Skip to content

Conversation

@xueweic
Copy link
Collaborator

@xueweic xueweic commented Nov 6, 2025

Major change: Updated highlighting parameters in colocboost_plot function

The colocboost_plot function now provides flexible highlighting options with the following parameter changes:

  • change add_vertical = TRUE to add_highlight = TRUE.
  • change add_vertical_idx to add_highlight_idx.
  • add one additional add_highlight_style = "star", displaying highlights as red star markers.

For detailed examples, see the Visualization of ColocBoost Results. This update reflects user feedback and suggestions. We thank all users for their valuable suggestions that improved this functionality.

Minor changes:

  • Thanks efforts from @kal26 to enhance the documentation and test examples for our wrapper pipeline. We move the entire vignettes with the enhanced documentation to pecotmr.

  • Computational Improvements
    Using profvis to check the computational time for each part of function components.

    • check_pair_jkeach: reduced computational burden by checking the upper triangle first instead of all pairs.
    • Addressed LD checking issues in LD-free mode.
    • Improved matrix multiplication efficiency using tcrossprod function.
    • Pre-calculated LD values for outcome-specific parameter updates at each boost iteration
  • Bug Fixes

    • Recalculate purity for merged CoS and filtered merged CoS whenpurity < min_abs_corr.
    • Fixed error when focal outcome fails multiple testing correction at the first iteration in disease-prioritization mode
    • Fixed bug in top variable selection within uCoS details
    • Fixed one-iteration model for single-trait fine-mapping
    • Fixed variables duplication issues: only keep one if duplicated variables
    • Fixed missing summary statistics issues: remove summary data if Z score or variant names are missing.

xueweic and others added 13 commits September 12, 2025 20:58
Using profvis to check the computational time for each part of functions. Improve on several way:
1. check_pair_jkeach: instead of checking every pair, we check the upper tri first to reduce the burden.
2. Still some LD checking issue caused the LD-free mode.
3. change matrix multiplication using tcrossprod function
The highlighting feature has been updated to provide greater flexibility in visualization options. In previous versions, additional highlight variables in the colocboost plot could only be displayed as vertical lines by setting add_vertical = TRUE.
The updated colocboost_plot function now offers more versatile highlighting capabilities through the new add_highlight = TRUE parameter, which allows users to display highlight variables as either vertical lines or red star markers, providing improved customization for data visualization needs.
To enhance computational efficiency, we pre-calculate LD values for each pair of outcome-specific best parameter updates at each boost update.
For the merged CoS, we recalibrate the CoS variants based on the merged weight. We should recalculate purity for the merged CoS and filter unpured merged CoS.
fix error when focal outcome does not pass multiple testing correction at the first place.
fix bug for choosing top variable in ucos details
fix fine-mapping one iteration issue
Comment out to avoid running this code here, as we do not have real data files in examples
handle duplication variables - remove one of them
1. Move colocboost wrapper to pecotmr.
2. Improve colocboost function - validation of the dataset.
3. Fix one error about the missing value or duplicated value in summary statistics
@xueweic xueweic changed the title Several update in Nov 6th Several update on Nov 6th Nov 6, 2025
@xueweic xueweic changed the title Several update on Nov 6th Several updates on Nov 6th Nov 6, 2025
@xueweic xueweic merged commit 4d4ea5b into StatFunGen:main Nov 7, 2025
1 of 7 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant