CnrLwlss
commented
1 year ago Hi Miquel,
Thanks for your email. Always great to hear about the software being used! It was a treat to see these images and growth curves. Glad you are finding it useful.
I have a three things to say about differences between CFW and YPD:
1) We have found sometimes that the colour of the medium affects these photography-derived growth curves in subtle and surprising ways. Though I can't see any colour differences in your case. We usually include some kind of control strain on each plate in our analysis so that we have some basis to control for differences between medium types or between batches of the same medium (or even differences between plates: CSM thickness can vary, for example). Anyway, I suspect that's not the main issue here.
2) Overall, your strains are growing far less well in CFW. Many of the cultures didn't grow at all it seems. At least not to a size detectable by eye. Nevertheless, the cultures you have highlighted did grow. The difficulty now is that the remaining cultures experience a significantly different environment to their equivalents in the very full YPD plate. In 2017 I did some work to correct for this effect with a masters student: https://doi.org/10.1101/086835 However, I have never followed up.
3) You are right, sometimes cells of all kinds do better under a little bit of stress. There is this strange concept of hormesis that I wonder about sometimes: https://en.wikipedia.org/wiki/Hormesis Exactly what "a little bit" is depends on the genotype I suppose.
Your point ii) and 2) above are the same issue I think.
In principle, you could minimise this competition effect by finishing sampling the growth curve very early, 10 hours, for example and calculate growth rates based only on these partial curves. However, the blue and orange cultures are still ahead of the rest, even at that stage. It is worth noting that those two cultures are on the edge of the plate. Usually we would grow cultures around those edge positions but discount them from analysis (e.g. fill the edges with wild-type strains). We find that those edge positions have access to more nutrients and always grow significantly more quickly than those in the centre rows and columns.
Another way to address this effect is to grow many multiple plates (8 or more) with the positions of genotypes randomly shuffled (ideally only in the non-edge positions) so that all replicates have a chance of growing with and without competition effects. The average growth rate across all replicates might be more representative of the phenotype and genotype in that case.
This is a difficult problem. You have induced a nice strong phenotype with CFW, which is good. But by doing that, you have also induced some very strong competition effects, which are difficult to deal with. Another way to lessen this problem might be to decrease the concentration of CFW so that all strains grow up fairly well?
You could also abandon solid medium and grow these strains in liquid plates and measure growth curves with a spectrophotometer. However then you can have issues around mixing, flocculation, splashing and contamination.
I suspect that developing Daniel Boocock's approach where you fit a model of all growth curves PLUS the competition for available nutrients is ultimately the best way to deal with this situation. It is a really nice and interesting problem. Unfortunately I've moved away from microbiology entirely now and can't really justify going back to Daniel's work.
Not sure how much help this email has been, but I'd love to hear how you get on with this project.
All the best,
CONOR.
On Wed, 15 Nov 2023 at 16:07, MikiSchikora @.***> wrote:
Good afternoon,
First of all, thanks for developing this tool, it is being very useful! I found something strange in my analysis, and I wonder if you can help me understand it. I did an experiment with different yeast strains growing in YPD and Calcofluor White (CFW). Surprisingly, I find that some species reach higher growth in CFW than in YPD. This is an example of the output I am getting:
[image: growth_curves_and_images_CFW_Cg-wt1] https://user-images.githubusercontent.com/30113759/283170360-0e3e59e3-5406-46bf-9f0f-3aba33caa92c.png
The left plate is rich medium (YPD) and the right is CFW. The growth curves are from 8 technical replicates of Cg-wt1 (Candida glabrata). The images below refer to these 8 replicates in different timepoints. As expected, most yeasts do not grow in CFW. However, Cg-wt1 grows slower, but reaches a higher growth at the end. This is puzzling because we'd expect YPD to be maximum growth. I have seen this in other stressors as well for different yeast species. I wonder if this is i) a true biological observation (i.e. some yeasts grow better on CFW), ii) a biological observation that is an artifact of the plate-based design (i.e. perhaps the higher number of spots growing in the YPD plate generate nutrient deprivation, which results in lower final growth) or iii) a technical artifact of how I measure cell density (or growth). This is not something that I can clarify by looking at the plates, as there is no visual differences between the YPD and CFW growing spots.
For reference, this is how I got this data:
1.
I ran colonyzer with colonyzer --lc --diffms --greenlab --plots --remove --initpos --fmt 96 2.
I loaded the colonyzer data into R with data_colonyzer = colonyzer.read() from qfa package. 3.
I measured growth (Cell Density in the plots) as (data_colonyzer$Trimmed/(data_colonyzer$Tile.Dimensions.Xdata_colonyzer$Tile.Dimensions.Y255))
- 1e7. Note that 1e7 is just a scaling factor to have a value between ~0-1.
I noticed that changing most analysis parameters (i.e. not using --diffms, and/or --greenlab, or using --edgemask in addition to --lc --diffms --greenlab) did not affect this general observation. Similarly, enhancing the contrast of the images also did not change the result. Conversely, I found that reducing --slopefill (i.e. to 0.5 instead of the default 0.9) reduced a bit the differences, but CFW spots had still higher growth.
What would you say is going on? For example, do you think that it is a good idea to use a different --slopefill parameter (I am not sure what it is doing)? I ask because perhaps you also saw something like this in the past.
Thanks for your time!
Best,
Miquel Àngel Schikora Tamarit Barcelona Supercomputing Center
— Reply to this email directly, view it on GitHub https://github.com/CnrLwlss/Colonyzer/issues/6, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAJNHVKAFRWUKG65KWU2N5DYETSERAVCNFSM6AAAAAA7MXRQZGVHI2DSMVQWIX3LMV43ASLTON2WKOZRHE4TKMBYGA2DAMQ . You are receiving this because you are subscribed to this thread.Message ID: @.***>
MikiSchikora
Good afternoon,
First of all, thanks for developing this tool, it is being very useful! I found something strange in my analysis, and I wonder if you can help me understand it. I did an experiment with different yeast strains growing in YPD and Calcofluor White (CFW). Surprisingly, I find that some species reach higher growth in CFW than in YPD. This is an example of the output I am getting:
The left plate is rich medium (YPD) and the right is CFW. The growth curves are from 8 technical replicates of Cg-wt1 (Candida glabrata). The images below refer to these 8 replicates in different timepoints. As expected, most yeasts do not grow in CFW. However, Cg-wt1 grows slower, but reaches a higher growth at the end. This is puzzling because we'd expect YPD to be maximum growth. I have seen this in other stressors as well for different yeast species. I wonder if this is i) a true biological observation (i.e. some yeasts grow better on CFW), ii) a biological observation that is an artifact of the plate-based design (i.e. perhaps the higher number of spots growing in the YPD plate generate nutrient deprivation, which results in lower final growth) or iii) a technical artifact of how I measure cell density (or growth). This is not something that I can clarify by looking at the plates, as there is no visual differences between the YPD and CFW growing spots.
For reference, this is how I got this data:
1) I ran colonyzer with
colonyzer --lc --diffms --greenlab --plots --remove --initpos --fmt 962) I loaded the colonyzer data into R with
data_colonyzer = colonyzer.read()from qfa package.3) I measured growth (Cell Density in the plots) as
(data_colonyzer$Trimmed/(data_colonyzer$Tile.Dimensions.X*data_colonyzer$Tile.Dimensions.Y*255)) * 1e7. Note that 1e7 is just a scaling factor to have a value between ~0-1.I noticed that changing most analysis parameters (i.e. not using
--diffms, and/or--greenlab, or using--edgemaskin addition to--lc --diffms --greenlab) did not affect this general observation. Similarly, enhancing the contrast of the images also did not change the result. Conversely, I found that reducing--slopefill(i.e. to 0.5 instead of the default 0.9) reduced a bit the differences, but CFW spots had still higher growth.What would you say is going on? For example, do you think that it is a good idea to use a different
--slopefillparameter (I am not sure what it is doing)? I ask because perhaps you also saw something like this in the past.Thanks for your time!
Best,
Miquel Àngel Schikora Tamarit Barcelona Supercomputing Center