Open Samuel-herr opened 4 years ago
Hi Sam,
Thank you for using our code! We are continuing functions improvements of BRM. The visualization scripts of BRM results are not robust enough to share and we are still working on it. But one of the easiest ways is just to plot one chromosome at a time, by using the position (Pos. in result1.xls) as x, the LOESS fitted results (AFD-Expected in result11.xls) as y. For example, if we want to draw a plot of chromosome one:
chr <- read.table("configureExample/chr_length.tsv",sep = "\t")
tab <- read.table("result1.xls",sep = "\t")
idx <- which(tab[,1] == chr[1,1])
loc <- tab[idx,2]
val <- tab[idx,8]
points(loc, val, type="l", lwd=3, col="black")
As for the threshold too low, we need to check the parameters in configure file first. Consider the threshold is 0.0211, the pool size must be very big. Could you tell the approximate number of each pool? Furthermore, could you share the configure file (like BRM_conf.txt) so that we can figure it out?
When talking about the LOESS usage, let's see the document of it:
https://www.rdocumentation.org/packages/stats/versions/3.6.2/topics/loess
The "tri-cube weight function" is the default weighting strategy of LOESS. We set the span from 0.05 to 0.95, which means α <1, so we didn't show the weighting setting in our code.
Very appreciate for your asking, Huang
Hi Huang,
Thank you for the quick response and thoughtful answers.
With regards to my first question, your response made me realized I was graphing the wrong thing (AFD-observed instead of expected) so thank you for clarifying what to graph. My approximate pool size is 10000.
However, because my past mistakes I no longer think the threshold is a problem. As the graphs come out much closer to what I expected then before.
With regards to LOESS, that makes sense. I apologize for not researching further into LOESS before asking the question.
I am much happier with the results that I am currently recieving so thank you again for answering my questions they provided much clarity.
Best, Sam
Hello,
First off, thank you for the helpful paper and code! I found the paper best explained QTL mapping with BSA experiments out of the papers I have read similar to this.
I was wondering if you could share insight into how you visualized the output of BRM. When I try to use the output file (Result1) to graph the AFD using an adapted version of the QTL-seqR code, my plot is very noisy but looks similar to the gray region in Figure 6 from the paper. The theoretical threshold I get also seems low (0.0211, it is indigo line on the graph). I used the suggested mu value for Yeast given on this repository. Do you know of any reasons why the threshold could be so low?
I also have a question relating to the use of LOESS in your code. The BRM Paper suggests the use of a tri-cube smoothing function, however, in the code from what I see a tri-cube weighting function is not a default in the calling of LOESS. I was wondering if my understanding is correct that to implement a tri-cube weight function, I would modify the "locfit_by_loess" function in step 3 to include a weight based on a tri-cube function.
Thank you for taking the time to read these questions. Any advice will be greatly appreciated. If anything is confusing in my questions, please let me know.
Thanks again, Sam