GitHub

Install

Requirements

  • Julia 1.11 or newer.
  • Kinbiont.jl 1.3+ (BayesBiont reuses Kinbiont's GrowthData, ModelSpec, and MODEL_REGISTRY).
  • Turing.jl + Distributions.jl + MCMCChains.jl (pulled in automatically).

Install from the General registry

using Pkg
Pkg.add("BayesBiont")

This will install BayesBiont, Kinbiont, Turing, and the rest of the dependency tree.

Optional: SciMLSensitivity for ReverseDiff on ODE models

BayesBiont supports adbackend=:reversediff for hierarchical fits. For ODE models specifically, this requires SciMLSensitivity.jl to be loaded in your session because reverse-mode AD through SciML ODE solvers needs adjoint methods. It is not a hard dependency of BayesBiont (its Enzyme constraints can conflict with other packages in some configurations).

To opt in:

using Pkg
Pkg.add("SciMLSensitivity")
using SciMLSensitivity
using BayesBiont
# Now `BayesFitOptions(adbackend=:reversediff)` works on ODE models too.

If you try ODE + :reversediff without SciMLSensitivity loaded, BayesBiont raises an ArgumentError with a clear hint.

For NL closed-form models (logistic, Gompertz, Richards, etc.) :reversediff works out of the box, no SciMLSensitivity needed.

Development install

To work against a local development tree (e.g. to modify BayesBiont source or to track an unreleased Kinbiont):

using Pkg
Pkg.develop(path="path/to/BayesBiont.jl")
# Optional: also develop Kinbiont
Pkg.develop(path="path/to/Kinbiont.jl")

Verify the install

using BayesBiont, Kinbiont
@assert isdefined(BayesBiont, :bayesfit)
@assert haskey(Kinbiont.MODEL_REGISTRY, "NL_Gompertz")