Clarify provenance of models in benchmarks.jl

[sumiya11]

Currently some models have comments containing links to sources, but this info is not systematic and may be not convenient to use.

So, I asked GPT-5:

Please create a .bib entry for each model in benchmarks.jl. Some useful sources are:

• https://github.com/Xabo-RB/Local-Global-Models
• https://github.com/alexeyovchinnikov/SIAN-Julia
• https://ieeexplore.ieee.org/document/10160017 (Table 1)

You may cross-reference the models by their names and equations. You may search on the internet. Make sure to verify DOIs and URLs. Only create a .bib entry if you are certain that it is correct.

It produced benchmarking/bib/. Note: some entries are not correct. My plan currently is to go model by model, verify the .bib file, and then add it in the :cite field in benchmarks.jl, as below:

Example

:JAK_STAT => Dict(
    # https://arxiv.org/pdf/2207.09745.pdf
    # https://github.com/Xabo-RB/Local-Global-Models/blob/main/Models/Cellular%20signalling/JAKSTAT1.jl
    :name => "JAK-STAT 1",
    :ode => @ODEmodel(...),
    :cite => "bib/JAK_STAT_1.bib"  # <-- NEW
),

Design

• In the example above, there is some potentially useful info in the comment. Maybe these could be incorporated in the :cite field somehow.
• In the same vein: allow several references per model? allow custom notes? In the case when we can pinpoint the precise equation/theorem in the paper, it would be nice to specify it.
• Many small .bib files or one large file?

Progress

Model Verification Checklist

• Modified LV for testing
• SIWR original
• SIWR with extra output
• Pharm
• SEAIJRC Covid model
• MAPK model (5 outputs)
• MAPK model (5 outputs bis)
• MAPK model (6 outputs)
• Goodwin oscillator
• HIV
• SIRS forced
• Akt pathway
• CD8 T cell differentiation
• Chemical reaction network
• QWWC
• SLIQR
• Fujita
• LLW1987_io
• Bilirubin2_io
• HIV2_io
• Biohydrogenation_io
• Treatment_io
• SEIR_1_io
• JAK-STAT 1
• SIR 24
• SIR 21
• SIR 19
• SIR 6
• SEIR 34
• SEIR 36 ref
• TumorHu2019
• TumorPillis2007
• SEIR2T
• SEIRT
• SEUIR
• Bruno2016
• Transfection_4State
• p53
• Crauste_SI
• HighDimNonLin
• KD1999
• CGV1990
• LeukaemiaLeon2021
• Ruminal lipolysis
• cLV1 (2o)
• cLV1 (1o)
• generalizedLoktaVolterra (1o)
• Pivastatin
• Linear_compartment_hard_1
• Linear_compartment_hard_2
• Covid model (Gevertz et al)
• Ovarian follicle population dynamics
• Immune response to influenza (MB1 model)
• Immune response to influenza (MB2 model)
• Immune response to influenza (MB3 model)
• Immune response to influenza (MB4 model)
• Immune response to influenza (MD1 model)
• Immune response to influenza (MD2 model)
• Immune response to influenza (MD3 model)
• Immune response to influenza (MD4 model)
• EIHRD epidemiological model
• NFkB

Notes

The AI fared fairly well. For example, it recognized that Akt and Fujita is actually the same model. It did not create entries for QY and St because it could not verify their origin (rightfully so).