beasleyjonm / AOP-COP-Path-Extractor

Construct Adverse/Clinical Outcome Pathway queries and extract path information from ROBOKOP or other Biomedical Knowledge Graphs
1 stars 2 forks source link

Improvements #13

Open beasleyjonm opened 2 years ago

beasleyjonm commented 2 years ago

For each query pattern, display generalized, copyable Cypher query and link to Neo4j browser to allow users to explore the more deeply, if so desired.

Store node and edge properties in Dash datatable as dropdown/Tooltip options. Just split all properties into name + everything else.

Include optional (click button to display) boxes to apply constraints to intermediate nodes/edges. These would be as simply as specifying intermediate nodes, or as complicated as applying arguments that filter based on node/edge properties.

Rearrange app layout to simplify query builders and table view, such that everything is more center-aligned and requires less horizontal scrolling.

Adjust "positive class" box to allow partial string matching to include positives. That way, users can type just start, or just end, nodes to be included in positive class during PCA. -----Decided against this. Instead, you must use the ":" to separate start and end entities. But you can save state of positives to avoid typing everything back in on refresh.

Implement Random Forest module to classify start-end pairs based on metapath embeddings. ------Working. Just specify positives and random forest with 100 trees, class weight selection, Y-randomization, and 5-fold cross validation will be performed. Model statistics will be displayed as bar chart.

Add hover-over icons to most features that explain the meaning and output of every action. ------Implemented with Dash Bootstrap library and the DB Tooltip component. Still need to add ALL explanations.

Allow ability to export figures, toggle edges, other cool features.

Show all possible IDs for a given found name.