General : is SLiM able to simulate plant breeding ?

[Augustin Desprez]

Dear Ben Haller,
    Dear all,

Currently working on my PhD studying diversity and haplotypes transmission in plant breeding, I am using markers data on a few thousands of plants covering around two decades (20 generations) of plant breeding. Therefore, I have access to genotypic data of a good amount of founders and their progeny. My questions are not exactly in the same time scale as ecology and evolutionary mechanism, and I know some tools are not always easy to adapt, that's why before diving into SLiM, I need the feedback of users. If I am not wrong, SLiM is able to simulate genome and their evolution, which is the main thing I am interested in.

In order to give a proper meaning to the data that I can observe, (i.e. diversity, effective population size, Haplotype size...), I was wondering if I could use simulations from SliM to answer several questions :
1- What demographic scenario (pop size, number of generations, selection pressure...) is best fitting the data I am observing now, when starting from founders (i.e. 10 to 20 founders) ?
Here SLiM mission could be to start from founders and try many combinations to create the result of 20 years of crosses and see witch model(s) are most probably providing the same data that I am able to observe today.

2- Comparing my data from data created without hard selective pressure (only drift for example), which genomic regions are the farthest from this neutral conditions ? And what may have led to be that far (founder effet, selection pressure, recombinations...) ?
Here I thought SLiM mission could to create genotypes obtained with neutral models starting from known founders.

3- Could SLiM be able to recreate pedigrees with founders and progeny (even if not all founders are genotyped) ?
Then, we could be able to know how actual plants could have been obtained with the least amount of generations, and compare if breeding methods were the most efficient or not (efficient could be in generations as making progress fast is a big issue, but could also be regarding having the highest result with few crosses between plants).

Thank you for reading and I would be pleased that you enlighten me on this subject,

Sincerely,

Augustin

[Augustin Desprez]

Gregor Gorjanc replied (on another page, so I transfer his response here) : 

Hi Augustin, 

SLiM is a fantastic software for stochastic simulation in genetics, and could in principle do what you are asking for. I am quite interested in how a SLiM script for a plant breeding programme would look like! We have developed an R package AlphaSimR dedicated for breeding simulations - see https://cran.r-project.org/web/packages/AlphaSimR/index.html, which is probably more suited to what you are asking for;) 

 With regards, 

 Gregor

[Augustin Desprez]

Thank you  Gregor for your answer, I was indeed hesitating with using AlphaSimR. 

As time is running fast, it will be hard to try both unfortunately, but I will dig further before choosing between AlphaSim and SLiM.

The first one, which I am thankful you highly contributed, seemed to be more adapted for breeding programs "by default", whereas SLiM may represent more work to get performing in those conditions however seems to be extremely adaptative when masterfully used. Is it also what you would think of it, knowing those tools better than I do ?

Best regards,

Augustin

[Ben Haller]

Hi Augustin!  I think that's the tradeoff, yes.  AlphaSimR will be much smarter about the details of running a breeding program specifically, because it's designed for that.  Managing the pedigree, running different demographic scenarios, etc. – it's built for that kind of thing.  SLiM is a general-purpose eco-evolutionary simulator, and as such it is capable of those things and lots of other things, but it is not specifically designed to run breeding experiments, and so getting it to do what you want it to do will require much more work.  I would suggest skimming through the AlphaSimR doc to see what it is capable of, and using it if it can do everything you need; you'll get up and running quickly.  If not, using SLiM instead is probably necessary; in that case, I recommend that you begin not by trying to read the manual (which is almost 900 pages, not even counting the separate manual for its scripting language, Eidos), but rather by doing the SLiM Workshop, which you can find online at the SLiM home page, and can download and follow for yourself quite easily, for free.  That will give you a relatively quick introduction to the basics of SLiM and Eidos, although it doesn't cover nearly everything that is in the manuals.  Good luck!

Cheers,
-B.

Benjamin C. Haller
Messer Lab
Cornell University

[Augustin Desprez]

Thank you very much for your answer, I will look into it.

I may do a feedback in a few weeks/months here, as it could be useful for everybody.

Cheers,

Augustin D.