adapting epigrowthfit to fit microbial growth curves

[bbolker]

Ghenu et al (2022) evaluated methods of fitting microbial growth curves.

Based on a conversation with the lead author, epigrowthfit could probably contribute a lot here if we:

• make it possible to match observations with $F(t)$ rather than $F(t+1) - F(t)$ (i.e., treating $F$ as the observed outcome rather than as the cumulative sum/integral of the observed outcome
• implementing continuous response/observation distributions (e.g. Normal, log-Normal, Gamma) [and removing any hard-coded tests that throw an error if the observed value is non-integer)
• possibly implement some other standard growth curves used in this field (e.g. Gompertz)
• (maybe?) allow for heteroscedasticity, or allow for a link function (or something else sensible that can account for a pattern of increasing variability around, say, a logistic function as the population size increases)
• think carefully about how we're treating baseline values, and whether we need to do something different
• censoring for below-detection observations? allowing for 'lag phase' (Bruslind 2018)?

Bruslind, Linda. 2018. “Microbial Growth.” In Microbiology. LibreTexts Biology. https://bio.libretexts.org/Bookshelves/Microbiology/Microbiology_(Bruslind)/09%3A_Microbial_Growth.

Ghenu, Ana-Hermina, Loïc Marrec, and Claudia Bank. 2022. “Challenges and Pitfalls of Inferring Microbial Growth Rates from Lab Cultures.” bioRxiv. https://doi.org/10.1101/2022.06.24.497412. [also in press at/coming soon to https://www.frontiersin.org/articles/10.3389/fevo.2023.1313500/abstract ]

[davidearn]

It would certainly be great to broaden the scope of epigrowthfit so it can be used for microbial growth curves. It would be good to have a prioritized major task list that we can all see. I'll start a new issue about that.

[bbolker]

FWIW the first few bullet points should be fairly straightforward/low-hanging fruit, although I'm sure that if we actually got into all of the details of/peculiarities of microbial growth curve data we would find that there is also harder stuff.

[dushoff]

Not a huge fan of censoring. Also, it took me a while to figure out that the Bruslind cite was only supposed to be about the lag phase... maybe separate those two bullet points. Ghenu is a big fan of lag phase.

…

On Mon, Jan 8, 2024 at 3:49 PM Ben Bolker ***@***.***> wrote: Caution: External email. Ghenu et al (2022) evaluated methods of fitting microbial growth curves. Based on a conversation with the lead author, epigrowthfit could probably contribute a lot here if we: - make it possible to match observations with $F(t)$ rather than $F(t+1) - F(t)$ (i.e., treating $F$ as the observed outcome rather than as the cumulative sum/integral of the observed outcome - implementing continuous response/observation distributions (e.g. Normal, log-Normal, Gamma) [and removing any hard-coded tests that throw an error if the observed value is non-integer) - (maybe?) allow for heteroscedasticity, *or* allow for a link function (or something else sensible that can account for a pattern of increasing variability around, say, a logistic function as the population size increases) - think carefully about how we're treating baseline values, and whether we need to do something different - censoring for below-detection observations? allowing for 'lag phase' (Bruslind 2018)? Bruslind, Linda. 2018. “Microbial Growth.” In Microbiology. LibreTexts Biology. https://bio.libretexts.org/Bookshelves/Microbiology/Microbiology_(Bruslind)/09%3A_Microbial_Growth . Ghenu, Ana-Hermina, Loïc Marrec, and Claudia Bank. 2022. “Challenges and Pitfalls of Inferring Microbial Growth Rates from Lab Cultures.” bioRxiv. https://doi.org/10.1101/2022.06.24.497412. [also in press at/coming soon to https://www.frontiersin.org/articles/10.3389/fevo.2023.1313500/abstract ] — Reply to this email directly, view it on GitHub <#7>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAW5E6OCDYKPDN3KB6EADJDYNRLWJAVCNFSM6AAAAABBSBTRWOVHI2DSMVQWIX3LMV43ASLTON2WKOZSGA3TCMJZHA3TQOA> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

-- Jonathan Dushoff (https://tinyurl.com/jd-pronouns) McMaster University Department of Biology https://mac-theobio.github.io/ http://jd-mathbio.blogspot.com/

[jaganmn]

Thanks. Yes, some low-hanging fruit here. @davidearn unless you've already started, I'll make a TODO (*.org, unless people would hate that) tomorrow.

[davidearn]

I scribbled the following, intending to think further before posting, but I'm happy to let you make your TODO.

This is intended to be a big picture list, to help focus our eyes on the prizes...

• get epigrowthfit up on CRAN
• write and submit a first publication applying epigrowthfit to COVID data
• write an R journal or PLoS Comp Biol paper about the package/methodology
• decide what other paper(s) we want to write applying epigrowthfit to epidemic data
• implement changes required to apply epigrowthfit to microbial growth data
• revise/expand this priority list