openmolecules.org

 
Home » DataWarrior » Functionality » Similarity heatmap (Visual heatmap)
Similarity heatmap [message #1150] Wed, 02 December 2020 13:54 Go to next message
cfuttrup is currently offline  cfuttrup
Messages: 2
Registered: December 2020
Location: Copenhagen
Junior Member
We are looking at the possibility of using openmolecules as a way of determining with proteins we need to analyse closer based on similarity and heatmap. I can see that Openmolecules doesn't have a heapmap currently and I am wondering, were to look to implement this functionality.
Re: Similarity heatmap [message #1151 is a reply to message #1150] Wed, 02 December 2020 20:33 Go to previous messageGo to next message
thomas is currently offline  thomas
Messages: 711
Registered: June 2014
Senior Member
Currently, DataWarrior cannot show heatmaps also it is able to perform a hierarchical clustering. Probably, heatmaps will come some time in the future. Protein similarity is another thing that is not likely to be added soon, because DataWarrior focuses on small molecules. Proteins will play a role soon, but as target for docking only.
Re: Similarity heatmap [message #1153 is a reply to message #1151] Thu, 03 December 2020 10:47 Go to previous messageGo to next message
cfuttrup is currently offline  cfuttrup
Messages: 2
Registered: December 2020
Location: Copenhagen
Junior Member
If I wanted to contribute and implement a heatmap where so I look and begin?
Re: Similarity heatmap [message #1162 is a reply to message #1153] Fri, 04 December 2020 16:29 Go to previous message
thomas is currently offline  thomas
Messages: 711
Registered: June 2014
Senior Member
It would need two parts: First the columns from a hierarchical clustering result would have to contain more information: at least a reference to the parent column and the similarity value to its partner in the similarity graph. The clustering in DataWarrior is hierarchical, but doesn't store these values. One might reconsider, whether there are alternatives to the algorithm, which is brute force and needs the full similarity matrix.

Second and much more effort: Currently all 2D-views are calculated and drawn by one class JVisualization2D (with support of its parent class JVisualization). It is overdue to break up the various 2-dimensional views (scatter, bars, boxes, whiskers, pies) into separate classes, which still share a lot of properties (marker size, color, labels, font size, ...), but I didn't ever have the time, because a lot would be involved including reorganising the way view properties are saved in templates and how macro properties interact with views.

The fastest, but messy approach would be to add another preferred view (heat map) to the JVisualization2D class, but working on the foundation first would be the better approach. I suggest you look into the source code and get a picture. If you then really like to contribute substancial time, then we should closer discuss options. Of course, I am grateful for any support.

Thomas

[Updated on: Fri, 04 December 2020 16:29]

Report message to a moderator

Previous Topic: Red Box
Next Topic: Loop over columns for calculations
Goto Forum:
  


Current Time: Sun Nov 03 17:39:25 CET 2024

Total time taken to generate the page: 0.03607 seconds