- Enable signature extraction without intersecting targets.
- Enable calculation of communities on non-square matrices.
- Results are now stored in database instead of a folder.
- Correlations are calculated alongside importances to give an intuition about the sign of the relationship.
- Release version for Bioconductor 3.17.
- Discontinue the use of .data due to its deprecation since tidyselect 1.2.0.
- Update minimum required versions for some tidyverse related packages due to function deprecation.
- Release version for Bioconductor 3.16. See changes for 1.5.x.
- Added different modeling functions. Might not be completely backwards compatible!
- Release version for Bioconductor 3.15. See changes for 1.3.x.
- Switched from Louvain to Leiden algorithm for community detection (requires igraph >= 1.2.7).
- The metamodel is now build by ridge regression. Intercept p-values are not calculated, the values are set to NA for backwards compatibility.
- Unique value error for cv folds is downgraded to warning.
- Prefix can be added to the column names generated by the fucntions add_juxtaview and add_paraview. This allows modeling the maker expression by its own juxtaview and paraview.
- Bug fixes.
- Fixed a separator issue in results aggregation and signature generation that might cause issues with variable names containing "_".
- Release version for Bioconductor 3.14. See changes for 1.1.x.
- Added functions for view manipulation, including view filtering and marker selection.
- Added functions for performance, contribution and importance signature extraction from results.
- Aggregation and signature generation is generalized for samples with non-identical targets by working on the intersection.
- Modeling of intraview can be bypassed.
- Added families of distances to calculate paraview.
- Paraview can exlcude measurements within a used defined zone of indifference around each spatial unit.
- Improved plotting control.
- Complete test coverage.
IMPORTANT: R2 is now reported in percentages for intra, multi and gain. Collecting results from running mistyR < 1.1.11 will lead to miscalculation of gain.R2. Update the performance.txt files by multiplying the values in columns intra.R2 and multi.R2 by 100.
- Fixed display of messages and progress during view generation.
- Improved plotting control and display.
- Fixed handling of NaN in results.
- Vignette output switched from BiocStyle to rmarkdown for pdfs due to BiocStyle issue.
- Bugfix: models built with different parameters stored and retrieved from the same cache file.
- Avoid calls to os-dependent file.info in tests.
- Bugfix: passing arguments to ranger.
- Warnings on clearing nonexistent cache folders and tests of performance.
- Increased test coverage.
- Release version for Bioconductor 3.13.
- Fixed a bug in Nystrom approximation for creating paraview.
- Added a suite of tests with high coverage.
- Cleaner cache control.
- Removed magrittr from dependencies. Reexported pipe operator. Removed from examples.
- Caching turned off by default.
- Added parameter for appending performance and coefficient files in run_misty.
- Internal functions in views.R are now explicit.
- Removed alternatives and additional examples for function use from documentation.
- Removed redundant messages, escalated to warnings where required.
- run_misty cleans up empty cache directories.
- Vignette compatibility with the new release of SpatialExperiment.
- Remove Seurat vignette from package due to missing hdf5r binary for R 4.1 on CRAN for MacOS.
- README.md figures moved to the cloud.
- All passed paths and cache location are normalized.
- Changes in the mistyR vignette to reflect changes to insilico evaluation from the paper.
- Version with vignettes ready to submit to Bioconductor.
- Initial beta release of mistyR (named as MISTy) with function documentation.