- Fixed test for the
`update_biomarker_files`

function (writes to`tmpdir()`

instead of the user’s library directory)

- Finally added tests to the package!
**Coverage is now 97%**. - Used the pkgdown package to create static html documentation for emba. Check it here!
- Change MCC calculation to return 0 when undefined/
`NaN`

MCC scores were produced (which is the correct limiting value - see Chicco at al. (2020)). Thus, the previous versions handling of`NaN`

MCC scores, is now deprecated. - Add the
`penalty`

parameter to account for the difference in model group size when calculating the average activity or link operator data differences. This minimizes the bias in the returned biomarkers.- For the implementation check the function
`emba::get_vector_diff`

and the corresponding StackOverflow question. - To get the same results as with previous versions of this library, use
`penalty=0`

in the general`emba::biomarker_*`

functions (though the results will probably be very biased and that’s why the default value for the`penalty`

is now**0.1**).

- For the implementation check the function
- Changed documentation to specify that the
`models.stable.state`

parameter used in various functions can take any values in the [0,1] interval and not just 0 (*inactive*) and 1 (*active*). - The following functions do not take the redundant parameter
`models`

anymore:`emba::get_avg_link_operator_diff_mat_based_on_tp_predictions`

`emba::get_avg_activity_diff_mat_based_on_tp_predictions`

`emba::get_avg_activity_diff_based_on_tp_predictions`

- Refactor several of the functions that load the results from the DrugLogics pipeline:
- If a model has less or more than 1 stable state, it’s discarded and a message is printed.
- Return value is now a
`data.frame`

object instead of a`matrix`

. - The models names do not have the annoying
`.gitsbe`

extension anymore. - These changes affect the following functions:
`emba::get_link_operators_from_models_dir`

,`emba::get_stable_state_from_models_dir`

and`emba::get_model_names`

.

- The general functions
`emba::biomarker_mcc_analysis`

and`emba::biomarker_tp_analysis`

do not use the`calculate.subsets.stats`

input option anymore. The`emba::biomarker_synergy_analysis`

continues to do so and now also calculates and returns all possible synergy set and subset pairs that miss just one of the model predicted synergies (`emba::get_synergy_comparison_sets`

). - Various small bug fixes and other code refactoring :)

`get_synergy_scores`

now supports reading both*ensemble-wise*and*model-wise*synergies files- add
`calculate.subsets.stats`

option to the general analysis functions (`biomarker_*`

) that decides if the powerset of the observed synergies and the number of models predicting each subset is going to be calculated. The default value is set to`FALSE`

to save computation time :)

- add function
`get_synergy_scores`

- fixed test that used a randomly generated matrix

- add function
`get_avg_link_operator_diff_based_on_synergy_set_cmp`

- add function
`get_avg_link_operator_diff_based_on_specific_synergy_prediction`

- add function
`filter_network`

- to use for visualizing induced subgraphs - update dependencies (set
`usefun`

min version to 0.4.3)

- Optimized
`count_models_that_predict_synergies`

function and added tests for it. For a benchmark see relative Stack Overflow thread

- Added a
`NEWS.md`

file to track changes to the package - Transferred functions from separate R scripts to the package
- Finished code refactoring and splitting to different modules
- Finished writing documentation for all functions