Growth models to integrate in the μGrowthDB GUI

[hariszaf]

@AndrewRadev @hallucigenia-sparsa
Based on a relatively recent book chapter on microbial growth models Mahdinia et al. (2020), here is a short list of the models mentioned:

• Gompertz model (Jefferies and Brain, 1984)
• Logistic model (Jason, 1983)
• The Rosso model (Rosso et al., 1993)
• Baranyi and Roberts model (Baranyi et al., 1993)
• Monod model (Monod, 1949)
• Compartmental model (Vanier and Bower, 1999)
• Weibull model (Farewell, 1982)

Also, from another book chapter by Bassoto Xavier et al. (2023)
], I found out about the

• Buchanan Model (Buchanan et al., 1994)

Last, I found a bit earlier chapter by Pérez-Rodríguez & Valero, in which I appreciated its organisation and I think brings some extra models too:

• McKellar model (a simplified version of the Baranyi model and has the same parameters)

Inactivation Models
Models that try to describe microbial inactivation,i.e. how a dense population goes extinct.
The Weibull model is an inactivation model.

• Bigelow Model (Linear Model)
• Shoulder/Tail Models

Please, feel free to suggest any missing or point out if a model should not be included.

🎗️ we call primary models those aiming to describe the four phases of a typical microbial population growth, while secondary are those that try to describe the effect of environmental conditions (i.e., physicochemical and biological factors) on the values of the parameters of a primary model.

[hariszaf]

I also found out about dashing growth curves.
This web-based tool gets as input an excel file to return growth curves and other growth stats.
(in their video they show all tool's features).

From their repo we could use a thing or two.

[hallucigenia-sparsa]

Thanks for the list. As discussed, we can keep it simple and refine ourselves to 3-4 growth models supported by growthrates (including Baranyi because it deals with lag phases, logistic because it's a classic and standard linear fit to log phase because it's widely used) unless there are good reasons to support additional growth models. If dashing growth curves or another software is easier to integrate, then of course we can use it rather than the growthrates R package. Please also have a look at the comparison of growth models by Ana-Hermina, Loic and colleagues. There's also the question whether to fit models once upon data upload and store model parameters or whether to re-fit them on the fly.

[hallucigenia-sparsa]

growthrates also reports indicators for the quality of the model fit, which we could store together with the parameters if we decide to store those. It's probably best to fit growth models to each biological replicate separately, so it's possible later on to visualise the model fit for each replicate and to report standard deviations on parameters across replicates. For me, the main motivation of model fitting here is to get a rough idea of growth rates & carrying capacity, so we can display interaction strengths on the fly when appropriate. If users want to go deeper, they can download the data.

[hariszaf]

Hi @hallucigenia-sparsa and @AndrewRadev
I found out about Kinbiont and I think it would be worthwhile to take a look.

Here is their preprint

[hallucigenia-sparsa]

Thanks for sharing; Kinbiont (and the group behind) is definitely good to know about. Kinbiont looks like it's equivalent to the growthrates R package, so the choice can be one of convenience (whichever is easier to work with).