NIH Data Management and Sharing Policy (DMSP)

A. Types and amount of scientific data expected to be generated in the project:

Example 1:

Demographic, clinical, MRI, and fMRI imaging data will be acquired from 110 affected youth and 110 matched healthy controls.  All data will be de-identified prior to receipt by the repository, but the information needed to generate a global unique identifier for the NIMH Data Archive (NDA) will be collected for each subject. 

Example 2:

This project will produce 256-channel electroencephalograms (EEG) data and functional magnetic resonance imaging (fMRI) images. Data will be collected from 500 research participants and 500 datasets totaling approximately 10 Tb in size will be produced. The following data files will be produced in the course of the project:

• EEG files
• fMRI files
• Metadata files

Raw data will be transformed by data processing methods and the subsequent processed dataset used for statistical analysis. To protect research participant identities, de-identified data will be made available for sharing.

Example 3:

Our genomic study will be registered with dbGaP, and our raw whole genome sequencing data and derived data will be submitted to the NIMH Data Archive (NDA). Phenotypic and clinical data for all 500 research subjects will be collected and deposited in NDA using the data dictionaries available in NDA.

Example 4:

The data to be shared will include MRI images and clinical assessments from human research participants.  This application is focused on secondary data analysis from existing data but will also deposit privately held data to a public repository.  The existing data is available from the NIMH Data Archive (NDA) in collections 2134 (148 subjects) and 2433 (47 subjects).  In addition, we have data from a previous study involving 155 research participants with major depressive disorder that have not yet been shared with the research community but will be uploaded to NDA during the second quarter of the first year of funding.  As discussed in the application, structural MRI scans are available for time points before and after treatment along with relevant clinical data.

Example 5:

As detailed in the Research Strategy Section, we propose the generation of a spatially mapped single-cell atlas of the developing mouse brain and include specific deliverables.  Our primary deliverable for each modality will be a matrix of cells × (counts in peaks for ATAC, UMIs in genes for RNA, or methylation status for DNAm) along with a dense metadata table with information for each cell. This includes the animal sex, developmental time point, punch of origin with x,y,z coordinates, assigned cluster and inferred cell type, assigned subcluster and inferred cell type, as well as a number of QC metrics (total reads, passing reads, reads in peaks, TSS enrichment, cell barcode combination, date of preparation for each stage, sequencing platform, likelihood of being a doublet, and any other relevant metrics that arise during the project). 

The amount and type of data from human cells will depend on the results from the mouse studies.  Data sharing plans will be updated when appropriate (likely at the start of year 4 of the grant award).

B. Scientific data that will be preserved and shared, and the rationale for doing so:

Example 1:

Sufficient data from this project will be preserved to enable sharing via NDA data of sufficient quality to validate and replicate research findings described in the Aims.  NIMH requires data measured from human subjects to be shared using the NDA. 

Example 2:

Based on ethical, and legal, considerations, the following data produced in the course of the project will be preserved and shared:

• EEG files
• fMRI files
• Metadata files

Example 3:

All raw and processed genomics files and all clinical and phenotypic data will be shared.

Example 4:

This is a secondary data analysis application, so new data is not being measured. Much of the data is already available through NDA. Clinical and imaging data from 155 new subjects will be shared.

Example 5:

The data described in section A will allow researchers to reproduce our publications and will allow them to collect additional data in a similar way to extend our results.

C. Metadata, other relevant data, and associated documentation:

Example 1:

In addition to the subject level data described above, all fMRS and fMRI task related paradigm designs and experiment definitions will be deposited in the NDA.

Example 2:

To facilitate the interpretation of the data, the study protocol will be deposited for sharing, providing details on the instruments and techniques used to collect the 256-channel EEG from the 500 subjects. Metadata files explaining the position of EEG electrodes and how to process collected values will be shared and associated with the datasets.

Example 3:

The Institutional Certification will be submitted to NIH during the dbGaP registration process once we have been told that a grant award is likely.  Within the first six months following the award, we will submit the Data Submission Agreement to NDA and will create the Data Expected list in our new NDA Collection.  A brief study protocol will also be submitted to NDA and will be made freely available.

Example 4:

Preparation for submitting existing data to NDA is largely complete.  Within the first six months following the award, we will submit the Data Submission Agreement to NDA and will create the Data Expected list (see Standards section) in our new NDA Collection. The policies of our institution mandate that exact dates will not be shared (see Access section).

Example 5:

In addition to a detailed methods section for any publications associated with this work, we will provide a detailed step-by-step protocol as a Supplementary Protocol document and maintain active protocols.io protocols for each technology and workflow. 

We will additionally release protocol links as metadata to be associated with single-cell data deposited to the Neuroscience Multi-omic Archive (NeMO, https://nemoarchive.org/).

In addition to providing detailed protocols, our laboratory has hosted visiting scientists to train on the data analysis pipelines developed and deployed by the lab.  We welcome the opportunity to continue these training efforts.

Example 1:

The clinical data will be analyzed with custom Python code written using the statsmodels, numpy, and pandas packages, all of which are freely available. fMRI images will be analyzed using the SPM8 toolbox (https://www.fil.ion.ucl.ac.uk/spm/software/spm8/)  for MATLAB (http://www.mathworks.com/products/matlab/). While MATLAB is commercial software, most universities have site licenses available and the SPM8 toolbox is free. It is also possible that the toolbox might run in Octave, an open-source alternative to MATLAB (https://www.gnu.org/software/octave/index), but we have not tried it. All code will be shared on our lab website at https://github.com/labname the main readme.md file for the project will also include instructions and parameter choices for the GUI-based analyses.

Example 2:

The 256 channels EEGs and fMRI images data will be made available in CSV files, and DICOM format files respectively, and will not require the use of specialized tools to be accessed or manipulated.

Example 3:

Genotypic data undergo an extensive automated data cleaning process in the laboratory.  Our replication plan for observed associations is outlined in the Research Strategy.  While all sequencing data from this proposal will be generated using Illumina pipelines, differences in read depth and primer libraries between studies will require joint re-calling of all genotypes from raw read files to yield the highest possible quality calls and a harmonized dataset for future use in follow-up and unrelated studies.  Using the Broad Institute’s Genome Analysis Toolkit (GATK), we will apply standard Best Practices workflows for single nucleotide variant (SNV) and Indel discovery from whole genome sequence alignment files (SAM/BAM).  These steps should ensure that final association results are representative of “true” genotypes rather than miscalls or confounded genotypes that are unlikely to replicate in independent populations.  

Example 4:

The basic statistical analyses described in the application will be done using R. We plan to use the MRI data analysis tools in the FMRIB Software Library (FSL) for multi-level modeling of group effects.  BrainVoyager software will be used for anatomical segmentation to isolate regions of interest within individual subjects, and the AI-powered analyses described in the application will use custom code written with the PyTorch library for Python. R, FSL, Python, and PyTorch are all freely available to the research community. BrainVoyager is commercial software, with licenses available for purchase from https://www.brainvoyager.com/. 

All R and Python code (including trained model weights) will be available on our lab Bitbucket page (located at https://www.thelab.edu/) no later than when publications are submitted. The Bitbucket page is publicly assessable and will be hosted for at least 5 years after the grant award ends.

Example 5:

All code and software that will be written to analyze the data will be deposited on GitHub (https://github.com/labname) for public access and be provided as Supplementary files for any publications. Code will be available no later than when a publication has been submitted.

Example 1:

Participant age, sex, ethnicity, height, weight, socioeconomic status, and other demographic data will be collected using the following instruments as defined in NDA:

1. Research Subject and Pedigree
2. Demographics Short Form
3. Ethnic Group Questionnaire
4. Height and Weight
5. Hollingshead Socioeconomic Rating Scale
6. Pubertal Development Scale 
7. Edinburgh Handedness Inventory
8. WASI-2
   

In compliance with NOT-MH-20-067, the following data will be collected to facilitate aggregation of this data set with other data sets:

1. DSM Crosscutting for Youth
2. RCADS-25
   

The clinical assessments we plan to collect for this study include:

1. Kiddie-SADS-Present and Lifetime Version
2. Children’s Yale-Brown Obsessive Compulsive Scale
3. Schedule for Obsessive-Compulsive and Other Behavioral Syndromes
4. Dimensional Obsessive Compulsive Scale
5. Yale Global Tic Severity Scale
6. Child Behavior Checklist
7. Multidimensional Anxiety Scale for Child Parent and Self
8. Conners 3
9. Adolescent Depression Rating Scale

Example 2:

Formal standards for 256 channels EEGs data have not yet been widely adopted. However, our data and other materials will be structured and described according to best practices. Data will be stored in common and open formats, such as CSV and JPEG files for our 256 channels EEGs and fMRI images. Information needed to make use of this data will be recorded in data dictionaries that will be accessible to the research team and will subsequently be shared alongside the final datasets.

Example 3:

In compliance with NOT-MH-20-067, the following common data elements will be collected to facilitate aggregation of this data set with other data sets:

1. Age (ndar_subject01)
2. Sex at Birth (ndar_subject01)
3. DSM Crosscutting (dsm5crossa0)
4. WHODAS 2.0 (whodas201)
5. PHQ-9 (phq901)
6. GAD-7 (cde_gad701)
   

As described in the Research Plan, the additional phenotypic and clinical information will be collected using the following data dictionaries obtained from NDA:

1. Genomics Subject (genomics_subject02)
2. Genomics Sample (genomics_sample03)
3. Structured Clinical Interview for DSM-V (scidv_dep01)
4. MATRICS Consensus Cognitive Battery (matrics01)
   

The sequence data will be stored in standard formats FASTQ, SAM/BAM, BED, and VCF. Those data files will all be deposited into NDA.  The description of the genomics experiment will be submitted using the NDA genomics_sample03 data structure.  Additional experimental protocols will be described in NDA Experiments associated with our NDA Collection.

Example 4:

The data that will be used for some of the proposed secondary data analysis is already in NDA and is formatted using NDA data dictionaries.  The new data we will deposit will also use existing NDA data dictionaries.  Since the data set to be deposited into NDA was collected prior to the publication of NOT-MH-20-067, not all of the common data elements expected by NIMH are available.  However, we will transform some existing demographic and clinical data into the formats expected for:

1. Age (ndar_subject01)
2. Sex at Birth (ndar_subject01)
3. Patient Health Questionnaire-9 (PHQ-9, cde_phq901 NDA data dictionary).
   

In addition, information from the Beck Depression Inventory will be deposited for all 155 research participants using the NDA bdi01 data dictionary.  Deposited images will use the NDA image03 data dictionary.  Data derived from the MRI images will be deposited into NDA using the imagingcollection01 data dictionary.

Example 5:

We will use the standards that are adopted or defined by NeMO.

A. Repository where scientific data and metadata will be archived:

Example 1:

All data will be deposited to NDA starting 12 months after the award begins and will be deposited every six months thereafter following the usual NDA data submission dates.

Example 2:

The data and metadata arising from the project are going to be stored at the NeuroImaging Tools and Resources Collaboratory (NITRC).

If NIH repositories or other free repositories cannot be used to store the sharable datasets, researchers must plan on their budgets an amount of $0.023 per GB per month (≈ $300 per TB per year) to share their datasets. The prices to store 1 GB at Amazon Web Services are available here.

Example 3:

All data will be deposited to NDA starting 12 months after the award begins and will be deposited every six months thereafter following the usual NDA data submission dates.

Example 4:

All previously unshared data will be deposited to NDA no later than 12 months after the award begins.

Example 5:

Mouse single-cell datasets: All single-cell epigenomics and transcriptomics data will made available through NeMO after initial data processing. 

Human single cell data will be deposited to NeMO.

Upon publication we will host processed data matrixes and associated metadata as compressed downloadable archives at NeMO and, when appropriate, as supplementary information in journal publications.

We will release datasets associated with the technological advances proposed in the application once protocols are established and initial analysis performed, at which point data will be released along with a preprint prior to manuscript submission. 

B. How scientific data will be findable and identifiable:

Example 1:

Data will be findable for the research community through the NDA Collection that will be established when this application is funded.  For all publications, an NDA study will be created.  Each of those studies is assigned a digital object identifier (DOI).  This data DOI will be referenced in the publication to allow the research community easy access to the exact data used in the publication.

Example 2:

The 256-channel EEG from approximately 500 subjects will be available at the NeuroImaging Tools and Resources Collaboratory (NITRC). The NITRC provides metadata, persistent identifiers, andlong-term access. This repository is supported by NIH and the dataset(s) are available under an open-source license.

Example 3:

Data will be findable for the research community through the NDA collection that will be established when this application is funded.  In addition, the dbGaP study, which will point to NDA, will help researchers find the data. For all publications, an NDA study will be created.  Each of those studies is assigned a digital object identifier (DOI).  This data DOI will be referenced in the publication to allow the research community easy access to the exact data used in the publication.

Example 4:

Data will be findable for the research community through the NDA collection that will be established when this application is funded.  For all publications, an NDA study will be created, and the data relevant to that publication will be shared immediately.  Each of those studies is assigned a digital object identifier (DOI).  This data DOI will be referenced in the publication to allow the research community easy access to the exact data used in the publication.

Example 5:

Data will be findable for the research community through searches at NeMO.  NeMO assigns unique identifiers for each sample.

C. When and how long the scientific data will be made available:

Example 1:

The research community will have access to data when the award ends.  As required by NDA, studies will also be created that contain the data used for every publication.  Those studies will be shared when the pre-print is available.  NDA studies have digital object identifiers (DOI) to aid in findability.  We will include that DOI in relevant publications.  NDA will make decisions about how long to preserve the data, but that data archive has not deleted any deposited data up to now.

Example 2:

The scientific data will be made available after the end of the project for a period of 3 years.

Example 3:

The research community will have access to data at the end of the grant award or when a publication has been submitted. Once the data are submitted to NDA, that archive will control the long-term persistence of the data set.  Currently, NDA has no process for deleting or retiring data sets.

Example 4:

The research community will have access to the previously unshared data at the end of the grant award.  Researchers will request data using the standard processes at NDA, and the NDA data access committee will decide which requests to grant.  The standard NDA data access process allows access for one year and is renewable.  Once the data are submitted to NDA, that archive will control the long-term persistence of the data set.

Example 5:

The research community will have access to data as soon as NeMO is able to release it.  NeMO will control the deletion of the data sets.

A. Factors affecting subsequent access, distribution, or reuse of scientific data:

Example 1:

All research participants will be consented for broad data sharing.

Example 2:

• Informed consent
• Privacy and confidentiality protections consistent with applicable federal, Tribal, state, and local laws, regulations, and policies
• Whether access to scientific data derived from humans will be controlled
• Any restrictions imposed by federal, Tribal, or state laws, regulations, or policies, or existing or anticipated agreements

Example 3:

All research participants will be consented for broad data sharing.

Example 4:

The two existing data sets from NDA used consents that allow broad data sharing.  The new dataset to be uploaded to NDA also was collected using informed consent terms that allow broad data sharing.  Access to data housed by the NDA requires the completion of a Data Use Certification (https://nda.nih.gov/faq.html#dac.3), which prohibits any redistribution or attempts to re-identify research participants.

Example 5:

There are no special considerations related to accessing or distributing the mouse data to be generated in this award.  

B. Whether access to scientific data will be controlled:

Example 1:

To request access of the data, researchers will use the standard processes at NDA, and the NDA Data Access Committee will decide which requests to grant.  The standard NDA data access process allows access for one year and is renewable.

Example 2:

User registration is required in order to access or download the 256-channel EEG data files. As part of the registration process, users must agree to the conditions of use governing access to the public release data, including restrictions against attempting to identify study participants, destruction of the data after analyses are completed, reporting responsibilities, restrictions on redistribution of the data to third parties, and proper acknowledgment of the data resource. The information provided to users will not be used for commercial purposes, and will not be redistributed to third parties.

Example 3:

To request access of the data, researchers will use the standard processes at NDA, and the NDA Data Access Committee will decide which requests to grant.  The standard NDA data access process allows access for one year and is renewable.

Example 4:

To request access of the data, researchers will use the standard processes at NDA, and the NDA Data Access Committee will decide which requests to grant.  The standard NDA data access process allows access for one year and is renewable.  Once the data are submitted to NDA, that archive will control the long-term persistence of the data set.  Currently, NDA has no process for deleting or retiring data sets.

Example 5:

Access to the human data sets in NeMO is controlled. The NIMH Data Archive Data Access Committee serves as the data access committee for NeMO.

C. Protections for privacy, rights, and confidentiality of human research participants:

Example 1:

The NDA GUID tool allows researchers to aggregate data from the same research participant without different laboratories having to share personally identifiable information about that research participant.  The NDA data dictionaries do not permit personally identifiable information to be shared.  NDA maintains a Certificate of Confidentiality.

Example 2:

In order to ensure participant consent for data sharing, IRB paperwork and informed consent documents will include language describing plans for data management and sharing data, describing the motivation for sharing, and explaining that personal identifying information will be removed.

To protect participant privacy and confidentiality, shared data will be de-identified using the “Safe Harbor” method provided in Section 164.514 (b) of the HIPAA Privacy Rule. The safe harbor method under the HIPAA Privacy Rule de-identification standard requires covered entities or business associates to remove all 18 identifiers of PHI from data in order to ensure that the data cannot be traced back to one person.

Example 3:

The NDA GUID tool allows researchers to aggregate data from the same research participant without different laboratories having to share personally identifiable information about that research participant.  The NDA data dictionaries do not permit personally identifiable information to be shared.  NDA maintains a Certificate of Confidentiality.

Example 4:

The NDA GUID tool allows researchers to aggregate data from the same research participant without different laboratories having to share personally identifiable information about that research participant.  The NDA data dictionaries do not permit personally identifiable information to be shared.  NDA maintains a Certificate of Confidentiality.

For the 155 participants from our previous study, exact dates have been obscured via the Shift and Truncate method [1], which preserves within-case temporal relations.

[1] Hripcsak, G., Mirhaji, P., Low, A. F., & Malin, B. A. (2016). Preserving temporal relations in clinical data while maintaining privacy. Journal of the American Medical Informatics Association, 23(6), 1040-1045.

Example 5:

The data are deidentified, and there is data access committee (see 5B) that is used by NeMO to evaluate the proposed use of the sensitive data from human subjects.  The data archives which will hold the data are funded by NIH, so they have a Certificate of Confidentiality.

Example 1:

Compliance with the DMSP will be realized via multiple means. First, the investigators, with institutional oversight and support via the Research Informatics Office (RIO) and Office of Regulatory Compliance, will review and update the DMSP as necessary in concordance with NIH/reviewer feedback and/or relevant changes dictated by scientific need and/or available institutional resources. Second, annual review and update will be performed as part of the RPPR with the grants manager for the unit ensuring sign-off on the DMSP. Third, where institutional resources are utilized, the RIO will perform audits for access control and sharing settings according to the specifications for (X) resource. [Any data sharing request for human subject data access will be reviewed prospectively by the IRB to ensure compliance with human subject and privacy requirements before being provided.] Finally, as part of grants management and institutional compliance, we will deploy a process whereby the investigator can log any data shared in public repositories as part of their annual RPPR process and in closing out awards, when appropriate.

The University of Colorado Anschutz has several institutional resources that can be used to store and share data. Please see a short description and links to the resources available. 

O365 OneDrive – OneDrive for Business is an integral part of Microsoft 365 and provides a place in the cloud where it is possible to store, share, and sync highly confidential, confidential, and public research data. OneDrive for Business is configured for HIPAA compliance, so files are all private by default. All Faculty and Staff members of the University of Colorado can get a 5 TB OneDrive account for free.

Dell Isilon – Dell Isilon is an On-premises data server where it is possible to store, highly confidential, confidential, and public research data. Dell Isilon is configured for HIPAA compliance, so files are all private by default. The cost to store 1 GB per month is $0.025.

Alpine – Alpine is an On-premises high performance computing (HPC) cluster. Alpine can be used to store and process confidential and public research data. Currently Alpine is not HIPAA compliant. There are no costs associated to use Alpine.

PetaLibrary – PetaLibrary is an On-premises data server that supports the storage, archival, and sharing of confidential and public research data. Currently Alpine is not HIPAA compliant. There is a cost of $50 per TB store in PetaLibrary.

SLCE – The Secure Local Compute Environment (SLCE) is a high performance computing (HPC) cluster designed to store and process highly confidential, confidential, and public research data. SLCE is HIPAA and NIST 800-171 compliant. For more information about SLCE including costs associated, please contact Dalan.Jensen@CUAnschutz.edu .

Egnyte – Egnyte is a cloud-based data storage where it is possible to store and share highly confidential, confidential, and public research data. Egnyte is configured for HIPAA compliance, so files are all private by default. There is a cost of $150 per user per year to use Egnyte.

REDCap - REDCap (Research Electronic Data Capture) is a secure, HIPAA-compliant web-based application designed for highly confidential, confidential, and public data collection for research studies. REDCap is free of charge.

OnCore – OnCore is a cloud base platform that enables users to manage highly confidential research data. OnCore is configured for HIPAA compliance, so files are all private by default. There is no cost associated to use OnCore.

Eureka - Eureka is a cloud base high performance computing (HPC) cluster that can be used to store and process highly confidential, confidential, and public research data. Eureka is HIPAA, NIST 800-171, and GDPR compliant. The cost to use Eureka can be calculated here.

Virtual Server Hosting – Virtual Server Hosting is provided by OIT as an alternative to physical servers. Highly confidential, confidential, and public research data can be hosted in virtual servers provided by OIT. Virtual Server Hosting is configured for HIPAA compliance, so files are all private by default. The cost to use a Virtual Server can be calculated here.

NIH recognizes that making data accessible and reusable for other researchers may incur costs. For that reason, investigators may request funds toward data management and sharing in the budget and budget justification sections of their applications.

Reasonable, allowable costs may be included in NIH budget requests for:

• Curating data
• Developing supporting documentation
• Formatting data according to accepted community standards, or for transmission to and storage at a selected repository for long-term preservation and access
• De-identifying data
• Preparing metadata to foster discoverability, interpretation, and reuse
• Local data management considerations, such as unique and specialized information infrastructure necessary to provide local management and preservation (for example, before deposit into an established repository).
• Preserving and sharing data through established repositories, such as data deposit fees.

Budget requests must NOT include:

• Infrastructure costs that are included in institutional overhead (for instance, Facilities and Administrative costs) 
• Costs associated with the routine conduct of research, including costs associated with collecting or gaining access to research data. 
• Costs that are double charged or inconsistently charged as both direct and indirect costs 

To request funds toward DMS costs, investigators should include:

1. A line item in the budget form
2. A brief summary of the DMS Plan and a description of the requested DMS costs in the budget justification

More information about the DMSP budgeting can be found here.

Effective for applications submitted for due dates on or after October 5, 2023, NIH will no longer require the use of the single DMS cost line item. Find additional information about these changes here.

For more information on How to Apply, click here. 

Plans must address oversight of data management and sharing, describing how compliance with the plan will be monitored and managed. During the funding period, compliance with the DMS plan will be reviewed and determined by the NIH ICO.

After the end of the funding period, non-compliance with the NIH ICO-approved Plan may be considered by NIH for future funding decisions for the recipient institution.

The CRIO Council created a Working Group (WG) composed of faculty and staff members from different CU Anschutz Schools and from the Strauss Health Sciences Library. The WG has been working intensively to develop DMP templates and other working materials that will help the Anschutz Research Community to adopt the new NIH Data Management and Sharing Policy.

See these resources for more information:

2023 NIH Data Management and Sharing Policy Guide

NIH Data Management and Sharing Policy Webinar Series

Elements of an NIH Data Management and Sharing Plan (NOT-OD-21-014)

Implementation Changes for Genomic Data Sharing Plans (NOT-OD-22-198)

Allowable Costs for Data Management and Sharing (NOT-OD-21-015)

Selecting a Repository for Data Resulting from NIH-Supported Research (NOT-OD-21-016)