favicon.ico

U.S. Geological Survey

Data standards are the guidelines by which data are described and recorded. In order to share, exchange, combine and understand data, we must standardize the format as well as the meaning.

HEAL Data Stewards

[HEAL Semantic Search (HSS) 2.3](https://www.healdatafair.org/resources/semantic-search) is now available, bringing a refreshed interface that makes it easier to explore related studies, variables, Common Data Elements (CDEs), and biomedical concepts from a single search; compare how concepts and variables are defined and measured across studies; and access insights from 20 newly added data dictionaries, over 1000 HEAL studies, and 233 CDEs. Get started today by [entering a search term](https://www.healdatafair.org/resources/semantic-search) relevant to your research.

Explore HEAL studies, variables, CDEs, and concepts in [HEAL Semantic Search (HSS) 2.3](https://www.healdatafair.org/resources/semantic-search)!

HEAL Semantic Search Resources

This page contains resources on HEAL Semantic Search including FAQs, tutorials, and other instructional materials. HEAL investigators interested in submitting their study information should visit the "Submit Your Study Information" section.

[Click here](https://www.healdatafair.org/resources/semantic-search) to navigate to the HEAL Semantic Search interface.

### Dig Deeper into HEAL Data

Recognizing the urgency and difficulty of addressing the pain and opioid public health crises, the NIH HEAL Initiative® has funded more than 1000 research studies to speed solutions. This wealth of information provides unprecedented opportunities for new discoveries in the pain and addiction research landscape, enabling novel treatment development. However, searching for data across HEAL studies is challenging due to the diverse research methods and data types.

**HEAL Semantic Search (HSS) provides a smart way to search across HEAL studies to uncover concept, study, and variable relationships within the pain and addiction research landscape and spark knowledge discovery and novel research ideas.**

HSS can further your research by helping you:
- Discover related variables, studies, Common Data Elements (CDEs), and biomedical concepts
- Examine how a concept/variable is defined and measured in different studies
- Ensure your metadata aligns with broader research standards

This tool complements the [HEAL Platform’s](https://healdata.org/portal/discovery) free-text HEAL study and dataset search. The HEAL Platform delves deeper into the studies and variables uncovered through HSS. Together, these tools provide a detailed search experience of the HEAL Data Ecosystem.

### How does HEAL Semantic Search Work?

HSS ingests information about HEAL studies, variables, and research concepts, then utilizes advanced artificial intelligence techniques to enhance data discoverability. This allows HSS to identify relevant results even when the exact words in the query aren't used. For example, HSS understands "heart attack" means the same thing as  "myocardial infarction", and recognizes that "chronic obstructive pulmonary disease" is relevant to a search for "lung cancer."

HSS first uses natural language processing (NLP) to recognize biomedical concepts in user queries and identify synonyms, ensuring users do not need to enter an exact biomedical term. The search term and synonyms are assigned ontological identifiers. Then, HSS uses biomedical knowledge graphs with the ontological terms to find related concepts. 

What is HEAL Semantic Search?

Read our [user guide](https://docs.google.com/document/d/1MjFXg0fENFJxifEFDJ3vZGfT-aKvqBZLsQ4tDkMukl8/edit?usp=sharing) to learn more about:
- How HSS works
- How to perform a search
- How to navigate results
- Additional resources 

You can also watch our demo video for a tutorial on how to navigate HEAL Semantic Search.
- [Watch the demo video](https://www.youtube.com/watch?v=0oV4ZriaEX0)


User Guide and Demo Video

The following use cases provide insight into how HSS can elevate and expand your research. This is not an exhaustive list.

### Research Landscape Analysis

Dr. Lee is researching muscle  pain in older adults. They are preparing a HEAL research proposal and want to conduct a landscape analysis to inform study planning. Dr. Lee searches for the concept “muscle pain.” 

- Because HSS uses semantic search, which looks beyond exact keywords to include synonyms, definitions, and related concepts, it returns studies that include “myofascial pain syndrome”, “fibromylalgia” and “myalgia” as well. This expands Dr. Lee’s search to include potentially relevant results they might not get from traditional search options.

- In the **Studies tab**, Dr. Lee finds related studies and detailed information including study description, award information, and variables and standardized questionnaires used. This helps Dr. Lee understand the current state of HEAL research on muscle pain and find potential collaborators. 
 
- HSS provides links to these studies on the HEAL Platform, which contains further study and variable information, including if a study has submitted data. 

### Cohort Building for Secondary Analysis

Dr. Ray and Dr. Wang are biostatisticians looking for data on how chronic pain affects quality of life among adolescents and teens. They search for “chronic pain teens”. 
- In the **Studies tab**, HSS returns a list of studies that have studied chronic pain among adolescents and provides study information, including Common Data Elements (CDEs) used by each study. They see studies using the Pain Catastrophizing Scale and the PROMIS Pediatric Anxiety Scale.

- They then navigate to the **CDEs tab** where they find further information about relevant CDEs including exact measures and responses. This helps them to identify datasets for building a cohort for secondary analysis.  
 
### Designing Questionnaire For Multiple Clinical Trials Sites

Jordan, a data coordinating center manager for a HEAL research network, is preparing a questionnaire that will be used across the different clinical trials sites for the network.  He is trying to identify standardized measures for concepts such as anxiety, depression and quality of life and pain acceptance. He searches for the concept “depression”.

- In the **CDEs tab**, HSS returns a list of CDEs related to depression. Jordan is able to see the exact measures and responses for the CDEs, and a list of studies using each CDE. 

- Jordan uses this information to identify standardized questionnaires for the HEAL research network. 
 



Use Cases

### What is semantic search?

Semantic search is “search with meaning.” Search results are enhanced by incorporating knowledge about the search concepts, including synonyms and relationships to other concepts. For example, a semantic search for “high heart rate” would find results containing those three words or their variants (such as “frequency” instead of “rate”) as well as synonyms (such as “tachycardia” or “heart racing”) and related concepts (such as “arrhythmia”).

### Why should I use HEAL Semantic Search (HSS)?
HSS provides a smart way to search disparate HEAL datasets to uncover related studies, variables, Common Data Elements (CDEs), and research concepts. Use HSS if you are looking for studies, variables, or Common Data Elements (CDEs) related to biomedical concepts, rather than a particular researcher, project or dataset.

### When should I use HEAL Semantic Search (HSS), and when should I use the HEAL Platform’s search function?

[HSS](/resources/semanticsearch) and HEAL Platform search are complementary search functions providing quick and effective HEAL study data search. The table below highlights HSS and HEAL Platform functions and how each can be used to dive deeper with the other. Our goal is to eventually integrate these two tools so users can query the HEAL Data Ecosystem seamlessly using either.

These tools complement each, enhancing result exploration. The HEAL Platform delves deeper into the studies and variables uncovered through HSS. Meanwhile, HSS aids in discovering related information pertaining to studies, variables, or concepts of interest identified within the HEAL Platform.

HEAL Platform Search
- Discover HEAL-funded projects
- Find researchers (PIs) working on these projects
- Conduct keyword search 
- View study metadata 
- Discover datasets that might be of interest

HEAL Semantic Search
- Find connections between biomedical concepts
- Expand the search by looking at synonyms and related terms
- Identify HEAL studies, variables, and Common Data Elements (CDEs) related to search terms and search results 
- Explore knowledge graphs

### What is a “concept” in HEAL Semantic Search (HSS)?

In HSS, a concept is a named entity, defined in an ontology or other formal knowledge source. Concepts have a unique identifier, preferred label, and a precise definition. For example, chronic pain is identified as [HP:0012532](http://purl.obolibrary.org/obo/HP_0012532), which is labeled “chronic pain” and is defined as “Persistent pain, usually defined as pain that has lasted longer than 3 to 6 months.” In HSS, concepts include categories such as disease, phenotype, biological process, or anatomical entity.

### Can I search multiple terms at the same time on HEAL Semantic Search (HSS), like “knee pain AND obesity?”

Yes, you can search for multiple terms by using the "+" operator. For example, to search for chronic pain and back, enter "chronic pain + back" into the search bar. 

### How does HEAL Semantic Search (HSS) find related concepts or studies for a search term?

HSS uses named-entity recognition tools to identify biomedical concepts in the variable names, descriptions, and measurement instruments (question text, permissible values). When you search for a term, it will start by looking for concepts, studies, CDEs, and variables that contain your search term in its name, description or list of synonyms. It will also use relationships between concepts from the [ROBOKOP Knowledge Graph](https://robokop.renci.org/) to suggest additional items that might be related to your search term. 

### How does HEAL Semantic Search (HSS) rank or score search results?

Search results are scored based on their search term relevance/closeness. Higher scoring, more relevant results are first. Further down the search results screen, the returned results become more obscure, are further from the search term, and score lower. HSS understands related biomedical concepts and search synonyms, and it uses these to prioritize search results. 

### Where does HEAL Semantic Search (HSS) ingest data/information from?

HSS ingests HEAL study variable-level metadata (VLMD) and integrates them in existing knowledge graphs ([ROBOKOP](https://robokop.renci.org/)) to combine distinct concepts and disparate data types.  You can learn more about metadata in the HEAL Data Ecosystem on the [Metadata page](https://www.healdatafair.org/resources/metadata).

In addition, HSS includes common data element (CDE) information collected by the [HEAL CDE program](https://heal.nih.gov/data/common-data-elements), so we can point you to relevant CDEs.

### Why is it critical that investigators share their data dictionaries with the HEAL Stewards for HEAL Semantic Search (HSS) development?

HSS quality and robustness depends on ingested data dictionary volume. More ingested data dictionaries will result in richer, more relevant user search results, since HSS indexes data dictionary information to well-established biological ontologies (including biomedical concepts and their synonyms) to connect biomedical concepts. 

Please note, data dictionaries contain information about study variables, not the data itself. Sharing your data dictionary allows others to discover your study and may help identify potential collaborators and novel research questions. It also contributes to NIH’s overarching goal of making HEAL data FAIR (Findable, Accessible, Interoperable, and Reusable). 

### A concept I searched for doesn’t have many results, but I know the data is out there. Why doesn’t HEAL Semantic Search (HSS) know about it?

HSS returns search results related to the data dictionaries it ingests, indexes, and builds knowledge graphs for. If HSS hasn’t ingested related information, it cannot connect between your search term and what is available. We are always looking for new data dictionaries to add, improving search results. We encourage you to submit your own data dictionaries for HSS through the [HEAL Platform’s variable-level metadata (VLMD) submission system](https://heal.github.io/platform-documentation/vlmd_submission/) or by [emailing the HEAL Stewards](mailto:HEALStewards@renci.org). If you’re sure that the concept you searched for must exist in HSS, please report this issue on the [HSS Support Portal](http://bit.ly/HEALSemanticSearch_Help).

### Why does HEAL Semantic Search (HSS) return results in the order it does?

HSS returns results based on their relevance to your search term. Usually, the result order is because:
1. The concept name or description includes the search term.
2. It knows the search term is a synonym for the concept name, e.g., “arthralgia” instead of “joint pain.”
3. A concept related to your search term is also displayed (e.g., bleomycin is displayed when searching for “joint” because the knowledge graph shows it can cause arthritic joint disease as a side-effect).

### How can I contribute my study’s information to HEAL Semantic Search (HSS)? Are there any legal restrictions on what I can submit?

Please submit your data dictionaries through the [HEAL Platform’s variable-level metadata (VLMD) submission system](https://heal.github.io/platform-documentation/vlmd_submission/) or by [emailing the HEAL Stewards](mailto:HEALStewards@renci.org). Your study does not have to be complete; we are happy to accept any data dictionary, even if it is not finalized. Data dictionaries inherently should not contain sensitive information such as personal health information (PHI) or personally identifiable information (PII). For more information about sensitive data, [click here](https://www.healdatafair.org/sensitive-data).

### How can I share feedback or suggestions for future enhancements for HEAL Semantic Search (HSS)?

We welcome your suggestions to build a better HSS! Please visit our Help Portal and [submit an improvement ticket](https://renci.atlassian.net/servicedesk/customer/portal/3/group/7/create/25). 



FAQs

To report a bug, request technical assistance, submit user feedback, or submit a different request, visit our [Help Portal](https://renci.atlassian.net/servicedesk/customer/portal/3).

Help Desk

If you are a HEAL Investigator or part of a HEAL study team, you can help unlock the potential of the HEAL Initiative by sharing information about your study to make it findable in HEAL Semantic Search. Data dictionaries precisely specify which variables you are planning to collect and how they will be measured. Sharing your data dictionary not only streamlines data reuse but also ensures your research is Findable, Accessible, Interoperable, and Reusable (FAIR). Your contribution will play a pivotal role in advancing research and promoting data reuse and collaboration.

 Submit your data dictionary through the [HEAL Platform’s variable-level metadata (VLMD) submission system](https://heal.github.io/platform-documentation/vlmd_submission/) or by [emailing the HEAL Stewards](HEALStewards@renci.org) and be a catalyst for transformative research to address the opioid epidemic.

Even if your study is a work in progress, your data dictionary can help fellow investigators discover your research. Data dictionaries inherently should not contain sensitive information such as personal health information (PHI) or personally identifiable information (PII).
