pharmpy-core 0.109.0

Pharmacometric modeling

Pharmpy is a library and toolkit for pharmacometrics. It can be used as a regular Python package, in R via the pharmr package or via its built in command line interface.

Current features:

• A model abstraction as a foundation for higher level operations on models

• Functions for manipulation of models, e.g. changing model components like elimination or absorption

• Reading NONMEM models and results

• Running models and complex workflows (with NONMEM or to some extent nlmixr)

0.109.0 (2024-04-10)

New features

• The structsearch tool can now take an mfl string as search space

• Allow list of lists of parameters in iovsearch options

• Explicitly handle CMT columns for observations in the expression for F for NONMEM models

Changes

• Deafault to keeping eta on CL in iivsearch

Bugfixes

• Handle cases with covariates on MU when parsing phi-file with PHI columns

• Fix issues with allometry and structural covariates for PKPD models in amd

• Fix broken conversion to RxODE for models without ODE system

• Fix handling of datasets with multiple DVIDs in ruvsearch and sructsearch for TMDD models

0.108.0 (2024-03-18)

New features

• New simulation tool tools.run_simulations

• Add reports with various plots for final model in all AMD subtools

• Add VPC plot to AMD

• Add functions modeling.add_predictions_residuals and modeling.remove_predictions_residuals

Changes

• Force positive definitiveness in retries tool

• Make initial estimates in AMD mandatory

• Add default search space for PKPD models

• Remove TMDD models with less than 2 DVs

Bugfixes

• Correct extraction of CL/VC parameters in PSC metabolite models

• Remove unused K-parameters from NONMEM code

• Correct number of expected models in BIC calculation for bottom up algorithm in IIVSearch

• Use input model results when updating initial estimates for first model in bottom up algorithm in IIVSearch

• Add keep-option to bottom up algorithm in IIVSearch

0.107.0 (2024-03-04)

New features

• Add bottom up algorithm in IIVSearch

• Add modeling.set_dataset

• Add AMD results plots for each DV

• Add default search space for TMDD in AMD

• Support EFIM as parameter uncertainty method

• Allow specific (covariate, parameter) combinations as part of input for mechanistic covariates in AMD

Changes

• Use Pharmpy class Expr instead of sympy/symengine

• Remove ModelfitResults attribute from Model

• Change logic of IIVsearch algorithm argument, add correlation_algorithm argument

• Ignore datainfo fallback in AMD (to avoid automatically filling in information not given by user)

• Raise error instead of warn when expression is invalid in modeling.filter_dataset

• Change strategy names in AMD (to “default” and “reevaluation”)

Bugfixes

• Keep IIV on all clearance parameters of central compartment in AMD

• Fix bug in AMD for TMDD models where dir_name was not specified

• Fix bug with naming of K-parameters in models with 9 compartments

• Skip first order absorption with 1 transit (no depot) combination

• Fix bug which caused added IOVs to be removed in covsearch when running AMD (edited)

0.106.0 (2024-01-11)

New features

• Change to mBIC as default ranking function in modelsearch, iivsearch and iovsearch

• Add modeling.get_central_volume_and_clearance

• New option parameter_uncertainty_method to amd

• New option ignore_datainfo_fallback to amd

• Handle conversion to ETA/ETC for PHI/PHC in NONMEM phi files

Changes

• Remove the order option in amd and instead add strategy with “fixed” orders and options to subtools

Bugfixes

• Allow ~ in paths in write_csv

• Have non-linear elimination models in default search space for amd TMDD models

• Fix issue causing removed off-diagonal omegas being transformed into thetas

• Fix issues in frem postprocessing when using mu-referencing

0.105.0 (2023-12-07)

New features

• Add modeling.bin_observations

• Add modeling.plot_dv_vs_pred

• Add modeling.plot_abs_cwres_vs_ipred

• Support strictness for thetas, omegas and sigmas separately

• Support stagewise addition of covariates in amd

• Support multiple DVs for TMDD models

• Add retries tool

• Use retries in the amd

Changes

• Always keep an iiv eta on clearence in amd

0.104.0 (2023-11-06)

New features

• Add modeling.replace_non_random_rvs

• Add option keep_index to modeling.get_observations to allow keeping the original dataset index

• Add path-option to tools.fit

• Add function tools.is_strictness_fulfilled

• Add strictness option to AMD and subtools

• Add TMDD models to AMD

• Add option for TMDD models in structsearch

• MFL for COVSearch works the same way as for Modelsearch

Bugfixes

• Properly handle 0 FIX etas in calculate_bic

• Fix crash when setting 0 transit compartments

• Fix various bugs in TMDD models (including allometry)

0.103.0 (2023-10-12)

Changes

• Update initial estimates in structsearch

• Add option dv to modeling.get_individual_parameters

• Add default search space for drug-metabolite models in AMD

Bugfixes

• Fix bug in calculate_bic where parameters were incorrectly set to fixed for PKPD and drug-metabolite models

• Fix bug in COVSearch step numbering

• Fix bug in COVSearch where the final_model was set incorrectly

• Fix bug in COVSearch where p-value wasn’t displayed for backward models

0.102.0 (2023-09-28)

New features

• Add modeling.plot_dv_vs_ipred

• Add modeling.plot_cwres_vs_idv

• Add modeling.add_indirect_effect

• Add option for presystemic circulation for modeling.add_metabolite

• Add bic for multiple testing in modeling.calculate_bic

• Support PKPD models in the amd tool

• Support Drug-metabolite models in the amd tool

• Add first version of report for amd

• Add PKPD models to MFL

• Add modeling.filter_dataset

Changes

• Change default p-value in ruvsearch from 0.05 to 0.001

• Change default p-values in covsearch from 0.05 and 0.01 to 0.01 and 0.001

• Change the mfl for modelsearch to mean search space and not which transformations to do

• Change the syntax for LAGTIME in the MFL

Bugfixes

• Allow ~ for home directory in read_modelfit_results

0.101.0 (2023-09-01)

New features

• Add modeling.load_dataset and modeling.unload_dataset

• Add @BIOAVAIL to MFL

• Add support for iv-oral administration for amd tool

0.100.0 (2023-08-25)

New features

• Support for M5, M6 and M7 methods for blq data

• New symbols @PK and @PD in MFL

• Internal support for multiple doses to one compartment

Changes

• Change the blqdv type to blq in datainfo

• Better usage of BLQ and LLOQ columns for blq data

0.99.0 (2023-08-23)

New features

• Add function modeling.set_reference_values

• Add function modeling.set_lloq_data

• Parse IV+oral models using CMT column

• Specify DV in RUVSearch

• Option to add logit IIV in add_iiv

• New options for remove_loq_data

Changes

• Make Task and Workflow immutable

• Ignore fixed IIVs/IOVs in IIVSearch and IOVSearch

Bugfixes

• Fix bug where epsilons where removed in remove_iiv

• Fix bug in create_basic_pk_model to handle space separated datasets

0.98.0 (2023-07-21)

New features

• Support for multiple doses

• Add function modeling.add_bioavailability

• Add function modeling.remove_bioavailability

• Support for PKPD models in structsearch

• Option to keep IIVs in IIVSearch

• Option to test uncertainty methods in Estmethod

• Autogenerate CMT column

Changes

• Rename BLQ flag datainfo typ to blqdv

0.97.0 (2023-06-28)

New features

• Support BLQ transformations in RUVSearch

• New tool structsearch and support for TMDD models

• Add function modeling.set_direct_effect

• Add function modeling.add_effect_compartment

Changes

• Reorganizing of modeling module

• Support changing error model with BLQ transformation

• Add max_iter option for RUVSearch

0.96.0 (2023-05-26)

Changes

• Rename functions handling the precision matrix (was previously referring to information matrix which was an error)

• Remove saddle reset for default AMD model

• Let LLQ column takes precedence over BLQ column

New features

• Add tools.load_example_modelfit_results

Bugfixes

• Fix bug where if-statements were reordered incorrectly

0.95.0 (2023-05-22)

Changes

• ModelfitResults.ofv_iterations and ModelfitResults.parameter_estimates_iterations have NaN rows in failed runs

Bugfixes

• Fix bug causing changes in FIX from model1 to model4 to crash frem

• Fix bug causing individual parameters in $ERROR to crash frem

• create_report now does not assume that results.json already exists

• ~ for $HOME is now supported in write_model and create_report

• Fix bug where LLOQ value did not override column in dataset in transform_blq

• Correct BLQ indicator column condition in transform_blq

• Fix bug where modelfit results were not connected to model after a fit

0.94.0 (2023-04-26)

New features

• Support parsing assignments other than DADT in $DES in NONMEM

• Fix parsing of some complex ODE-systems in NONMEM

Changes

• Drop support for Python 3.8

Bugfixes

• Fix bug causing BIC calculation to fail for models having first order absorption and lag_time after going into zero order absorption

0.93.0 (2023-04-19)

New features

• Add function modeling.get_zero_order_inputs

• Add function modeling.set_zero_order_input

• Add function modeling.set_tmdd

• Added plugin to convert models to RxODE

• Support conversion of more models to nlmixr

Changes

• modeling.generate_model_code was renamed to modeling.get_model_code since the code is not generated by this function

• Do not use ADVAN7 because models that should work with ADVAN7 didn’t were found

Bugfixes

• Fix multiple bugs in parsing $TABLE headers

0.92.0 (2023-04-05)

New features

• Add function modeling.is_linearized

• Add function modeling.plot_transformed_eta_distributions

• Add function modeling.create_config_template

• Add function modeling.get_dv_symbol

• Add function modeling.get_initial_conditions

• Add function modeling.set_initial_condition

• Add function modeling.transform_blq

Bugfixes

• Fix bug where $ABBR wasn’t added for etas

0.91.0 (2023-03-03)

New features

• Add function modeling.create_basic_pk_model

• Add function modeling.add_metabolite

• Add function modeling.set_dvid

• Add function modeling.has_weighted_error_model

Changes

• model.dependent_variable becomes model.dependent_variables

Bugfixes

• Fix regression causing DEFDOSE to sometimes be put on the wrong compartment

• Fix ruvsearch crashing in case of bad modelfit_results (#1551)

0.90.0 (2023-02-24)

New features

• Add function modeling.has_odes

• Add function modeling.has_linear_odes

• Add function modeling.has_linear_odes_with_real_eigenvalues

• Add function modeling.is_real

• Support for more types of models in the nlmixr plugin

• Automatic selection between ADVAN5 and ADVAN7 for NONMEM models

Changes

• Remove modeling.copy_model

• Support nlmixr2 instead of nlmixr for the nlmixr plugin

• The Model class is now immutable

• update_source is run by all transformation functions

0.89.0 (2023-01-26)

New features

• Add function modeling.display_odes

• Add support for Python 3.11

Changes

• Naming of parameters for NONMEM models reworked. Configuration options removed.

• Only allow MFL as input to run_covsearch

• Remove read_model_from_database from pharmpy.modeling

• Merge ExplicitODESystem into CompartmentalSystem

0.88.0 (2022-12-21)

New features

• Add algorithm exhaustive_only_eval to Estmethod tool

• Add replace methods to Assignment, Compartment, Bolus and Infusion

Changes

• Rename algorithms in Estmethod tool: reduced -> exhaustive. exhaustive -> exhaustive_only_eval

• Always add iteration 0 in ofv_iterations and parameter_estimates_iterations for eval models with FO/FOCE

0.87.0 (2022-12-14)

Changes

• Allometry model will update initial estimates in allometry tool

• Base model in IIVSearch tool will update initial estimates

• Do not update initial estimates from model that did not minimize successfully (except rounding errors), this affects all AMD subtools

• Rename derive to replace in some base classes

0.86.0 (2022-11-30)

Changes

• Add description to proxy-models (#1314)

• Input check covariates (#1355), allometric variable (#1378) occasion-column before running AMD

Bugfixes

• Fix typo in COVSearch that caused it to select model with highest OFV (#1377)

• Ignore NaNs when selecting models in COVSearch (#1381)

• Fix issue where initial estimate for KM (in MM-elimination) was set outside of NONMEM’s bounds (#1064)

• Fix issue where individuals without observations were not filtered for general model objects (afd7707, #1139)

• Fix issue where saddle reset was not added in start model for AMD (#1394)

0.85.0 (2022-11-18)

Changes

• Covariates are defined in search space option in AMD-tool

• Store name of final model instead of final model in AMD

• Change methods and solvers option in estimation method tool: None means none should be tested

• Add FORMAT option if length of IDs are too long (#1139)

• Make Result classes immutable

Bugfixes

• Fix bug in results parsing where extracting whether parameters are fixed (#1117)

• Add timeout-loop to wait for .lst-file when renaming

0.84.1 (2022-11-13)

Changes

• Much faster parsing of NONMEM models

• 4 times faster parsing of NONMEM phi files

Bug fixes

• Have correct F-statement in $ERROR for $DES NONMEM models

• Read compartment names correctly when having both NCOMP and COMP in $MODEL of NONMEM models

0.84.0 (2022-11-09)

New features

• Add modeling.deidentify_data

Changes

• Change CLI anonymize into deidentify

0.83.0 (2022-11-01)

Changes

• Only test IOV on statements before ODE

Bug fixes

• Allow spaces in DADT definitions when parsing ODE

• Fix issue where expression setter was used

• Fix issue in IIVSearch where tool doesn’t continue to next step if there is a multivariate distribution

• Input dataset into model constructor in convert model (fixes #1293)

• Modelfit should not crash if .lst-file does not exist, warns if .lst and .ext-files do not exist (#1302, #1303)

0.82.0 (2022-10-24)

Changes

• modeling.summarize_modelfit_results takes results objects instead of model objects

• Do not include aic and bic in summarize_modelfit_results

0.81.1 (2022-10-24)

Bug fixes

• Handle NM-TRAN datasets with one and two digit year in DATx column using default LAST20 (50)

0.81.0 (2022-10-21)

New features

• Add tools.read_modelfit_results

Changes

• run_modelsearch, run_iovsearch, run_ruvsearch, run_allometry, run_covsearch, run_amd and run_iivsearch now need results as a separate input

Bug fixes

• Correct scaling for F in NOMEM models for ADVAN 2,4,5,7 and 12

0.80.0 (2022-10-19)

Changes

• modeling.fit returns ModelfitResults instead of Model

• Let zero_protect default to True for modeling.set_proportional_error_model

• Faster parsing of NONMEM table files

Bug fixes

• Let bioavailability parameters be part of rhs of ode_system

• Make sure initials are non-zero for absorption parameters

0.79.0 (2022-10-16)

New features

• Relative paths in files, absolute paths in Python objects (#1180, fixes 887)

• Validate tool inputs (#1162, fixes #1032)

• Add allow_nested flag to add_covariate_effect (#1004)

• Add has_covariate_effect and remove_covariate_effect (#1004)

• Generalize get_rv_parameters (#1181)

• 9fd701521 Store input models in tool database

• Replace best_model with final_model_name and retrieve_final_model

• a7fbcbfe2 Handle results and databases as input to retrieve_models

• Add modeling.update_initial_individual_estimates function

Changes

• Include input model as step 0 in summary_models for AMD tools

• Rename all AMD tool candidates such that modelsearch_candidate1 -> modelsearch_run1

• Add columns for number of parameters and delta parameters in summary_tool for AMD tools

• Only include chosen models in AMD summary_tool

• New names and description for COVSearch candidates

• Modify COVSearch summary_tool to include information from the steps-table, remove ranking

• New candidate descriptions in IOVSearch

• Add multiindex to RUVSearch which include step/iteration, remove ranking.

• Add algorithm column to IIVSearch summary_tool, remove algorithm from candidate name

• Compare final model in IIVSearch to input model, return input if worse

• Generalize detection of existing effects in add_covariate_effect (#1004)

• calculate_bic and calculate_aic will need the -2LL as input instead of modelfit_results

• calculate_eta_shrinkage needs the explicit arguments parameter_estimates and individual_estimates

• calculate_individual_shrinkage needs the explicit arguments parameter_estimates and individual_estimates_covariance

• check_parameters_near_bounds needs the parameter estimates given in the arguments

• check_high_correlations needs the correlation matrix as an explicit argument

• plot_iofv_vs_iofv takes two iofv series instead of two models as input

• plot_individual_predictions takes the predictions dataframe as input

• create_joint_distribution takes an option individual_estimates argument and does not use modelfit_results directly

• evaluate_expression to get parameter estimates from optional argument instead of from modelfit_results

• evaluate_population_prediction will not take parameter estimates from modelfit_results

• evaluate_individual_prediction will not take parameter estimates from modelfit_results

• evaluate_eta_gradient will not take parameter estimates from modelfit_results

• evaluate_epsilon_gradient will not take parameter estimates from modelfit_results

• evaluate_weighted_residuals will not take parameter estimates from modelfit_results

• sample_parameters_from_covariance_matrix will need parameter_estimates and covariance_matrix as explicit arguments

• sample_parameters_uniformly will need parameter_esimtates as explicit arguments

• sample_individual_estimates will need individual_estimates and individual_estimates_covariance as explicit arguments

• calculate_individual_parameter_statistics and calculate_pk_parameters_statistics will need parameter estimates and covariance matrix

• update_inits need explicit estimates as argument and does not use modelfit_results

• update_inits does not update initial individual estimates

• Move predict_* functions from modeling to pharmpy.tools

• Move summarize_individuals and summarize_individuals_count_table to pharmpy.tools

• Move print_fit_summary to pharmpy.tools

• Move write_results to pharmpy.tools

• Move summarize_errors to pharmpy.tools

• Move rank_models to pharmpy.tools

• Move summarize_modelfit_results to pharmpy.tools

• Speedup parsing of NONMEM results

Bug fixes

• 297a64041 Handle individual_ofv is None in dofv (fixes #1101)

• 57fc4fee8 Fix adding categorical covariate effects (#1004)

0.78.0 (2022-09-20)

Changes

• fd417aaf Always return a new model in convert_model

• d5458e36 Raise KeyError in LocalModelDirectory#retrieve_model (instead of FileNotFoundError)

• 1193bd39 Remove unused pharmpy.symbols submodule

Bugfixes

• bb96a13c Fix update_parameters when parameters are added

• 0ca786c5 Fix backward search of covsearch

• dd056da3 Fix for models with bioavailability parameters

• 915bc9c7 Fix get_config_path output when config file is disabled

• 82b32278 Remove some unwanted debug printing

• 1131a610 Fix issue in PsN SCM results parsing

• ebfafb45 Assign ODE as compartmental system to variable (#1173)

0.77.0 (2022-09-08)

Changes

• fb070ee1 Return input model if allometry model fails (#1049)

0.76.1 (2022-09-06)

Bugfixes

• Fix issue with adding allometry to models with MM elimination

• Make pyreadr an optional dependency, making Pharmpy easier to install on Mac M1

0.76.0 (2022-09-05)

New features

• Add modeling.get_evid to get or create evid from a model

• Add modeling.get_cmt to get or create a cmt column from a model

• New column type: “rate”

Changes

• Rename “resmod” tool to “ruvsearch”

• Return only DataFrame in modeling.rank_models

• Fall back to rank value if model fails in modeling.rank_models (fix #916)

• Rename “strictness” to “errors_allowed” in modeling.rank_models

• Only allow “rounding errors” by amd, iivsearch, iovsearch, modelsearch and covsearch (fix #1055)

• Add attibute significant_digits to ModelfitResults

Bugfixes

• Serialize modelfit results #1092

• Exlude “unreportable number of significant digits” in modeling.rank_models (fix #1076)

0.75.0 (2022-08-10)

New features

• SCM forward search followed by backward search in covsearch (#988)

Changes

• Change initial estimates of IIV parameters of start model in AMD tool (1c65359)

• Change default order of subtools in AMD tool (42fe72f)

Bugfixes

• Make NONMEM column renaming work in more cases (#1001)

• Fix issue when search spaces which lead to uneven branch length in reduced stepwise algorithm (#694)

• Fix issue with error record not numbering amounts properly with non-linear elimination (#708)

• Fix issue with comments being removed in omega blocks (#790, #974)

• Fix ranking issue when candidate models do not produce an OFV (#1017)

• Fix issue with reading datasets in AMD with RATE column (#989)

0.74.0 (2022-07-18)

Changes

• Rename pharmpy.parameter to pharmpy.parameters (71f4cf23)

• Merge COVsearch DSL into MFL (#932, #973)

• Add ZO absorption to default search space in AMD (cfc09bad)

Bugfixes

• Make run_amd work in more cases (#975)

• Make run_iovsearch work in more cases (#917, #977)

• Make remove_iov work in more cases (#917)

• Make get_pk_parameters/run_covsearch work in more cases (#908)

• Make NONMEM .mod parsing work in more cases (#917, #975, #977)

• Make NONMEM .mod updating work in more cases (fd564168)

• Make NONMEM dataset column dropping work in more cases (088a046a)

• Make ODES updates work in more cases (c76fa476, 430f1d2e)

0.73.0 (2022-06-21)

New features

• Add covsearch tool

• Add function tools.retrieve_models to read in models from a tool database

• Add functions modeling.get_individual_parameters, modeling.get_pk_parameters, modeling.get_rv_parameter, and modeling.has_random_effect

Changes

• Include covsearch tool in AMD

• Add results for AMD tool

• Move fit, run_tool, run_amd, and all tool wrappers from modeling module to tool module

• Rename ‘diagonal’ -> ‘add_diagonal’ in iiv strategy option for iivsearch and modelsearch tool

• Include column for selection criteria in rank_models

0.72.0 (2022-06-08)

New features

• Add iovsearch tool

• Add function modeling.summarize_errors to get a summary dataframe of parsed errors from result files

• Add modeling.make_declarative

• Add modeling.cleanup_model

• Add modeling.greekify_model

Changes

• Use ‘no_add’, ‘diagonal’, ‘fullblock’, or ‘absorption_delay’ instead of numbers for iiv_strategy in iivsearch and modelsearch-tool

• Add results and documentation for allometry tool

• Add error summaries to iivsearch, modelsearch, resmod, and allometry tools

• Add algorithm argument in estmethod tool (‘exhaustive’ and ‘reduced’)

Bugfixes

• Handle etas after ODEs in iivsearch-tool

0.71.0 (2022-05-24)

New features

• Add functions find_clearance_parameters and find_volume_parameters

Changes

• Rename candidate models in estmethod tool

Bugfixes

• Add upper limit to VP parameter in modelsearch tool

• Fix issue with matrices not being considered positive semidefinite but considered positive definite

0.70.1 (2022-05-17)

Bugfixes

• Require pandas 1.4 or newer for multiindex joins. (Fixes #820)

0.70.0 (2022-05-13)

New features

• New tool allometry added

• Add modeling.summarize_individuals_count_table

• Add modeling.calculate_ucp_scale

• Add modeling.calculate_parameters_from_ucp

• Add description attribute to model objects

• Add wrappers for iivsearch and modelsearch tools (run_iivsearch and run_modelsearch)

• Add documentation for iivsearch tool

Changes

• resmod can now iterate and add multiple residual error models

• Automatically generate R examples in a seprate tab in documentation

• Merge iiv functions into one iivsearch algorithm (‘brute_force’)

• Use parameter names instead of eta names in iivsearch tool features

Bugfixes

• Make sure dropping of DATE columns in NONMEM models are handled correctly

• Solve issue with sporadic crashes because of a database race condition

• Solve issue with sporadic crashes caused by race in lazy parsing of NONMEM records

• Fix issues with converting some piecewise functions to NONMEM code correctly

• Fix issue with generating candidate models for iivsearch tool

• Fix issue with duplicate candidate models in iivsearch tool (#745)

0.69.0 (2022-04-29)

New features

• Add 1st and 3rd quantiles of residual to simeval results

Changes

• Rename mfl -> search_space in modelsearch and amd

• Use BIC as default ranking function in modelsearch

• Start model in modelsearch is not fitted

• Update modelsearch documentation

Bugfixes

• Fix bad odes when adding two peripheral compartments to model with MM elimination

• Fix bug in block splitting in IIV-tool (fixes #745)

0.68.0 (2022-04-27)

Bugfixes

• Fix bad odes when adding peripheral compartment to model with MM elimination (fixes #710)

0.67.0 (2022-04-25)

New features

• Add modeling.get_thetas, modeling.get_omegas and modeling.get_sigmas

• Add configuration option for NONMEM license file path

Bugfixes

• Correct parsing of ADVAN=ADVANx in $SUBROUTINES in NONMEM models

• Fix issue with duplicated TAD in $INPUT after add_time_after_dose

• Fix issue with not being able to use models with assignments in $DES in estmethod tool

• Set an upper limit for intercompartmental clearances in the modelsearch tool (fixes #695)

0.66.0 (2022-04-20)

Bugfixes

• Fix NONMEM model parsing issue causing ADVAN not to change for models with DEFOBS in $MODEL

0.65.0 (2022-04-14)

New features

• Add option in add_iiv and add_pk_iiv to choose initial estimate

Changes

• Replace different iiv-options in IIV-tool with iiv_strategy

• Use 0.01 as initial estimate for added IIVs in modelsearch tool

Bugfixes

• Add K-parameters in NONMEM model when changing to general linear (GL) solvers

0.64.0 (2022-04-12)

New features

• Add modeling.solve_ode_system

• Add documentation for .datainfo file

• Add iofv plot to linearize results

• Store tool meta data in metadata.json

Changes

• New options for modelsearch tool: switch order of mfl and algorithm, replace different iiv-options with iiv_strategy

0.63.0 (2022-04-07)

New features

• Support ~ as HOME in file paths input by users

• Add modeling.read_dataset_from_datainfo

• Store unique datasets for tool runs

Bugfixes

• Fix problem with TAD calculation for datasets with ADDL

• Handle LinAlgError when updating initial estimates in modelsearch (#656)

0.62.0 (2022-04-04)

New feature

• Store unique datasets in models/.datasets

Changes

• New name for final model in resmod

Bugfixes

• Use NaN in summary_individuals if tflite cannot be used

0.61.1 (2022-03-31)

Bugfixes

• Fix time after dose calculation for steady state dosing

• Fix issue where create_joint_distribution could create matrices that are not positively definite (#649)

• Keep IIV from MAT in MDT when adding a transit (#654)

0.61.0 (2022-03-29)

New features

• Add modeling.summarize_individuals

Changes

• Change initial estimates for QP1/QP2 ratio to 0.1/0.9

Bugfixes

• Handle ADDL columns for add_time_after_dose

0.59.0 (2022-03-25)

New features

• Add modeling.expand_additional_doses

0.58.4 (2022-03-24)

Bugfixes

• Fix issue with start model not being selected if no candidates are better in IIV- and modelsearch-tool

• Fix issue with ranking models by dBIC in IIV-tool

0.58.1 (2022-03-22)

Bugfixes

• Fix ordering of TAD values for dose at some time as observation

• Fix TAD values for datasets with reset time event

• Handle models with no covariates for predict_outliers and predict_influential_individuals

0.58.0 (2022-03-22)

New features

• Add modeling.add_pk_iiv to add iiv to all pk parameters of a model

Changes

• Change cutoff for zero protection in proportional error ModelSyntaxError

• Change to checking for positive semidefiniteness instead of only positive definiteness when validating omegas

Bugfixes

• Fix BIC-mixed calculation to not count thetas related to non-random etas (0 FIX) towards random parameters

• Read 0 FIX diagonal etas as random variables

0.57.0 (2022-03-21)

Bugfixes

• Keep thetas/etas when going across absorption transformations (#588, #625)

• Fix missing ALAG-parameter in non-linear elimination (#578)

• Fix issue with added VC1-parameter when adding transits to non-linear elimination (#577)

• Fix missing D1-parameter and RATE-column when adding zero order absorption to non-linear elimination (#578)

• Only do update_inits if start model was successful in IIV-tool (#632)

• Fix issue where etas where added to KA/K-parameters instead of MAT/MDT (#636)

0.56.0 (2022-03-17)

Changes

• Remove ZO elimination from the default search space in model search

• Do not apply resmod mode if no change on the full model

Bugfixes

• Fix bad calculation of number of observations for datasets with both EVID and MDV

• Properly handle observations and dose at same time for time after dose calculation

• Handle DATE column for time after dose calculation

• Handle NONMEM models with no ETAs

0.55.0 (2022-03-16)

New features

• Option in modeling.update_inits to move estimates that are close to boundary

Changes

• Set different initial estimates of clearance for peripherals (#590)

Bugfixes

• Fix issue with duplicated features with IIV-options in modelsearch-tool

• Fix issue where $MODEL was not added when setting ODE solver to GL or GL_REAL

• Fix issue where reduced_stepwise failed for certain search spaces (#616)

• Fix issue with reading in sampled_iofv in simeval

• Use the same time varying cutoff for resmod models and best model

0.54.0 (2022-03-08)

New features

• New IIV version of BIC in calculate_bic

• Use IIV BIC in iiv tool

• Add allometry step in amd tool

• Reduced stepwise algorithm in modelsearch

• Add cutoff option to predict_outliers etc

Bugfixes

• Fix issue with failing to creating correct subblocks of fullblock of random variables

• Set index name to id-name in predict_outliers

0.53.0 (2022-03-04)

New features

• Add option to remove specific IOV random variables

• Support Python 3.10

• Add modeling.check_dataset

Changes

• modeling.calculate_bic: Count epsilons interacting with etas to random parameters

• Updated tensorflow models for prediction of outliers and influential individuals

• Only consider parameters with etas for covariate modelbuilding in amd

• Include AIC/BIC in modeling.summarize_modelfit_results

Bugfixes

• Update solvers in estmethod-tool

• Handle Q-parameters when adding IIV on structural parameters in iiv-tool

• Only add IIV on MDT-parameter with add_mdt_iiv-option in modelsearch-tool

0.52.0 (2022-02-25)

New features

• Add covariate search to amd tool

0.51.0 (2022-02-24)

New features

• Add option to add IIV to start model in iiv-tool

• Add solver option in estmethod-tool

• Add option to add IIV only on MDT in modelsearch-tool

Changes

• modeling.calculate_bic can Calculate three different versions of the BIC

  default has switched to be a mixed effects version

• Remove etas instead of setting to 0 fix in iiv-tool

• Parse more errors and warnings in .lst-file

• Rename add_eta -> add_iivs, etas_as_fullblock -> iiv_as_fullblock, add_mdt_eta -> add_mdt_iiv in modelsearch

0.50.1 (2022-02-16)

Bugfixes

• Handle long paths when fitting NONMEM model

0.50.0 (2022-02-16)

New features

• Add modeling.write_results

• Add modeling.print_fit_summary

• Add modeling.remove_loq_data

• Add first version of WIP scm wrapper

Changes

• Change in mfl in modelsearch such that transits that don’t keep depot will have additional transit

• Make it possible to set $DATA directly for NONMEM models (via datainfo.path) (#130)

0.49.0 (2022-02-10)

New features

• Add modeling.calculate_se_from_cov

• Add modeling.calculate_se_from_inf

• Add modeling.calculate_corr_from_cov

• Add modeling.calculate_cov_from_inf

• Add modeling.calculate_cov_from_corrse

• Add modeling.calculate_inf_from_cov

• Add modeling.calculate_inf_from_corrse

• Add modeling.calculate_corr_from_inf

• Add modeling.create_report

• Add modeling.check_high_correlations

• Add modeling.calculate_bic

• Add modeling.check_parameters_near_bounds

• Add option to choose search space in AMD-tool

Changes

• Use p-value instead of OFV cutoff in resmod

Bugfixes

• Fix issue with no conversion to $DES for some models (#528)

0.48.0 (2022-02-04)

New features

• Parse estimation step runtime from NONMEM results file

Changes

• Force initial estimates when reading model file to be positive definite

Bugfixes

• Random block was not split properly in some cases when random variable was removed

• Add $COV correctly in NM-TRAN models (#457)

0.47.0 (2022-01-28)

• Add modeling.drop_columns

• Add modeling.drop_dropped_columns

• Add modeling.undrop_columns

• Add modeling.translate_nmtran_time

0.46.0 (2022-01-27)

• Add modeling.calculate_aic

• Add modeling.print_model_code

• Add modeling.has_michaelis_menten_elimination

• Add modeling.has_zero_order_elimination

• Add modeling.has_first_order_elimination

• Add modeling.has_mixed_mm_fo_elimination

• Add parent_model attribute to Model object

• Support non-linear elimination in search space in modelsearch tool

• Rename summary -> summary_tool in IIV and modelsearch tool, add summary_models

• Update modelsearch algorithm to only run 2C if previous model is 1C

• Fix bug in transformation order in features column of summary in modelsearch tool

0.45.0 (2022-01-21)

• Add timevarying models to resmod

0.44.0 (2022-01-20)

• Add modeling.create_symbol

• Add modeling.remove_unused_parameters_and_rvs

• Add modeling.mu_reference_model

• Add modeling.simplify_expression

• Add option keep_depot to modeling.set_transit_compartments

• Add CLI for estmethod tool

• Add attributes isample, niter, auto and keep_every_nth_iter to EstimationStep

• Remove stepwise algorithm in modelsearch tool

0.43.0 (2022-01-12)

• Add modeling.bump_model_number

• Fix regression in detection of dv column when synonym was used

0.42.0 (2022-01-11)

• Add modeling.get_doseid

• Add modeling.get_unit_of

• Add modeling.get_concentration_parameters_from_data

• Add modeling.write_csv

• Add modeling.resample_data

• Add modeling.omit_data

• Add modeling.get_observation_expression

• Add modeling.get_individual_prediction_expression

• Add modeling.get_population_prediction_expression

• Add modeling.evaluate_individual_prediction

• Add modeling.evaluate_population_prediction

• Add modeling.calculate_eta_gradient_expression

• Add modeling.calculate_epsilon_gradient_expression

• Add modeling.evaluate_eta_gradient

• Add modeling.evaluate_epsilon_gradient

• Add modeling.evaluate_weighted_residuals

• Support for Python 3.7 dropped

0.41.0 (2021-12-21)

• Add modeling.get_individuals

• Add modeling.get_baselines

• Add modeling.get_covariate_baselines

• Add modeling.get_doses

• Add modeling.list_time_varying_covariates

• Add combined error model to resmod

• Add option to zero_protect to set_proportional_error_model

• Add tool estmeth to find optimal estimation method for a model

• Fix bug causing resmod models to be incorrect

• New model.datainfo object

0.40.0 (2021-12-16)

• Add modeling.add_allometry

0.39.0 (2021-12-15)

• Add AMD and IIV tool and respective functions run_amd and run_iiv

• Add function add_covariance_step and remove_covariance_step

• Add method insert_after to ModelStatements

• Add option to set limit or no limit for power_on_ruv theta

• Rename EstimationMethod to EstimationStep and add EstimationSteps class

• Parse eta and epsilon derivatives from $TABLE

• Fix bug where lag time is removed when changing to ZO or FO absorption

0.38.0 (2021-12-08)

• Add function to get path to user configuration file

• Add function to get missing DVs

• Add option to add IIV on structural parameters (as diagonal and block)

• Add guard for log(0) in proportional error for log data

• Avoid crash if plots cannot be created in CDD results

• Fix issue saving modelsearch results

• Fix bipp issues with etas outside of FREM matrix

0.37.1 (2021-11-26)

• Fix bug causing frem report to crash with #IDs > 5000

• Fix bug for shifted uncertainty in frem with bipp

0.37.0 (2021-11-24)

• First version of IIV-tool

• Rename set_lag_time to add_lag_time

• Include run type in summarize_modelfit_results

• Fix bug with force option in write_model

• Fix bug in parsing .ext-files with tables without header

• Fix bug with nested update_source crashing due to incorrect handling of diagonal records

• Fix bug with inserted IGNORE on comment lines

0.36.0 (2021-11-11)

• Add option to set_dtbs_error_model to fix parameters to 0 (i.e. get data on log-scale)

• Create model file when fitting a model that has no model file

• Fix bug where files are missing during e.g. modelsearch

• Fix crash when including a model with no results in summarize_modelfit_results

• Fix bug in Pharmr where integers where interpreted as floats

• Fix issue with extra IPRED on power_on_ruv model

0.35.0 (2021-11-02)

• Option to include all estimation steps in summarize_modelfit_results

• Use kwargs in set_estimation_step and add_estimation_step

• First version of logger (via model.modelfit_results.log)

0.34.3 (2021-10-28)

• Let parametrization of peripheral compartment rates be kept if volume parameter can be found in the expression for K.

• Fix bug causing crashes when parsing some lst-files due to mixed encodings.

0.34.2 (2021-10-26)

• Fix broken parallelization for tools (workflows)

• Fix bug causing parsing of some NM-TRAN datasets to set a column index

0.34.1 (2021-10-20)

• Fix issues with retrieving results after model runs

0.34.0 (2021-10-14)

• Remove the need for update_source. Instead use model.model_code or modeling.generate_model_code(model)

• str(model) can no longer be used to get the model_code

• Fix crash in model database when using copies of models

0.33.0 (2021-10-11)

• Add modeling.read_model_from_database

• Add modeling.print_model_symbols

• Add modeling.append_estimation_step_options

• Fix crash for $DES models with RATE in dataset

• Fix estimation status for evaluation steps to use latest estimation

0.32.0 (2021-09-28)

• Move plot_iofv_vs_iofv to modeling

• Add modeling.get_observations

• Add modeling.plot_individual_predictions

0.31.0 (2021-09-21)

• Move parameter_sampling-functions into modeling module

• Add run_tool function to modeling

• Add predict_outliers, predict_influential_individuals and predict_influential_outliers functions to modeling

• Update API documentation (e.g. add examples, and improved index)

0.30.0 (2021-09-06)

• Add modeling.load_example_model

• Move eta_shrinkage results method to modeling.calculate_eta_shrinkage

• Add first version of resmod tool

• Update documentation (including API reference)

• Rename summarize_models to summarize_modelfit_results

• Fix bug related in running NONMEM on Windows via Rstudio

0.29.0 (2021-08-25)

• Rename zero_order_absorption to set_zero_order_absorption

• Rename first_order_absorption to set_first_order_absorption

• Rename bolus_absorption to set_bolus_absorption

• Rename seq_zo_fo_absorption to set_seq_zo_fo_absorption

• Rename have_zero_order_absorption to has_zero_order_absorption

• Rename power_on_ruv to set_power_on_ruv

• Rename add_lag_time to set_lag_time

• Move individual_shrinkage results method to modeling.calculate_individual_shrinkage

0.28.0 (2021-08-24)

• Move method individual_parameter_statistics from Results to a function in modeling and rename to calculate_individual_parameter_statistics

• Move method pk_parameters from Results to a function in modeling and rename to calculate_pk_parameters_statistics

• Rename create_rv_block to create_joint_distribution

• Rename split_rv_block to split_joint_distribution

• New default option force=True for write_model

• Rename ninds to get_number_of_individuals

• Rename nobs to get_number_of_observations

• Rename nobsi to get_number_of_observations_per_individual

• Rename remove_error to remove_error_model

• Rename additive_error to set_additive_error_model

• Rename proportional_error to set_proportional_error_model

• Rename combined_error to set_combined_error_model

• Rename has_additive_error to has_additive_error_model

• Rename has_proportional_error to has_proportional_error_model

• Rename has_combined_error to has_combined_error_model

• Rename theta_as_stdev to use_thetas_for_error_stdev

• Rename set_dtbs_error to set_dtbs_error_model

• Rename boxcox to transform_etas_boxcox

• Rename tdist to transform_etas_tdist

• Rename john_draper to transform_etas_john_draper

• Rename iiv_on_ruv to set_iiv_on_ruv

• Rename add_parameter to add_individual_parameter

• Rename first_order_elimination to set_first_order_elimination

• Rename zero_order_elimination to set_zero_order_elimination

• Rename michaelis_menten_elimination to set_michaelis_menten_elimination

• Rename mixed_mm_fo_elimination to mixed_mm_fo_elimination

• Function summarize_models to create a summary of models

• Parse total runtime

• Revert to dask distributed

0.27.0 (2021-08-09)

• Use dask threaded for Windows, allow configuration of dispatcher type

• Filter out individuals without observations in .phi-file

0.26.1 (2021-08-04)

• Correct residual calculation in simeval

• Correct how laplace estimation method is written

0.26.0 (2021-07-13)

• Add functions to set, add, and remove estimation step

• Add supported estimation methods (ITS, LAPLACE, IMPMAP, IMP, SAEM)

• When updating estimation step, old options are kept

0.25.1 (2021-07-08)

• Read site path if user path doesn’t exist (previously read user path)

• Change return type of covariates to a list for easier handling in R

0.25.0 (2021-06-24)

• Add modeling.ninds, nobs and nobsi to get number of individuals and observations of dataset

• Add reading results for resmod and crossval

• Add structural bias, simeval and resmod results to qa results

• Update index of cdd case_results to plain numbers

• Support line continuation (&) in NM-TRAN code

• Fix error in calculation of sdcorr form of parameter estimates

• Fix crash of cdd results retrieval

• Various fixes for running NONMEM models

0.24.0 (2021-05-25)

• Added theta_as_stdev, set_weighted_error_model and set_dtbs_error

• Error models can be added with log transformed DV using data_trans option

• Added model attributes data_transformation and observation_transformation

• Protected functions in NM-TRAN translated to Piecewise. Should now give the same result as when evalutated by NONMEM.

• Bugfixes for frem, scm and bootstrap results generation

• Rename model attribute dependent_variable_symbol to dependent_variable

• Added simplify method on Parameter class to simplify expressions given parameter constraints

0.23.4 (2021-05-03)

• 10-100 times Speedup of modeling.evaluate_expression

0.23.3 (2021-04-29)

• Documentation fix for pharmr release

• Handle implicit ELSE clauses for NM-TRAN IF

0.23.2 (2021-04-28)

• Fix bug #177

0.23.1 (2021-04-28)

• Bugfixes

0.23.0 (2021-04-28)

• Add function modeling.evaluate_expression

• Some documentation for modelfit_results

• Reworked interface to RandomVariables and Parameters

• Bugfixes

0.22.0 (2021-03-29)

• Support COM(n) in NONMEM abbreviated code

• Fix stdin handling issue when running NONMEM from R

0.21.0 (2021-03-22)

• New function read_results in modeling

• Add method to convert ExplicitODESystem to CompartmentalSystem

• Support running NONMEM 7.3 and 7.5

• Bugfixes:

  • Allow protected functions in NONMEM abbreviated code

  • Fix bad rates when changing number of transit compartments (#123)

0.20.1 (2021-03-11)

• Fix regression for calling NONMEM

0.20.0 (2021-03-11)

• New function modeling.set_peripheral_compartments

• New tool Model Search

• New model attribute estimation_steps to read and change $ESTIMATION

• Bugfixes (#99, #118)

0.19.0 (2021-03-02)

• Add create_result to create results from PsN

• Add documentation for covariate effects

0.18.0 (2021-03-01)

• Add functions to fix and unfix values to a specified value

• Add documentation for using Pharmpy with NONMEM models

• New execution system for modelfit

• Support for single string input for transformations of etas and epsilons (e.g. add_iov)

• Various bugfixes, including running NONMEM via Pharmpy on Windows

0.17.0 (2021-02-15)

• Add function to split an eta from a block structure

• New names for covariance between etas in create_rv_block

• Clearer error messages when adding IOVs (if only one level of occasion) and for parameter_names config

0.16.0 (2021-02-08)

• Improve initial estimates for adding peripheral compartments

• Parameter names are set according to priority in config

• Avoid duplication of e.g. median/mean when having multiple covariate effects with the same covariate

• Change assignments when multiple covariate effects are applied to the same parameter to be combined in one line

• Do not change error model if it is the same error model transformation multiple times

• Add AggregatedModelfitResults

• Document scm results

0.15.0 (2021-02-01)

• Change parameter_names config option to be a list of prioritized methods

• Option to read names from $ABBR for NONMEM models

• Add option to give parameter names to methods.add_iiv

• Add calculation of elimination half-life to one comp models in modelfit_results.pk_parameters

• Document cdd results

• Add set_initial_estimates, set_name and copy_model to modeling

• Allow single str as input to add_iiv and add_iov

0.14.0 (2021-01-25)

• Support reading $DES-records

• Add individual_parameter_statistics to ModelfitResults

• Add pk_parameters to ModelfitResults

• Add add_iov to modeling

• Rename add_etas -> add_iiv

0.13.0 (2021-01-18)

• Change names of covariate effect parameters for add_covariate_effects

• Improve ordering of terms in created NONMEM expressions

• Add parameter_inits, base_parameter_change, parameter_variability and coefficients to frem_results

• Add SimevalResults class

• Add fit and read_model_from_string functions to modeling

• Add solver attribute to ODESystem to be able to select ODE-system solver. Currently ADVANs for NONMEM

• New method nonfixed_inits to ParameterSet

• Add residuals attribute to ModelfitResults

• Various bug fixes

• Migrate to github actions for continuous integration

0.12.0 (2020-12-18)

• Add modeling.update_inits, modeling.add_peripheral_compartment and modeling.remove_peripheral_compartment

• Update FREM documentation

• Switch to using modelled covariate values for baselines in FREM

• Add methods for retrieving doses and Cmax, Tmax, Cmin and Tmin from dataset

• Various bugfixes and support for more ADVAN/TRANS combinations

0.11.0 (2020-11-20)

• Method df.pharmpy.observations to extract observations from dataframe

• Add ColumnTypes EVENT and DOSE

• Add model.to_base_model to convert model to its raw base model form

• New functions in modeling: remove_iiv, zero_order_elimination, comined_mm_fo_elimination and add_parameter

• Split modeling.absorption_rate and error into multiple functions

• Add calculations of AIC and BIC to ModelfitResults

• Improved pretty printing

0.10.0 (2020-11-16)

• modeling.create_rv_block

• modeling.michaelis_menten_elimination

• modeling.set_transit_compartments

• First version of modelfit method

• Add first version of bootstrap method

• Add parameter estimates histograms to bootstrap report

• Add automatic update of $SIZES PD when writing/updating NONMEM model

• Additions to QAResults

• NMTRanParseError replaced with ModelSyntaxError

• Multiple bugfixes to frem and scm result calculations

0.9.0 (2020-10-26)

• Add error_model function to the modeling module

• Added more standard models for modeling.add_etas

• Improve BootstrapResults

• Add plots to bootstrap

• Add support for the PHARMPYCONFIGPATH environment variable

• Add QAResults and LinearizeResults classes

• Bugfixes for some Windows specific issues

0.8.0 (2020-10-08)

• Add basic modeling functions to the modeling module

• modeling.add_etas

• Improved bootstrap results generation and additional plots

• Bugfix: Labelled OMEGAS could sometimes get wrong symbol names

0.7.0 (2020-09-28)

• Add method reset_indices in Results to flatten multiindices. Useful from R.

• absorption_rate can also set sequential zero first absorption

• New functionsadd_lag_time and remove_lag_time in modeling

• Add basic functions fix/unfix_parameter, update_source and read_model to modeling API

• Updated reading of NONMEM results

• Bugfixes in add_covariate_effects and absorption_rate

• Fix crash in FREM results if no log option could be found in meta.yaml

0.6.0 (2020-09-18)

• Add eta transformations: boxcox, t-dist and John Draper

• Add results cdd and scm to CLI

• Add different views for scm results

• Add support for taking parameter names from comment in NONMEM model

• Remove assumptions for symbols

• Add modeling.absorption_rate to set 0th or first order absorption

• Add update of $TABLE numbers

0.5.0 (2020-09-04)

• Many bugfixes and improvements to NONMEM code record parser

• Add calculation of symbolic and numeric eta and eps gradients, population and individulal prediction and wres for PRED models

• Add option to use comments in NONMEM parameter records as names for parameters

• Reading of ODE systems from NONMEM non-$DES models

• Calculation of compartmental matrix and ODE system

• New module ‘modeling’

• Function in modeling and CLI to change ADVAN implicit compartmental models to explicit $DES

• Function in modeling and CLI to add covariate effects

• Functions for reading cdd and scm results from PsN runs

• Many API updates

• Extended CLI documentation

0.4.0 (2020-07-24)

• Add categorical covariates to covariate effects plot in FREM

• Better support for reading NONMEM code statements (PK and PRED)

• Support for updating NONMEM code statements (PK and PRED)

• Bugfixes for CLI

0.3.0 (2020-06-16)

• New CLI command ‘data append’

• Parameter names is now the index in Parameters.summary()

• FREM postprocessing

• Standardized results.yaml and results.csv

0.2.0 (2020-03-27)

First release

0.1.0 (2018-07-22)

Initial library development/testing directory structure.