mkborregaard / Microbiome.jl

Microbiome-related methods for julia

This package provides (or will provide) methods for microbial community analyses. For now, I'm adding stuff as I need it, but pull-requests are more than welcome.

Working with microbial abundances

The AbundanceTable type is treated like a 2D array where columns are samples and rows are features (eg species). Sample and feature names are also stored, and there's a convenience function if you want to convert a DataFrame to an AbundanceTable, assuming the first column contains feature names:

using Microbiome
using DataFrames

df = DataFrame(species=["E. coli", "B. fragilis", "L. casei"],
               sample1=[0.1, 0.4, 0.5],
               sample2=[0.3, 0.7, 0.0],
               sample3=[0.0, 0.2, 0.8])

abund = AbundanceTable(df)

@show abund
@show abund.samples
@show abund.features

Note: I've used a relative abundance table here, but that need not be the case. You can also use raw counts, RPK etc. If you want relative abundance, you can do relativeabundance(abund)

You can also filter on the n most abundant features accross the dataset. This function automatically generates an n+1 row for other containing the remaining features. Note - these doesn't modify in-place, so you've gotta reassign if you want to update:

abund2 = filterabund(abund, 1)
@show abund2
@show abund2.features

Working with Distances / Dissimilarity

Quite often, it's useful to boil stuff down to distances between samples. For this, I'm using an interface with Distances.jl to generate a symetric DistanceMatrix, which also contains a vector for samples, and a field specifying which type of distance was used to calulate it. You can load one in manually, or generate it from an AbundanceTable.

using Distances

dm = getdm(abund, BrayCurtis())
@show dm
@show dm.labels

I've also implemented a method to do a principle coordinates analysis. If necessary, you can include correct_neg=true to use the correction method described in Lingoes (1971)

p = pcoa(dm)

@show eigenvalue(p, 2)
@show principalcoord(p, 1)

Plotting

I've included some plotting recipes for convenience using RecipesBase.

using StatPlots

abund = AbundanceTable(
    rand(100, 10), ["sample_$x" for x in 1:10],
    ["feature_$x" for x in 1:100])

abund = relativeabundance(abund)
plot(abund, title="Random abundance")

dm = getdm(abund, BrayCurtis())
p = pcoa(dm, correct_neg=true)

plot(p, title="Random PCoA")