Given a unique identifier (CUI or gene id), return the most related set of concepts
1) User finds concept by keyword search or browse and click somehow (e.g. ERAD-degradation pathway)
2) System identifies a set of related concepts
3) System produces N by N distance matrix for those concepts
4) Views are produced for that distance matrix (heatmap, binary network, sortable table)
Comments (5)
-
reporter -
reporter - edited description
- changed title to Given a unique identifier (CUI or gene id), return the most related set of concepts
specifying issue better. replaced ambiguous co-ccurrence idea based on previous same same code below with intention to use implicitome data.
Prior example based on MeSh co-occurrence and the Google Distance formula was in the SameSameWeb application. method = gatherMeshCloudData . source code = https://bitbucket.org/bgood/samesame/src/e4463b4430c87830dd8d156b86124efed3ce8d4e/src/org/scripps/samesame/Same.java?at=default
-
reporter To use Implicitome concept profiles (weighted concept vectors) to measure the distance between two concepts X and Y they use the formula:
-
reporter - changed milestone to Experimental
-
- changed version to 2.0.0
- Log in to comment
Should adapt step 3) above to use Implicitome concept profiles.
Not sure how to use Implicitome for step 1,2. Suspect a major pre computation is needed and ucsd team will work on that step.
In interim, suggest tackling based on information in semmedb as a starting point. e.g. query-> selected concept -> all predications with that concept -> all pmids associated with those predications -> all concepts in predications associated with those pmids. then take that list of concept ids back to implicitome tables and execute the vector comparisons to produce the distances. (e.g. cosine similarity).