joey711
commented
10 years ago Yes, there are lots of ways to import tab-delimited tables of values into R. Some very quick googling will give you lots of options. In your case I suggest read.table, and you may need to fuss with a few of the default parameters to let it know that your file does have both row and column headers, etc.
Once you have imported the file, you will need to coerce it to a matrix if it isn't already, and then to a "dist" (distance matrix) object. From your photo of a portion of the matrix, it appears that you have it stored as a full symmetric pairwise square. Here is the line of code for converting the table/data.frame to a standard matrix, and then to a "dist" object.
# assume `x` is your imported table.
x1 <- as.dist(as(x, "matrix"))Now, if you did the above and it worked, class(x1) will return "dist" in your R session. Furthermore, you can no provide x1 as the dist argument to relevant phyloseq functions, like ordinate.
Performing MDS (referred to as PCoA by microbiome folks) is very fast, so I don't think there is much point in storing or importing the QIIME-calculated object. Use phyloseq::ordinate instead, and avail yourself of some of the other options and supported ordination methods (MDS is not the only game in town).
Finally, the premise of your question seems to imply that calculating weighted-UniFrac is implausible in phyloseq. Actually, I've implemented an optionally-parallelized version of fastUniFrac for both weighted- and unweighted-UniFrac. Furthermore, the QIIME-calculated UniFrac distances will, at least by default, be based on rarefied counts, which you generally should not use. Ever. In phyloseq you can perform admissible transformations of your counts prior to your choice of distance calculation, and this is what I recommend.
Hope that helps. Thanks for the feedback and interest in phyloseq,
joey
p.s. be sure to check out the extensive documentation, tutorials, and other closed issues related to the distance, ordinate, and transform_sample_counts functions. What you need here is already pretty well documented. Try ?phyloseq::distance for instance, from an R session, if you have phyloseq installed.
nsarode
mweand
cresil
ronenliberman
I am pretty new to R, but thanks to Phyloseq, atleast I don't dread it anymore :)
I am learning Phyloseq using output files generated by QIIME. While working with the ordinate plots section, I saw text mentioning that Unifrac dist measurements (esp. weighted) takes quite a while. But from what I know, QIIME does provide precomputed weighted distance matrices (Unifrac dist weighted matrix as well as Principal coordinate matrix). Is there any way to import these precomputed files directly instead of computing them ? I tried reading these files as matrix, as table, but get errors. I'll be happy to send example file to you if you wish, but am pasting snippets of the file below.
QIIME generated Distance matrix (weighted_unifrac)
QIIME generated Principal coordinate matrix (weighted_unifrac)
Thanks !