clmp is an R extension, mostly written in C, for extracting genetic clusters from a phylogeny using a Markov-modulated Poisson process to model variation in branching rates.
Our paper that describes and evaluates MMPP as a model-based clustering method for HIV epidemiology was recently accepted in PLOS Computational Biology.
clmp is also available as a Shiny web application that you can try out here:
https://shiny.filogeneti.ca/clmp/
A genetic cluster is a subset of nucleotide or amino acid sequences that are more similar to each other than to the remaining sequences in the sample population. For infectious diseases, a genetic cluster may correspond to an outbreak of cases that are clustered in space and/or time, which can imply that the cases are related by a common source.
> require(clmp)
Loading required package: clmp
Loading required package: ape
> t1 <- read.tree('data/test.nwk') # a simulated tree with 1000 tips, 100 in clusters
> res <- clmp(t1) # returns an ape::phylo tree object
log likelihood for 2 state model is 2238.543290
rates: 495.368085 1305.115860
Q: [ * 2.526691 ]
[ 23.309483 * ]
> index <- match(t1$tip.label, names(res$clusters))
> labels <- grepl("_1_", t1$tip.label) # extract truth from the tip labels
> table(labels, res$clusters[index])
labels 0 1
FALSE 860 3 # false positive rate, 3/(3+860)=0.34%
TRUE 13 98 # true positive rate, 98/(98+13)=88.2%For step-by-step instructions to compile the clmp R package from source, please refer to our installation documentation.