Scoring scenario + refinement scenario + alascan page

[sverhoeven]

Fixes #138 Fixes #139 Fixes #143

To test upload https://github.com/haddocking/haddock3/blob/main/examples/scoring/data/T161-rescoring-ens.pdb

TODO

• [x] latest haddock3 is used, with no 'The following error occurred 1' errors
• [x] alascan report, it generates plot per cluster, should have page that shows all clusters
• [x] show alascan score in 3D?
• [x] show score in 3D
• [x] show scores of molecules without clustering? look at 2_caprieval module output
• [x] allow user to adjust clustering?
• [x] include caprieval score without clustering?
• [x] scoring page
• [x] refinement page
• [x] handle non-easy parameters in workflow.cfg, see comment below

Scoring Scenario page

Scoring Scenario result

Refinement scenario page

Refinement scenario result

Alascan page

[sverhoeven]

@VGPReys could you have a look at the screenshots?

And have a look at the constructed workflow at https://github.com/i-VRESSE/haddock3-webapp/blob/9ddb448553574c3a62e427d49ab883cd7a2a74d7/app/routes/scenarios.scoring.tsx#L58-L95

[sverhoeven]

@VGPReys

1. How could I include scores of unclustered molecules, add caprieval before clustfcc?
2. Should an user be allowed to set clustering parameters? if so which ones and how much?

[VGPReys]

@sverhoeven For the workflow, I would rather go for something like:

# =================================================================================== 
 # CAPRI Scoring example 
 # =================================================================================== 
 # The Critical Assessment of PRedicted Interactions (CAPRI) experiment 
 #  aims to do test methods that model macromolecular interactions in 
 #  blind predictions based on the three-dimensional structures of proteins. 
 # For more information, please visit: https://www.ebi.ac.uk/pdbe/complex-pred/capri/ 
 # =================================================================================== 
 run_dir = "${JOB_OUTPUT_DIR}" 

 # molecules to be scored (an ensemble PBD) 
 molecules = ${molecules} 

 # =================================================================================== 
 [topoaa] 
  tolerance = 10

 [emscoring] 
  tolerance = 10
 [caprieval] 

 [clustfcc] 
 min_population = 1
 plot_matrix = true 

 [seletopclusts] 
 top_models = 4

 [alascan]

 [caprieval] 

 # =================================================================================== 

• removing [mdscoring] as it is too costly.
• adding a caprieval after emscoring (related to your point 1)
• removing top_cluster = 1, so we retrieve all clusters.
• adding tolerance to make sure the workflow will terminate successfully.

2. Should an user be allowed to set clustering parameters? if so which ones and how much?

Yes, a user could be able to tune clustering parameters. I would make visible:min_population and clust_cutoff. Maybe also top_models and top_cluster from [seletopclusts].

[amjjbonvin]

Do we need alascan in this scenario?

Could be interesting, but then probably after caprieval?

[VGPReys]

I was thinking that I could bring interesting insights on what are the key components for the interaction. Not required. Also [contactmap] could be added, for even higher degree of analysis performed on the scoring set. You have the last words !

[VGPReys]

@sverhoeven Also, this scenario could handle multiple input files (up to 20).

[VGPReys]

ok @sverhoeven , after some discussion with Alex, we decided to:

• remove both [alascan] and [contactmap] from the scoring scenario, as it may generate too many plots and take too much time (we could expect user providing thousands of models)

But those two modules will be part of the refinement scenario.

[sverhoeven]

ok @sverhoeven , after some discussion with Alex, we decided to:

• remove both [alascan] and [contactmap] from the scoring scenario, as it may generate too many plots and take too much time (we could expect user providing thousands of models)

But those two modules will be part of the refinement scenario.

OK, so refinement is another scenario, which is similar to https://github.com/haddocking/haddock3/blob/main/examples/refine-complex/refine-complex-test.cfg . I will create issue for it.

[sverhoeven]

Ah yes without mdscoring it takes 18 seconds for https://github.com/haddocking/haddock3/blob/main/examples/scoring/data/T161-rescoring-ens.pdb

[sverhoeven]

• adding tolerance to make sure the workflow will terminate successfully.

tolerance is a expert parameter with default 0. So a user with only easy expertise level will get an error that they are not allowed to use that parameter.

The webapp uses same endpoint to upload workflow from scenario, build, upload pages. I can not have an endpoint that accepts parameters not part of the users level, because that could be misused.

What should I do?

1. Do not set tolerance parameter when user has easy level, but do set when expert or guru
2. Change default to 10
3. Move tolerance to easy level

Where should I do it?

1. In haddock3 defaults.yaml
2. In workflow builder
3. In haddock3 webapp,

[sverhoeven]

For now picked 1 for what and 3 for where, but this gives users with easy level worst results.

[amjjbonvin]

I would go for 2 and 1 :-)

[sverhoeven]

Some other PRs where merged into this branch, please review by focusing on functionality or this pr not the other prs.

[sverhoeven]

Sorry review is taking to long, merging it, review can be done on merge pr if needed.

[github-actions[bot]]

Please delete the images belonging to this Pull Request as they are no longer useful.

• ghcr.io/i-vresse/haddock3-webapp
• ghcr.io/i-vresse/bartender
• ghcr.io/i-vresse/certmaker

Goto versions page of each image, find version called pr-140, filter on tagged versions by clicking "N tagged" button in table header, click on ... button and select Delete version.