Previous fix didn't cut it and I jumped the gun on the release. This works.
Fixed recombination bug where one of the two progeny genomes was being lost (thanks David Suárez!). Updated joblib version.
Revisions incorporated!
Change all mentions of "lethality" to "mortality".
Update Pillow to 9.0.1
Modify behavior of compositionPlot to take all hosts/vectors into account when plotting immunity/protection.
Change units of lethality_rate_host and lethality_rate_vector to be a rate like other parameters for better internal consistency, instead of a fraction of recovered cases. Affects only behavior of simulations with disease-caused mortality. Updated example accordingly.
Fix bug in updateVectorCoefficients() specific to natality and migration. Does not affect any simulation results as long as natality and migration are not functions of pathogen genome sequence.
Update compositionPlot additional arguments.
Update Pillow to 9.0.0.
Added biorXiv links.
Changed computation of inter-population contact rates to match logic of intra-population contact rates. Does not affect outcome of simulations not using inter-population contact.
Parenthesis error.
Another small change to the Gillespie algorithm, this time to avoid rare infinite loops when tampering with t_var.
- changed order in which time delta was added to t_var to be after interventions occur
- do not carry out interventions if t_var is past simulation end time
Same bug as below, the fix was incomplete.
There was a bug in the last release regarding interventions! in order to correctly impplement custom user killswitches, it is important to update the time variable t_var immediately before an intervention takes place, not after it. Fixed in Gillespie method now, does not change behavior of any previous simulations.
General simulation structure changes:
- changed handling of intra-population contact rate between hosts and vectors within the Gillespie function to fit a biting rate definition for the contact rate (i.e. constant per vector contact rate). Does not affect behavior of simulations if total number of hosts and vectors is equal (or if models have no vector-borne pathogens)
- made time variable t_var a property of Model objects instead of internal to Gillespie object, allowing users to modify simulation time (e.g. for killswitches); does not affect simulation results
Opqua structure changes:
- changed way interventions are executed to guarantee that the Model object being simulated carries the intervention upon itself
- changed runReplicates() to create independent copies of model object and run simulations on each (should not affect simulation results due to parallelization through joblib)
- added deepCopy() to reassign all internal model and population references in copied model objects
- changed runReplicates() and runParamSweep() to use deepCopy() (should not affect simulation results due to parallelization through joblib)
- added customModelFunction() function to allow users to add custom methods to specific Model instances (e.g. for killswitches and conditional interventions)
Opqua syntax changes:
- changed syntax of interventions to force all functions used to be methods of the Model object being intervened
All graphs in publication (title pending) generated with this stable version. [RETROACTIVE EDIT: first draft only]
General model structure changes:
- added transmission_efficiency_host_host, transmission_efficiency_host_vector, transmission_efficiency_vector_host as additional parameters
- made global_trackers copy into history object of a model
- adjusted computation of recombination probabilities for hosts and vectors
- skip recombining when parental genomes are the same
- changed genome sampling during inoculation of hosts and vectors
- added more flexibility, changed structure, and debugged migration/population contact options in runParamSweep()
In compositionDf():
- remove missing data
- change algorithm to replace combinations of genomes (more efficient when combinations are limiting factor)
- allow user to specify genome plotting order
Miscellaneous:
- add option to plot population fractions instead of absolute counts in compositionPlot()
- Gillespie algorithm now prints out event name rather than ID number
- changed error handling when adding pathogens to hosts and vectors
- removed a duplicate definition in Model newSetup()
- changed a group name in intervention_example.py
v0.2.5 created a major bug that escaped my attention with the division by zero error fix.
- corrected Host and Vector acquirePathogen() functions to restore correct behavior
- added a requirements.txt file purely for reference purposes in case a future dependency update breaks opqua
- added global_trackers dictionary to Model in order to track and return some global indicators when running replicates or parameter sweeps
- added addCustomConditionTracker() function to Model class in order to allow users to track custom events in model
- modified mutation() and recombination() in Vector and Host classes as well as addPathogensToHosts() and addPathogensToVectors() in Population class in order to track genomes seed for global_trackers
- added model attribute to Population class for the above reason as well
- fix way host and vector sampling is handled (apparently I forgot to actually implement it in Gillespie and Population classes after I added the arguments, haha)
- minor bug fix: modify acquirePathogen in Vector and Host classes to avoid division by zero errors when recalculating sum_fitness after it was zero
- corrected name of compositionDataframe argument in compositionPlot() to composition_dataframe
- fixed regex processing bug in compositionDf()
- added **kwargs argument passing to joblib functions to allow user to change backend and stuff
Trying to deploy on cluster so bear with me on the updates here
- added parameter sweep function runParamSweep()
- added id property and argument to Setup() in order to associate a Setup to its ID in a Model, so that runParamSweep() can edit the setups
- added getCompositionData() function to Model class to allow user output composition data without plotting compartments
- fixed bug in how runReplicates() computed and return output
- added verbose optional argument to saveToDf() to reduce console output
- added composition_dataframe optional argument to allow for pre-computed data
- added setup.cfg as per Joel Barmettler I guess?
- change compositionPLot remove_legend behavior to fix bug
- change pathogenDistanceDf seq_names behavior to fix bug
- reduce mean inoculum from hosts into vectors to reflect malaria cycle
- modify infectHost and infectVector inoculation behavior so that mean_inoculum does not affect overall transmission rate; each infection now results in at least 1 pathogen transfer (if not containing and not immune to the pathogen genome sampled)
- update version tags
- pathogen genome influences transmission, death, recovery, migration, mutation probabilities. Done, tested
- independent recombination of alleles -> make chromosome separators!! make reassortment parameter option as related but separate to recombination. Done, tested
- host/vector birth/death rates in populations -> make birth event. Done, tested
- separate natural death into an event that doesn't log deaths. Done, tested
- make RECEIVE_CONTACT and RECEIVE_POPULATION_CONTACT coefficient columns in arrays, modify how these events are handled. Done, tested
- migrate vectors. Done, test
- contact between populations (without migration). Done, tested
- make genome labels optional, if no labels write a file with the genomes in the same order for compositionPlot. Done, test
- compositionPlot – make custom groupings, eg. "genomes containing sequence AAA". Done, test
- make option to count only 1 fitness-dominant strain/host in compositionPlot. Done, test
- genomes and dates output for TDA. Done, test
- set seed. Done, test
- make all events except mutation, recombination and recovery 1-coefficient like contact and pop contact are. this way, coefficients are fraction of healthy rate. Modify dummy rows in coeff arrays to match this. Done, tested
- parallelizeable simulations. Done, test
- try Numba, Cython, JAX optimization
- update docs. Done
- update tutorials. Done, tested
- correctly update arguments in function documentation and README for compositionPlot family functions