search

monarch-initiative / monarch-legacy

Monarch web application and API
BSD 3-Clause "New" or "Revised" License
42 stars 37 forks source link

Patient clustering algorithm and visualization #1273

Open jmcmurry opened 8 years ago

jmcmurry commented 8 years ago

Think about what algorithms and UI would be needed to explore how the UDP patients may be phenotypically and or demographically clustered.

mellybelly commented 8 years ago

@cmungall @drseb we need this for our UDP work, but could be the basis for a nice monarch UI feature as well.

Create a tool that can be used by researchers and clinicians that can operate on patient data (for example, data extracted from the UDPICS database). The tool will not be web based, and is intended to be executed either within the researcher's desktop environment (assuming privacy requirements are met) or on servers behind the UDP firewall. The tool will generate reports about clusters of potentially similar UDP cases, similar known diseases including new ones, and similar matching models. The tool will be an extension of the Monarch Initiative owlsim phenotype similarity tool. We will implement standard clustering techniques using the phenotype distance output of owlsim, and we will provide basic visualization of the clusters, either through 3rd party libraries or via integration with environments such as Cytoscape.

Various methods of displaying results will be tested, including network diagrams. The software will be developed to be integrated with UDPICS, but will also accommodate an external data source so that UDN data outside the UDP can be searched a well (this would likely be done by creating a local data store with periodic updates from the UDN gateway).

harryhoch commented 8 years ago

I've been giving some thought to potential visualizations. The key will be to integrate not just display of the clusters, but some representation of the metrics contributing to the similarity measurements...

jmcmurry commented 6 years ago

Relevant to Aim 3 of R24

jmcmurry commented 6 years ago

@kshefchek to link to the prototypes. We of course were limited by the paucity of actual patients and by the compute needed to render all of the iterations and connections.