Add parameters for penalty and threshold #138

Jakob37 · 2024-10-18T13:06:56Z

Description

This addresses two issues in Lund with using Genmod compounds.

We use different threshold and score from those hard-coded here.
For single cases, we do not want to apply a penalty (i.e. we would use --penalty 0 here).

Close #137

Further discussed with @dnil and @northwestwitch in: Clinical-Genomics/scout#4823

Added

Optional command line parameter to set threshold and penalty applied in compound scorer.

Changed

Fixed

How to prepare for test

Ssh to relevant server (depending on type of change)
Use stage: us
Paxa the environment: paxa

Install on stage (example for Hasta):

bash /home/proj/production/servers/resources/hasta.scilifelab.se/update-tool-stage.sh -e S_[TOOL]-t [TOOL] -b [THIS-BRANCH-NAME] -a

How to test

Do ...

Expected test outcome

Check that ...
Take a screenshot and attach or copy/paste the output.

Review

Tests executed by
"Merge and deploy" approved by
Thanks for filling in who performed the code review and the test!

This version is a

MAJOR - when you make incompatible API changes
MINOR - when you add functionality in a backwards compatible manner
PATCH - when you make backwards compatible bug fixes or documentation/instructions

Implementation Plan

Document in ...
Deploy this branch on ...
Inform to ...

Jakob37 · 2024-10-18T13:17:09Z

Parameters are tested such that they respond.

First no penalty is applied, yielding score -6 (which matches the RankResult).

Then a penalty of 5 is applied, changing the score to -6.

Finally, the threshold for penalty is lowered below the low scoring compound, thus avoiding the penalty by having a passing compound variant. Score going back to -6.

jakob@laptop:~/src/genmod$ /usr/bin/python3 -m genmod.commands.base compound -p 1 test/3_scored.vcf --penalty 0 | grep Compounds | tail -1 | cut -f8 | sed "s/.*;Compounds=/Compounds=/" | tr ";" "\n" | cat -n
     1  Compounds=SAMPLE_ID:21_46602317_C_G>-11.0
     2  GeneticModels=SAMPLE_ID:AD|AR_comp
     3  ModelScore=SAMPLE_ID:99
     4  RankScore=SAMPLE_ID:-6.0
     5  RankScoreNormalized=SAMPLE_ID:0.3037974683544304
     6  RankScoreMinMax=SAMPLE_ID:-30.0:49.0
     7  RankResult=-12|0|0|1|1|0|3|0|0|0|1
     8  CompoundsNormalized=SAMPLE_ID:21_46602317_C_G>0.620253164556962
jakob@laptop:~/src/genmod$ /usr/bin/python3 -m genmod.commands.base compound -p 1 test/3_scored.vcf --penalty 5 | grep Compounds | tail -1 | cut -f8 | sed "s/.*;Compounds=/Compounds=/" | tr ";" "\n" | cat -n
     1  Compounds=SAMPLE_ID:21_46602317_C_G>-16.0
     2  GeneticModels=SAMPLE_ID:AD|AR_comp
     3  ModelScore=SAMPLE_ID:99
     4  RankScore=SAMPLE_ID:-11.0
     5  RankScoreNormalized=SAMPLE_ID:0.24050632911392406
     6  RankScoreMinMax=SAMPLE_ID:-30.0:49.0
     7  RankResult=-12|0|0|1|1|0|3|0|0|0|1
     8  CompoundsNormalized=SAMPLE_ID:21_46602317_C_G>0.5569620253164558
jakob@laptop:~/src/genmod$ /usr/bin/python3 -m genmod.commands.base compound -p 1 test/3_scored.vcf --penalty 5 --threshold -20 | grep Compounds | tail -1 | cut -f8 | sed "s/.*;Compounds=/Compounds=/" | tr ";" "\n" | cat -n
     1  Compounds=SAMPLE_ID:21_46602317_C_G>-11.0
     2  GeneticModels=SAMPLE_ID:AD|AR_comp
     3  ModelScore=SAMPLE_ID:99
     4  RankScore=SAMPLE_ID:-6.0
     5  RankScoreNormalized=SAMPLE_ID:0.3037974683544304
     6  RankScoreMinMax=SAMPLE_ID:-30.0:49.0
     7  RankResult=-12|0|0|1|1|0|3|0|0|0|1
     8  CompoundsNormalized=SAMPLE_ID:21_46602317_C_G>0.620253164556962

Jakob37 · 2024-10-18T13:18:50Z

genmod/commands/score_compounds.py

@@ -45,8 +45,10 @@
                    is_flag=True,
                    help='If variants are annotated with the Variant Effect Predictor.'
 )
+@click.option('--threshold', type=int, help="Threshold for model-dependent penalty if no compounds with passing score", default=9)


Feel very free to help me improve these help texts

Correctly spelling threshold is already a strong text! 😸

Jakob37 · 2024-10-18T13:21:47Z

Not sure how to turn to with reviewing here. Maybe @dnil , @northwestwitch or @torbjorgen could give this a look?

Jakob37 · 2024-10-21T09:08:47Z

Thanks for the review @torbjorgen !

I don't have merge-access to the repo. Maybe someone else could help me merge this PR?

Jakob37 · 2024-10-24T05:20:46Z

I don't have access to this repo, so I am unable to merge / release.

Not sure who to ask. Could maybe @torbjorgen or @dnil help me out here? With merging or giving me access to the repo so I can do it myself.

dnil

This looks safe enough - defaults are exactly as before. The PR exposes a previously somewhat internally hidden, and rather important aspect of the model.

dnil · 2024-10-24T05:45:25Z

genmod/commands/score_compounds.py

@@ -45,8 +45,10 @@
                    is_flag=True,
                    help='If variants are annotated with the Variant Effect Predictor.'
 )
+@click.option('--threshold', type=int, help="Threshold for model-dependent penalty if no compounds with passing score", default=9)


Correctly spelling threshold is already a strong text! 😸

dnil · 2024-10-24T05:45:39Z

genmod/commands/score_compounds.py

 @click.pass_context
-def compound(context, variant_file, silent, outfile, vep, processes, temp_dir):
+def compound(context, variant_file, silent, outfile, vep, threshold: int, penalty: int, processes, temp_dir):


dnil · 2024-10-24T05:51:53Z

I would call this a minor, since it adds functionality in a completely backwards compatible manner. It does not correct a bug per se, but almost. ;)

Jakob37 · 2024-10-24T08:15:01Z

Thanks for the review 🙏

* Add parameters for penalty and threshold * Cleaning things up * Cleaning up and changelog

Jakob37 added 4 commits October 18, 2024 14:59

Add parameters for penalty and threshold

10d93f5

Cleaning things up

bc19f45

Cleaning up and changelog

98935d6

Merge master branch

5127fb0

Jakob37 marked this pull request as ready for review October 18, 2024 13:08

Jakob37 commented Oct 18, 2024

View reviewed changes

torbjorgen approved these changes Oct 21, 2024

View reviewed changes

dnil self-assigned this Oct 24, 2024

dnil approved these changes Oct 24, 2024

View reviewed changes

dnil merged commit c30e149 into Clinical-Genomics:master Oct 24, 2024
3 checks passed

dnil mentioned this pull request Oct 24, 2024

Bump version to 3.9 #139

Merged

15 tasks

dnil pushed a commit that referenced this pull request Oct 24, 2024

Add parameters for penalty and threshold (#138)

f02a6f8

* Add parameters for penalty and threshold * Cleaning things up * Cleaning up and changelog

dnil mentioned this pull request Oct 24, 2024

Change or document AR_compound low scoring compound penalty #107

Open

fellen31 mentioned this pull request Nov 6, 2024

Rank score penalty genomic-medicine-sweden/nallo#492

Open

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Add parameters for penalty and threshold #138

Add parameters for penalty and threshold #138

Jakob37 commented Oct 18, 2024 •

edited by dnil

Loading

Jakob37 commented Oct 18, 2024

Jakob37 Oct 18, 2024

dnil Oct 24, 2024

Jakob37 commented Oct 18, 2024

Jakob37 commented Oct 21, 2024

Jakob37 commented Oct 24, 2024

dnil left a comment

dnil Oct 24, 2024

dnil Oct 24, 2024

dnil commented Oct 24, 2024

Jakob37 commented Oct 24, 2024

Add parameters for penalty and threshold #138

Add parameters for penalty and threshold #138

Conversation

Jakob37 commented Oct 18, 2024 • edited by dnil Loading

Description

Added

Changed

Fixed

How to prepare for test

How to test

Expected test outcome

Review

This version is a

Implementation Plan

Jakob37 commented Oct 18, 2024

Jakob37 Oct 18, 2024

Choose a reason for hiding this comment

dnil Oct 24, 2024

Choose a reason for hiding this comment

Jakob37 commented Oct 18, 2024

Jakob37 commented Oct 21, 2024

Jakob37 commented Oct 24, 2024

dnil left a comment

Choose a reason for hiding this comment

dnil Oct 24, 2024

Choose a reason for hiding this comment

dnil Oct 24, 2024

Choose a reason for hiding this comment

dnil commented Oct 24, 2024

Jakob37 commented Oct 24, 2024

Jakob37 commented Oct 18, 2024 •

edited by dnil

Loading