Online Narrative and Codified feature Search Engine (ONCE)
What is it?
For a target disease with its name as input, the ONCE system can generate a list of related EHR features:
- Standardized clinical concepts represented by concept unique identifiers (CUIs) according to the unified medical language system (UMLS)
- PheCodes mapped from ICD codes;
- Ingredient-level RxNorm codes mapped from Medication prescriptions;
- Grouped CPT procedure codes according to the Clinical Classifications Software;
- LOINC codes mapped from laboratory tests.
The output of ONCE can be used as input features for predictive modeling of a target disease phenotype (see a paired phenotyping app - KOMAP ;
Why does it work?
- Representative learning: The key component that powers the ONCE feature selection engine is the semantic representations of all EHR codified and NLP concepts, trained via multi-source knowledge graph representation learning.
- Online article knowledge: To identify an initial set of clinical concepts important to a target disease, the ONCE system additionally leverages a large knowledge base of medical articles that describe the relatedness between narrative features and the disease at a higher level. The current article corpus comprises data from seven online sources including Wikipedia, Medscape, and Merck Manuals.
A composite feature score:
The final ONCE selection criteria are based on a composite score that integrates information
- Relatedness of a candidate concept to the target concept as quantified by the semantic embeddings;
- Frequency of the candidate concept in the EHR;
- Weighted frequency of the concept in the disease-related knowledge source articles.
How does it work?
- Step 1: Type in your disease of interest;
- Step 2: Choose the main CUI(s) and the main PheCode corresponding to the disease. ONCE already select several candidates for you. You can select multiple main CUIs and a single main PheCode among them.
- Step 3: Just click the "Search" button and you are ready to go!