Studies

The data archived and distributed by dbGaP are organized as studies. Studies may be either stand-alone or combined in a “parent study/child study” hierarchy. Parent or “top level” studies may have any number of child studies (also referred to as substudies). However, study hierarchy is limited to two levels (parent and child only). In other words, substudies may not have substudies. Studies, whether parent or child, can contain all types of data ascertained in genetic, clinical or epidemiological research projects such as phenotype and molecular assay information that are linked via subject and sample IDs. Studies often contain documents, such as questionnaires and protocols, which help contextualize the phenotype and genotype data. Study data are distributed by consent groups, each of which contains all data from a set of study participants who have signed the same consent agreement. In other words, the data delivered for a single consent group will all have the same Data Use Limitations (DULs) for future research use.

Each study is assigned a unique accession number which should be used when citing the study. The general dbGaP accession format for a study is phs######.v#.p#. The first three letters [phs] denote the object type (‘s’ denotes study), followed by 6-digit, 0-padded object number (######) which is consecutively assigned by dbGaP. The version number (.v#) indicates updates of the object, where # is initially 1 and increments by 1 as the object is updated. The version number is followed by participant group (.p#) where # is initially 1 and increments by 1 as the participant group changes. The version number of a study will increment any time the version of an object contained by the study (such as a phenotype variable, genotype data, or a sub study) is updated. The participant group of a study will change when existing subjects are removed, or when an existing subject changes from one consent group to another, but not when additional subjects are added.

While the data found in studies can vary widely based on both the number of participants and the variety of deposited phenotypic and molecular data, all studies include basic descriptive metadata such as study title, study description, inclusion/exclusion criteria, study history, disease terms, publications related to the study, names and affiliations of the principal investigators, and sources of funding. This information is publicly available on the study’s report page at the dbGaP website.

Datasets and Variables

Phenotypic data values are submitted to dbGaP as tabular files or datasets (accessioned with pht#, where ‘t’ denotes table), where columns represent phenotypic variables (accessioned with a phv#, where ‘v’ denotes variable) and rows represent subjects. A dbGaP phenotype variable consists of two parts: the data values and the description of the data in the accompanying data dictionary. Each cell (value) in a dataset is stored in a relational database and is mapped to the appropriate phenotype variable and subject. Phenotype variable metadata are provided by the submitter via a data dictionary for each dataset and include: variable name, variable description, units, and a list of any coded responses. The variable’s data type (text string, integer, decimal or date) is automatically determined by calculating which type is in the majority. Conflicts between submitted and calculated data types, or other discrepancies in the data, are reconciled by dbGaP curators in consultation with the data submitter.

Variables are created from the columns of the dataset; each variable and dataset is accessioned using the general dbGaP format ph(v|t)######.v#.p#. A variable’s version (v#) will change when either values of data change or its entry in the data dictionary changes. A dataset’s version will change when a variable inside the dataset is added, updated or deleted. For both variables and datasets the participant set (p#), is inherited from the study to which it belongs. Variables, and sometimes datasets, are linked to appropriate sections of documents (please see the Website section below for details).

Individual level phenotype data is only available through the dbGaP Authorized Access System. Public summary-level variable information is available on the dbGaP website and ftp site.

Genotype data

Genotype data hosted at the dbGaP consist of individual level genotypes and aggregated summaries, both of which are distributed through the dbGaP Authorized Access System. The types of data available include DNA variations, SNP assay, DNA methylation (epigenomics), copy number variation, as well as genomic/exomic sequencing. RNA data types such as expression array, RNA seq, and eQTL results are also available. For details about the accepted format of submitted genotype files please see the dbGaP submission guide.

Genotype data are accessioned based on their data type and use the general dbGaP accession format ph(g|e|a)######.v# where ‘g’ denotes GWAS, ‘e’ expression, and ‘a’ analysis. Versioning of genotype data is triggered by addition or withdrawal of samples, sample consent status change, or error correction.

Genotype data files are compressed and archived into tar files for distribution. The files are explicitly named to indicate file content, such as image data (cel and idat), genotype calls (genotype), and locus annotations (marker info). Genotype calls are usually clustered according to file format and genotyping platform, including one sample per file (indfmt), multiple-sample matrix (matrixfmt) and pre-defined variant call format (vcf). They will be accompanied by sample-info file for subject lookup and consent status. The consent code and consent abbreviation are also embedded in the file name.

Examples of genotype data file names:

phe000005.v1.FHS_SHARe_project4_miRNA.sample-info.MULTI.tar

phg000006.v6.FHS_SHARe_Affy500K.genotype-calls-matrixfmt.c2.HMB-NPU-MDS-IRB.tar

The various pieces of the names can be parsed to extract meaningful content: the genotype accession (phe000005.v1 and phg000006.v6); the study (FHS_SHARe in both cases); the molecule type (miRNA); the platform/chip information (Affy500K); the content type (sample-info or genotype-calls-matrixfmt); the consent code (c2); and the consent abbreviation (HMB-NPU-MDS-IRB).

Analyses

Because of the large volume of data generated and concerns regarding participant confidentiality many genetic epidemiological analyses have not been published. But, because individual-level data is only accessed through Controlled Access, dbGaP can archive, integrate and distribute these results.

Analyses can either be provided by submitters or be pre-computed by dbGaP staff, though pre-computes account for a small number of the total analyses. Submitted analysis results are accessioned with the prefix “pha”. After removing identifiable elements, like counts and frequencies, analysis results are displayed in the public dbGaP browser that dynamically links to NCBI annotation resources, like dbSNP, Gene, RefSeq. These public views can be found through the “Analysis” link on the study page and they can be downloaded from the FTP site. The original submitted analyses, including updated marker info, are fully accessible through dbGaP Controlled Access.

Analysis files are typically formatted by population, trait and analysis method (typical), however some include surveys across multiple populations, and may use either SNPs or genes as the loci analyzed. However, in general, they all contain the following three parts which are also required for dbGaP submission.

1.

Metadata, which includes trait, population, sample size and brief descriptions on Analysis and Method.

2.

Marker information and genotyping summary, such as identifiers of loci (variation, gene and structure variant), alleles, genotype counts (frequency), call rate and p-value from Hardy-Weinberg-equilibrium testing.

3.

Testing statistics, including p-value, effect size (odds ratio/regression coefficient/relative risk) and direction (coding allele) if association results.

With these resources, other scientists can verify the discoveries, recalculate statistics under various genetic models, develop new hypotheses, and more importantly, construct a meta-analysis even though individual–level data are inaccessible. The details of data fields are listed in dbGaP submission guides and we welcome suggestions and comments from the scientific community. An interactive view of the analsysis results submitted to dbGaP is described in The dbGaP Genome Browser in the Related Tools section of this chapter.

Documents

The dbGaP encourages investigators to submit documents related to their studies, such as protocols, patient questionnaires, survey instruments and consent forms, along with their data. These documents provide valuable information and context for subsequent researchers who will apply for and download datasets. All submitted documents are available publicly and can be used by anyone interested in gaining a better understanding of the phenotypic data found in a study.

Each document is accessioned using the general dbGaP format phd######.v# where “d” indicates document. A document’s version (v#) will change when the variables annotated on the document change, or when the document itself is changed significantly. (For example, fixing a typo would not be considered a significant change unless it were to change the meaning of the document.)

Documents submitted to dbGaP are represented in a common XML format. Converting documents into a common format allows all documents to be treated uniformly in the database (aiding indexing and discovery) and to be displayed in a single HTML style. Additionally, the XML format allows curated information to be added to the documents. This curated information is used to create live links between the documents and other portions of the dbGaP website, such as variable report pages. Linking between documents and other objects will be discussed further in the section about the dbGaP website.

Documents are generally viewable on the dbGaP website in both HTML and PDF format (the PDF for a document may be the originally submitted object, if it was sent as PDF, or could be a PDF representation of another format such as Microsoft Word or Excel or a plain text file).

The XML used by dbGaP is an extension of NLM’s Archiving and Interchange Tag Set Version 2.3 (http://dtd.nlm.nih.gov/archiving/2.3/). The extension adds structures to code questionnaires and adds a number dbGaP-specific attributes to common document structures (such as sections, tables, and lists) to facilitate curation. A copy of our extension is publicly available at http://dtd.nlm.nih.gov/gap/2.0/wga-study2.dtd, and documentation for the extension is located at http://dtd.nlm.nih.gov/gap/2.0/doc/wga-document.html.

NIH-Funded Studies

Institutional certification, as well as basic information about the study, is required when submitting data to dbGaP.

• Institutional certification consists of a letter signed by the principal investigator and an institutional official that confirms permission to submit data to dbGaP. NIH has developed Points to Consider for IRBs and Institutions to assist institutions in their review and certification of an investigator’s plan for submission of data to dbGaP.
• Basic information consists of items like the title of the study, a description and history of the study, inclusion and exclusion criteria, listing of preview and certification of PI’s data
• rincipal investigators (PIs) and funding information.

Principal investigators (PIs) should familiarize themselves with the “NIH Points to Consider” document that provides information about: the NIH GWAS Data Sharing Policy; benefits of broad sharing of data through a central data repository; risks associated with the submission and subsequent sharing of such data; safeguards designed to protect the confidentiality of research participants; and specific points for institutional review boards (IRBs) to consider during review and certification of PIs’ data submission plans.

The principal investigators must contact their NIH program official (PO) to begin the submission process. If the study was not funded by the NIH, PIs should contact vog.hin.mln.ibcn@pleh-PaGbd for guidance.

Non-NIH-Funded Studies

To submit non-NIH-funded data to dbGaP the following information will need to be provided:

• Institutional certification as described in the last section. To provide this, someone from the institution or organization will need to be registered in eRA Commons. Information regarding registration is available from the eRA Commons website. (Note: The review of a PI’s request can be initiated without the certification, but the review process will be expedited if GWAS staff receives the certification at time of submission.)
• Basic information about the study, as described in the previous section.
• The NIH IC that most closely aligns with the research. A list of ICs can be found at http://www.nih.gov/icd/.
• Whether the study has been published or accepted for publication. If it has the PI should provide documentation (i.e., the publication citation or a copy of any correspondence indicating that an article about the study has been accepted for publication).

The PI should submit all information and the certification to vog.hin.liam@SAWG. Once GWAS staff receives the documents, they will forward them to the appropriate IC program administrator for consideration. The IC program administrator will contact the PI with any questions and/or to notify you of the IC’s decision.

The PI is encouraged to consult with the Program Officer/Director (PO/PD) and/or IC GWAS Program Administrator (GPA) at an NIH Institute or Center (IC) to discuss the project, data sharing plan, and data certification process (non NIH funded projects should contact pleH PaGbd) to complete the registration process.

Instructions for submitters

Phenotype processing

The phenotypic data are subjected to both automated and human-mediated assessment.

Before the data are loaded into the database, scripts are used to evaluate the following:

• Poor formatting - Each dataset submitted to dbGaP should be a rectangular table showing variables in columns and subject or sample IDs in rows.
• HIPAA violations - The datasets submitted to dbGaP should follow the Health Insurance Portability and Accountability Act (HIPAA) rules in order to protect the privacy of personally identifiable health information.
• Issues with Subject and Sample IDs - Submitters are required to use the subject consent file to list all the subjects who participated in, or are referred to by, the study, with their consent values. Subjects who had not directly participated in the study, e.g., parents of participants included in the pedigree file, should also be included in the consent file with consent value 0.
• Missing information in Data Dictionaries - The submitters are required to submit data dictionaries, in addition to datasets, to explain the meanings of the variables and data values. For each value in the datasets, dbGaP requires that the submitters provide a variable description, variable type, units of values for numerical variables, and logical minimum and maximum values if available. For each encoded value, a code meaning should be included in the data dictionary.
• Issues with Pedigree files- A pedigree file submitted to dbGaP should include the following columns: family ID, subject ID, father ID, mother ID, sex, and twin ID if available. We also require that all the subjects who appear in father ID or mother ID columns also be included in the subject ID column.

More detailed information about the particular automated testing performed to catch errors in the broad categories listed above can be found in Appendix – Phenotype Quality Control.

Reports from the automated quality control (QC) scripts are reviewed by curatorial staff. If necessary, curators will communicate with the submitters and ask that new files be submitted correcting the errors. Even if the automated QC checks do not detect problems, the curatorial staff check all data dictionaries manually to see if there are any problems that were not identified by automated checks.

Genotype processing

The genotype data processing and QC process consists of the following steps:

1.

Check for availability of Sample-Subject Mapping file and validate against subject list.

2.

Check for availability of sample genotype file and validate against SSM.

3.

Process and genotype file and generate PLINK (3) set.

4.

Check for data consistency in the submitted QC component against data from other submissions for the study.

5.

Conduct QC checks and generate genotype-QC component. This step includes checking:

a.

Missing call rates per sample/per marker.

b.

Minor allele frequency.

c.

Mendelian error rate when trios are available.

d.

Duplicate concordance check (Generate SNP and subject filters based on tests results

e.

when dups are available).

f.

Gender check.

g.

IBD analysis.

6.

Generate SNP and subject filters based on tests results.

7.

Verify genotype and phenotype data using NCBI GWAS pre-compute against similar pre-compute provided by PI/analysis group.

8.

Split PLINK sets according to consent and generate genotype-calls-mtrxfm components.

9.

Generate sample-info and marker-info release components.

10.

Partition and pack build of individual genotype files according to subject consent information and data type.

The quality of the genotype data is checked at both the genotype data file and the genotyping level. Typically at the file level a genotype release contains individual level data in both individual (one file per sample) and matrix (one matrix with all samples) formats. The genotype matrices are generated by dbGaP curators from submitted individual genotype files and subject related information, such as gender and pedigree data. These matrices then are used to generate pre-computes/metrics/QC-filters, which are further verified against similar pre-computes submitted by the investigators. When necessary, the genotyping quality of each sample is also verified using a B-Allele frequency (BAF) analysis pipeline which calculates and processes BAF values to identify samples with extremely “noisy” or failed genotyping.

The following quality assurance steps are implemented to facilitate cross-study and cross-technology data merging and analysis:

1.

Identify duplicated genotypes across all studies as well as generating data.

2.

Check data formats, annotation, and QC- metrics for genotype data derived using different technologies.

3.

Check ID reconciliation, gender, missing call rate, duplicate concordance, identity-by-descent, and Mendelian error rate at sample as well as SNP level.

Subjects and Samples

In dbGaP there are two similar yet distinct concepts that describe the participants in a study: subject and sample. A subject corresponds to an individual human. A sample in dbGaP corresponds to each analyte (DNA/RNA) that is put in the machine or on the chip, rather than the physical result of obtaining a tissue or blood sample, though this is accepted as well. Modeling samples this way allows dbGaP to track duplicates, centers, plates, wells, and any sequence of aliquots that precedes the actual aliquot used to produce the molecular data finally submitted to dbGaP. The tracking method also enables dbGaP to easily add, rename, or redact samples over time.

Submitters are required to assign de-identified IDs to subjects and samples; these are submitted subject and sample ids. However, dbGaP also assigns a unique id to samples and subjects; these are the dbGaP Subject and Sample IDs and are included in the final phenotype files available through controlled access. The dbGaP assigns its own IDs to accurately represent cases where a single subject (person) has participated in more than one study. In such a case the two submitted subjects will be assigned the same dbGaP Subject ID. This can only be done if the submitter provides the information that a subject in their study is the same subject as in an existing dbGaP study. Similarly, cell repository, or otherwise readily available samples such as Coriell samples, used as controls in multiple studies will typically receive the same dbGaP Sample ID.

All phenotype and molecular data are connected though the Subject Sample Mapping file.

Data ID mapping

The data submitted to the dbGaP are de-identified. The phenotype and genotype data are connected through the subject sample mapping file in which one sample is mapped to exactly one subject and one subject is mapped to any number of samples. The following is a partial list of the IDs and attributes included in dbGaP phenotype and molecular data files.

1.

SUBJECT_ID: This is the submitted Subject ID and it is included in the Subject Consent Data File, the Subject Sample Mapping Data File, the Pedigree Data File (if applicable), and all Subject Phenotype Data Files. SUBJECT_ID should be an integer or string value consisting of the following characters: English letters, Arabic numerals, period (.), hyphen (-), underscore (_), at symbol (@), and the pound sign (#). In addition to the submitted Subject ID, dbGaP will assign a dbGaP Subject ID that will be included in the final phenotype dump files along with the submitted Subject ID.

2.

SAMPLE_ID: This is the submitted Sample ID and is included in the Subject Sample Mapping Data File and the Sample Attributes Data File. This ID should be used as the key for the individual level molecular data. Each sample should be submitted with a single, unique, de-identified Sample ID. The acceptable characters in Sample IDs are the same as those in the Subject IDs. In addition to the submitted Sample ID, dbGaP will assign a dbGaP Sample ID that will be included in the final phenotype dump files along with the submitted Sample ID. The SAMPLE_IDs listed in the Subject Sample Mapping Data File should be identical to the samples found in the genotype, SRA, and other molecular data.

3.

dbGaP_SAMPLE_ID: This is the dbGaP assigned unique identifier assigned to the submitted Sample ID. The dbGaP Sample ID is included as a column in the final phenotype dump files whenever there is a submitted sample ID column.

4.

dbGaP_SUBJECT_ID: This is the dbGaP unique identifier assigned to the submitted Subject ID. The dbGaP Subject ID is included as a column in the final phenotype dump files whenever there is a submitted subject ID column. The dbGaP Subject ID is unique cross all dbGaP studies, which means that if a subject is known to have participated in multiple studies that have been submitted to dbGaP, the same dbGaP Subject ID will be assigned to the individual across multiple studies, though the submitted subject ID may be different.

5.

SOURCE_SUBJECT_ID and SUBJECT_SOURCE: The Source Subject ID (SOURCE_SUBJECT_ID) is the de-identified alias Subject ID used in the public repository, consortium, institute, or study from where the subject has been obtained. The Subject Source (SUBJECT_SOURCE) is the name of the third party source, public repository, consortium, institute, or study that corresponds to the subject. For subjects originating from a shared source (such as a public repository, consortium, institute, study, etc.) or for subjects with alias IDs, these 2 variables will be included in the Subject Consent Data File. The SOURCE_SUBJECT_ID maps to the SUBJECT_ID. For referencing HapMap subjects from Coriell, the SUBJECT_SOURCE value is written as “Coriell.” The SOURCE_SUBJECT_ID should be written as the de-identified subject ID assigned by Coriell (e.g., NA12711).

6.

FAMILY_ID: The Family ID is a column of de-identified Family IDs in the pedigree file. The Family ID is also referred to as the Pedigree ID. The family ID should be the same for individuals belonging in the same biological family. The family ID is found in the pedigree file if a pedigree file is available.

7.

SEX: The gender variable can be included in a subject phenotype data file or in a pedigree file if a pedigree file is available.

8.

FATHER and MOTHER: In the pedigree file, FATHER and MOTHER are the two columns of the unique, de-identified subject IDs of the participant’s biological father and mother. The Father ID and Mother ID may not be identical. 0 (zero) or blank is filled in for founders or marry-ins (parents not specified) in a pedigree. Each unique Father ID and unique Mother ID is also listed in the Subject ID column of both the Pedigree Data File and the Subject Consent Data File.

9.

TWIN_ID: Monozygotic twins or multiples of the same family have Twin IDs. Twins or multiples of the same family share the same TWINID, but are assigned different SUBJECT_IDs.

10.

CONSENT: Every subject that appears in a Subject Phenotype Data File must belong to a single consent group and every sample that appears in a Subject Sample Mapping File and in a Sample Attribute Data File must belong to a consented subject. The consent information is listed in the Subject Consent Data File. Consents are determined by the submitter, their IRB, and their GPA (GWAS Program Administrator) along with the DAC (Data Access Committee). All data is parsed into its respective consent groups for download.

Curatorial document annotation

One thing that sets dbGaP apart from similar databases is the extent of curatorial work done with the data and documentation we receive. For documents, this involves making connections between appropriate portions of text and other accessioned objects (such as variables, data tables, and other documents) and creating links to external resources. We refer to this process as “document annotation” and it involves embedding references into the XML for the documents.

Documents can either be annotated by the submitter or by the dbGaP curator responsible for a study. Types of annotations include adding variable IDs to particular sections of text so that the text can be linked to the variable report page or adding references so that hyperlinks can be made between chapters of a protocol document.

dbGaP Website

Searching dbGaP

All publicly released dbGaP studies can be queried from the search box on the top of the dbGaP homepage. Queries can be very simple, just keywords of interest (“cancer”), or complex, making use of search fields and Boolean operators (“cholesterol[variable] AND phs000001”). More complex searches can be facilitated by using the “Advanced Search” which helps create queries via a web form.

There are many search fields available in dbGaP. Table 1 shows a selection of the most commonly used fields, explains what they search for, and gives an example of how the search would be formed.

Table 1.

This table lists fields in the dbGaP advanced search that are likely to be useful to most searchers.

Additionally, complex queries can also contain Boolean operators. For example:

Cancer[Disease] AND True[Study Has SRA Components]

returns a list of all studies having SRA data and where the PI has assigned the keyword “cancer” as a disease term.

As with all other NCBI resources, the searches in dbGaP are performed using the Entrez search and retrieval system. Please see the Entrez chapter of the NCBI handbook for general guidance on forming Entrez queries.

Once a search query is executed and results returned (Figure 3), clicking on an item’s name or accession will lead to a page listing more specific information about that object. This information is of particular importance to those users who want to find out more about a study before deciding whether or not to apply for Authorized Access. (Note that on each of the different pages, one can examine other objects in the study by using the navigational aid along the right-hand edge of the page.)

Figure 3.

This shows the returns for a simple search on the word "diabetes". Note that specific results for Studies, Variables, Study Documents, Analyses, and Datasets can be accessed by choosing the appropriate tab.

Variables on the public website

Phenotype variables can either be found by doing a search on the dbGaP home page, and then linking to an individual variable page (see Figure 4 and Figure 5 for examples), or they can be found by choosing the “Variables” tab if you are already looking at the website of a study. If you are using the “Variables” tab the phenotypes are generally grouped into broad categories for ease of browsing. These categories can be found to the right-hand side of any variable report web page (see Figure 4). Most studies use the following categories, as appropriate:

Figure 4.

Top portion of the variable report for the variable phv000200829, CARDIOV. Variables can be browsed by category using the navigation to the right hand side of the page.

Figure 5.

Top portion of the variable report for the variable phv000200829, CARDIOV. This shows how variables are linked to appropriate sections of documents. If you follow the first link, you will be taken to the portion of the document show in Figure 6.

• Affection Status
• Sociodemography and Administration
• Medical History
• Physical Observations
• Lab Measurements
• Psychological and Psychiatric Observations
• Lifestyle and Environment
• Treatment

Exceptions to this grouping method are found in large studies such as the Framingham Cohort which have their own long-standing system for grouping data. When searching for variables in large studies, or if you have a very specific query, it can be more efficient to search for variables using the search box on the right side of the variable report page (or from the dbGaP home page if you want to perform a cross-study search), rather than attempting to browse through the hierarchy of folders.

Documents on the public website

There are multiple pathways to find documents through the dbGaP web site. On the dbGaP home page the newest studies are listed under the “Latest Studies” heading, with the most direct route to documents being the orange "D" icons. A gray icon means there are no documents associated with the study. Beneath that section, the "List Top Level Studies" link leads to a searchable listing of all studies and documents, with an advanced search option available for building document-specific queries. On a study page, clicking the Documents tab will open the study's default document, with a folder tree on the right to explore the rest, and a "Search Within This Study" box that will search document text (Figure 5). Variable pages may also link to documents in which they are annotated, through the "See document part in context" links. Documents for each study are also available on the dbGaP ftp site as a downloadable zip file, which includes the pdfs, xml, and images. The ftp site is accessible from study pages by clicking the link under "Publicly Available Data."

Using document annotation

As noted previously, curators establish connections between appropriate portions of text and other accessioned objects. As an example of the types of functionality that annotations can provide, imagine looking at a variable report page having to do with whether subjects take a multivitamin. If you scroll down to the bottom of the variable page, there is a section labeled “Document parts related to the variable” which is shown in Figure 5.

This shows that there are two documents, a Coding Manual and an Annotated Form, that have text which has been associated to the multivitamin variable. If you click on the “See document part in context” link you will be taken to the appropriate portion of the document. If you follow the link for the Annotated form, you would get taken to a page that looks like Figure 6.

Figure 6.

This show the portion of the web page for the document “GoKinD Study Diabetic Offspring” (phd000152.2) which you would be taken to if you clicked the first link from Figure 5.

In the image of the questionnaire shown in Figure 6, the icon of a red circle with a white question mark, , indicates that a section of the document is associated with one or more accessioned objects in a study. The accessioned objects are generally variables, but can include data tables. Clicking on the icon will either take the user to a variable report page (if only a single variable is associated with the icon) or to an Entrez search result page (if there are multiple objects associated with that icon).

dbGaP ftp site

The ftp site includes a directory for every study, which contains a directory for every version of a study, as well as a directory where analyses are found. Currently, each version of a study contains directories for documents, phenotype variable summaries, manifests, and release notes (Figure 7). Manifests describe the files available in each consent category while release notes describe the history of the released files as well as giving details of any changes made from previous versions.

Figure 7.

This figure shows the basic organization of the ftp site, with the study phs000001.v3.p1 opened up to show portions of the hierarchy of directories and files.

Please note that older versions of studies may have a different directory structure although they contain similar information.

The variable summaries and the data dictionaries are delivered as XML files with an accompanying XSL file which produces the HTML rendering of the file that can be viewed on a browser.

The documents directory contains at least one .zip file that holds the xml files, images, and pdf versions of the documents in a study. In cases where there are a large number of documents the files may be separated into separate .zip files for xml, images, and pdf.

Scope

The Phenotype-Genotype Integrator (PheGenI), (4) merges NHGRI genome-wide association study (GWAS) catalog data with several databases including Gene, dbGaP, OMIM, GTEx and dbSNP. This phenotype-oriented resource, intended for clinicians and epidemiologists interested in following up results from GWAS, can facilitate identification and ranking of variants that may warrant additional study.

History

PheGenI was first released in 2011. The major functionality has not changed, i.e. modes of search and categories of display, but functions have been added to improve both queries and data processing. For example, an autocomplete function was added to facilitate the phenotype queries, and download functions were added to the ideogram and tabular results sections. PheGenI is under active development, with contents and displays scheduled to be more closely integrated with additional web resources.

Data Flow

PhenGenI is populated automatically via feeds from NHGRI, dbSNP, dbGaP and NCBI’s genome annotation pipeline. Please note that PheGenI does not display all p-values from each dbGaP-hosted analysis. Specifically, only p-values <10^-4, and/or the lowest 100 p-values are included for each analysis. Currently, the phenotype search terms are based on MeSH, but will be enhanced with additional options in the future.

Access

Users can search based on chromosomal location, gene, SNP, or phenotype and then view and download results. Association results can be filtered by p-value, and genotype data can be filtered by location of variant site relative to gene annotation. The results are separated into several categories, including association results, genes, SNPs, eQTL data, a dynamic genome view and dbGaP studies. Each section provides a download function.

As a tool to find data in dbGaP, the view of all analysis results is accessed by clicking on the dbGaP link in the source column of the Association Results table. For full analysis and aggregate statistics such as allele frequencies, apply for controlled access.

Gene’s Phenotypes section also provides links to PheGenI, via the anchor “Review eQTL and phenotype association data in this region using PheGenI”.