GitHub

Utility functions

AbstractFBCModels export several utilities for easier definition of the model contents; here we explore some of them in individual sections.

import AbstractFBCModels as A

Which accessors should I overload?

You can query the current set of accessor functions as follows:

A.accessors()
38-element Vector{Method}:

You do not need to overload all of them (e.g., if you model does not have any genes you can completely omit all gene-related functions). The main required ones (esp. the reaction- and metabolite-related ones) will throw an error if not implemented, and the automated tests will fail with them.

Why are there 2 representations of gene-reaction associations?

Gene-reaction association is, in general, a Boolean function – you give it a list of genes that are available (or, symmetrically, knocked out) and a reaction, and it tells you whether the reaction can work or not. This representation is captured in accessor reaction_gene_products_available.

Some model construction methods, on the other hand, require a relatively streamlined view of the gene-reaction associations, roughly describing "ways in which several possible enzymes that catalyze the reaction can be assembled from individual gene products" – these are sometimes dubbed "isozymes". This view is isomorphic to a Boolean formula in a disjunctive normal form (DNF) (put simply, a DNF would be "a big OR of ANDed identifiers", such as (a && b && c) || (a && d)). The DNF view of the gene association is accessed via reaction_gene_association_dnf.

Both representations have pros and cons; the main can be summarized as follows:

  • DNF may not be able to represent some valid boolean formulas without an exponential explosion in size – in particular, the formula (a1||b1) && (a2||b2) && (a3||b3) && ... && (aN||bN) will explode to roughly 2^N terms in DNF.
  • There's no simple universal format to store and exchange the descriptions of general boolean functions; we thus do not want to enforce a single one.

Your model is primarily supposed to overload the reaction_gene_products_available accessor. If you are able to reliably convert your Boolean functions to DNF, you should also overload the reaction_gene_association_dnf.

In some cases when the model directly stores the DNF form, you may also utilize the reaction_gene_products_available_from_dnf function to easily implement the reactiongeneproducts_available` by just forwarding the arguments.

Downloading models for testing

For reproducibility, it is often viable to check downloaded files for validity, using e.g. checksums. Since this is a common operation, we provide download_data_file, which is an appropriate wrapper for Downloads.download:

mktempdir() do dir
    origin = joinpath(dir, "origin")
    url = "file://$origin"
    dest = joinpath(dir, "model")
    open(origin, "w") do f
        write(f, "hello")
    end
    A.download_data_file(
        url,
        dest,
        "2cf24dba5fb0a30e26e83b2ac5b9e29e1b161e5c1fa7425e73043362938b9824",
    )
end
"/tmp/jl_2CJqBq/model"

This page was generated using Literate.jl.