Government U.S. Territory or Possession Independent School Districts Public Housing Authorities/Indian Housing Authorities Native American Tribal Organizations (other than Federally recognized tribal governments) Faith-based or Community-based Organizations Non-domestic (non-U.S.) Entities (Foreign Organizations) are not eligible to apply.

Non-domestic (non-U.S.) components of U.S. Organizations are not eligible to apply. Foreign components, as defined in the NIH Grants Policy Statement , are allowed. Applicant organizations must complete and maintain the following registrations as described in the How to Apply- Application Guide to be eligible to apply for or receive an award.

All registrations must be completed prior to the application being submitted. Registration can take 6 weeks or more, so applicants should begin the registration process as soon as possible. Failure to complete registrations in advance of a due date is not a valid reason for a late submission, please reference the NIH Grants Policy Statement Section 2.

3. 9. 2 Electronically Submitted Applications for additional information.

System for Award Management (SAM) – Applicants must complete and maintain an active registration, which requires renewal at least annually . The renewal process may require as much time as the initial registration. SAM registration includes the assignment of a Commercial and Government Entity (CAGE) Code for domestic organizations which have not already been assigned a CAGE Code.

NATO Commercial and Government Entity (NCAGE) Code – Foreign organizations must obtain an NCAGE code (in lieu of a CAGE code) in order to register in SAM. Unique Entity Identifier (UEI) - A UEI is issued as part of the SAM. gov registration process.

The same UEI must be used for all registrations, as well as on the grant application. eRA Commons - Once the unique organization identifier is established, organizations can register with eRA Commons in tandem with completing their Grants. gov registrations; all registrations must be in place by time of submission.

eRA Commons requires organizations to identify at least one Signing Official (SO) and at least one Program Director/Principal Investigator (PD/PI) account in order to submit an application. Grants. gov – Applicants must have an active SAM registration in order to complete the Grants.

gov registration. Program Directors/Principal Investigators (PD(s)/PI(s)) All PD(s)/PI(s) must have an eRA Commons account. PD(s)/PI(s) should work with their organizational officials to either create a new account or to affiliate their existing account with the applicant organization in eRA Commons.

If the PD/PI is also the organizational Signing Official, they must have two distinct eRA Commons accounts, one for each role. Obtaining an eRA Commons account can take up to 2 weeks.

Eligible Individuals (Program Director/Principal Investigator) Any individual(s) with the skills, knowledge, and resources necessary to carry out the proposed research as the Program Director(s)/Principal Investigator(s) (PD(s)/PI(s)) is invited to work with their organization to develop an application for support.

For institutions/organizations proposing multiple PDs/PIs, visit the Multiple Program Director/Principal Investigator Policy and submission details in the Senior/Key Person Profile (Expanded) Component of the How to Apply-Application Guide. This NOFO does not require cost sharing as defined in the NIH Grants Policy Statement Section 1. 2 Definition of Terms .

3. Additional Information on Eligibility Applicant organizations may submit more than one application, provided that each application is scientifically distinct. The NIH will not accept duplicate or highly overlapping applications under review at the same time, per NIH Grants Policy Statement Section 2.

3. 7. 4 Submission of Resubmission Application .

This means that the NIH will not accept: A new (A0) application that is submitted before issuance of the summary statement from the review of an overlapping new (A0) or resubmission (A1) application. A resubmission (A1) application that is submitted before issuance of the summary statement from the review of the previous new (A0) application.

An application that has substantial overlap with another application pending appeal of initial peer review (see NIH Grants Policy Statement 2. 3. 9.

4 Similar, Essentially Identical, or Identical Applications ). Section IV. Application and Submission Information 1.

Requesting an Application Package The application forms package specific to this opportunity must be accessed through ASSIST, Grants. gov Workspace or an institutional system-to-system solution. Links to apply using ASSIST or Grants.

gov Workspace are available in Part 1 of this NOFO. See your administrative office for instructions if you plan to use an institutional system-to-system solution. 2.

Content and Form of Application Submission It is critical that applicants follow the instructions in the Research (R) Instructions in the How to Apply - Application Guide except where instructed in this notice of funding opportunity to do otherwise. Conformance to the requirements in the Application Guide is required and strictly enforced.

Applications that are out of compliance with these instructions may be delayed or not accepted for review. Although a letter of intent is not required, is not binding, and does not enter into the review of a subsequent application, the information that it contains allows IC staff to estimate the potential review workload and plan the review. By the date listed in Part 1.

Overview Information, prospective applicants are asked to submit a letter of intent that includes the following information: Descriptive title of proposed activity Name(s), address(es), and telephone number(s) of the PD(s)/PI(s) Names of other key personnel Participating institution(s) Number and title of this funding opportunity The letter of intent should be sent to: All page limitations described in the How to Apply- Application Guide and the Table of Page Limits must be followed.

Instructions for Application Submission The following section supplements the instructions found in the How to Apply- Application Guide and should be used for preparing an application to this NOFO. All instructions in the How to Apply - Application Guide must be followed. SF424(R&R) Project/Performance Site Locations All instructions in the How to Apply- Application Guide must be followed.

SF424(R&R) Other Project Information All instructions in the How to Apply- Application Guide must be followed.

This NOFO requires an attachment: A Meaningful Stakeholder Engagement Plan Meaningful Stakeholder Engagement Plan (Required - 2 page maximum) Include a detailed plan for meaningfully engaging community members and other implementing partners, including health system and public health leaders, clinical or public health informaticians, patients/clients, families, providers or other clinical care staff, payers, and community leaders, as appropriate to the specific goals of the study.

SF424(R&R) Senior/Key Person Profile All instructions in the How to Apply- Application Guide must be followed. All instructions in the How to Apply- Application Guide must be followed. All instructions in the How to Apply-Application Guide must be followed.

PHS 398 Cover Page Supplement All instructions in the How to Apply- Application Guide must be followed. All instructions in the How to Apply- Application Guide must be followed, with the following additional instructions: Describe how the proposed research advances the state of the science in multimodal AI for HIV research, clinical care, or public health.

Describe the benefits of using multimodal AI over other methods for the proposed HIV-related application. Describe how the proposed HIV application is likely to impact HIV research, clinical care, or public health. Describe the innovative methods that will be taken to ensure the model is extensible to allow for future HIV applications.

Describe the extent to which the data modalities and data sources selected are appropriate for model building. Describe the methods used for developing, adapting, and integrating the knowledge graph in the multimodal AI. Describe how the proposed methods for developing the new or adapted knowledge graph strengthens the interpretability and explainability of the model outputs.

In addition, describe the metrics and measures used for demonstrating model performance and enhanced domain explainability. Describe how the timeline for the model building, development and integration of the knowledge graph, and model evaluation are technically feasible and realistic.

Within the Meaningful Stakeholder Engagement Plan, describe plans for how relevant community members and other implementing partners (which may include patients, physicians, healthy system, public health, and community leaders) will be engaged in model design and testing. Describe the metrics that will be used to measure the impact of the engagement.

Resource Sharing Plan : Individuals are required to comply with the instructions for the Resource Sharing Plans as provided in the How to Apply- Application Guide. Specific to this RFA. The NIMH Division of AIDS Research embraces open science principles to promote the sharing of scientific resources with the broader community.

As such, a resource sharing plan is required for this NOFO. Applicants must describe their plan for open dissemination of protocols, methods, models, software, and/or code, findings and related tools to the community such that they are readily usable and extensible, where applicable.

The multimodal AI model, as well as the processes for developing and preparing the model (e.g., data curation, model training, model assessment), must be documented using appropriate methods (e.g., model card) to enhance transparency, accountability, and reproducibility.

All instructions in the How to Apply-Application Guide must be followed, with the following additional instructions: All applicants planning research (funded or conducted in whole or in part by NIH) that results in the generation of scientific data are required to comply with the instructions for the Data Management and Sharing Plan.

All applications, regardless of the amount of direct costs requested for any one year, must address a Data Management and Sharing Plan. To advance the goal of advancing research through widespread data sharing among researchers, investigators funded