Clustering
Jeff Leek
Dependencies
This document depends on the following packages:
library(devtools)
library(Biobase)
library(dendextend)To install these packages you can use the code (or if you are compiling the document, remove the eval=FALSE from the chunk.)
install.packages(c("devtools","dendextend"))
source("http://www.bioconductor.org/biocLite.R")
biocLite(c("Biobase"))General principles
- How do we define close?
- How do we group things?
- How do we visualize the grouping?
- How do we interpret the grouping?
Load some data
We will use this expression set to look at how we use plots and tables to check for different characteristics
con =url("http://bowtie-bio.sourceforge.net/recount/ExpressionSets/bodymap_eset.RData")
load(file=con)
close(con)
bm = bodymap.eset
pdata=pData(bm)
edata=as.data.frame(exprs(bm))
fdata = fData(bm)
ls()## [1] "bm" "bodymap.eset" "con" "edata"
## [5] "fdata" "pdata" "tropical"Calculate distances between samples
- Most important step
- Garbage in -> garbage out
- Distance or similarity
- Continuous - euclidean distance
- Continous - correlation similarity
- Binary - manhattan distance
- Pick a distance/similarity that makes sense for your problem
First we log transform and remove lowly expressed genes, then calculate Euclidean distance
edata = edata[rowMeans(edata) > 5000,]
edata = log2(edata + 1)
# By default calculates the distance between rows
dist1 = dist(t(edata))
## Look at distance matrix
colramp = colorRampPalette(c(3,"white",2))(9)
heatmap(as.matrix(dist1),col=colramp,Colv=NA,Rowv=NA)
Now cluster the samples
Here we use the distance we previously calculated to perform a hierarchical clustering and plot the dendrogram:
hclust1 = hclust(dist1)
plot(hclust1)
We can also force all of the leaves to terminate at the same spot
plot(hclust1,hang=-1)
We can also color the dendrogram either into a fixed number of groups
dend = as.dendrogram(hclust1)
dend = color_labels(hclust1,4,col=1:4)
plot(dend)
Or you can color them directly
labels_colors(dend) = c(rep(1,10),rep(2,9))
plot(dend)
This can get very fancy and we won’t cover all the options here but for further ideas check out the dendextend package.
K-means clustering
Now we can perform k-means clustering. By default, the rows are clustered. You can either input the cluster means (often unknown) or the number of clusters (here I input 3 clusters)
kmeans1 = kmeans(edata,centers=3)
names(kmeans1)## [1] "cluster" "centers" "totss" "withinss"
## [5] "tot.withinss" "betweenss" "size" "iter"
## [9] "ifault"Now we can look at the cluster centers
matplot(t(kmeans1$centers),col=1:3,type="l",lwd=3)
We can also look at how many belong to each cluster
table(kmeans1$cluster)##
## 1 2 3
## 63 390 745And cluster the data together and plot
heatmap(as.matrix(edata)[order(kmeans1$cluster),],col=colramp,Colv=NA,Rowv=NA)
Note that this is a random start so you get a different clustering if you run it again you get a different answer
kmeans2 = kmeans(edata,centers=3)
table(kmeans1$cluster,kmeans2$cluster)##
## 1 2 3
## 1 3 60 0
## 2 15 0 375
## 3 151 0 594Session information
Here is the session information
devtools::session_info()## setting value
## version R version 3.2.1 (2015-06-18)
## system x86_64, darwin10.8.0
## ui RStudio (0.99.447)
## language (EN)
## collate en_US.UTF-8
## tz America/New_York
##
## package * version date
## acepack 1.3-3.3 2014-11-24
## annotate 1.46.1 2015-07-11
## AnnotationDbi * 1.30.1 2015-04-26
## assertthat 0.1 2013-12-06
## BiasedUrn * 1.06.1 2013-12-29
## Biobase * 2.28.0 2015-04-17
## BiocGenerics * 0.14.0 2015-04-17
## BiocInstaller * 1.18.4 2015-07-22
## BiocParallel 1.2.20 2015-08-07
## biomaRt 2.24.0 2015-04-17
## Biostrings 2.36.3 2015-08-12
## bitops 1.0-6 2013-08-17
## bladderbatch * 1.6.0 2015-08-26
## broom * 0.3.7 2015-05-06
## caTools 1.17.1 2014-09-10
## cluster 2.0.3 2015-07-21
## colorspace 1.2-6 2015-03-11
## corpcor 1.6.8 2015-07-08
## curl 0.9.2 2015-08-08
## DBI * 0.3.1 2014-09-24
## dendextend * 1.1.0 2015-07-31
## DESeq2 * 1.8.1 2015-05-02
## devtools * 1.8.0 2015-05-09
## digest 0.6.8 2014-12-31
## dplyr * 0.4.3 2015-09-01
## edge * 2.1.0 2015-09-06
## evaluate 0.7.2 2015-08-13
## foreign 0.8-66 2015-08-19
## formatR 1.2 2015-04-21
## Formula * 1.2-1 2015-04-07
## futile.logger 1.4.1 2015-04-20
## futile.options 1.0.0 2010-04-06
## gdata 2.17.0 2015-07-04
## genefilter * 1.50.0 2015-04-17
## geneLenDataBase * 1.4.0 2015-09-06
## geneplotter 1.46.0 2015-04-17
## GenomeInfoDb * 1.4.2 2015-08-15
## GenomicAlignments 1.4.1 2015-04-24
## GenomicFeatures 1.20.2 2015-08-14
## GenomicRanges * 1.20.5 2015-06-09
## genstats * 0.1.02 2015-09-05
## ggplot2 * 1.0.1 2015-03-17
## git2r 0.11.0 2015-08-12
## GO.db 3.1.2 2015-09-06
## goseq * 1.20.0 2015-04-17
## gplots * 2.17.0 2015-05-02
## gridExtra 2.0.0 2015-07-14
## gtable 0.1.2 2012-12-05
## gtools 3.5.0 2015-05-29
## highr 0.5 2015-04-21
## HistData * 0.7-5 2014-04-26
## Hmisc * 3.16-0 2015-04-30
## htmltools 0.2.6 2014-09-08
## httr 1.0.0 2015-06-25
## IRanges * 2.2.7 2015-08-09
## KernSmooth 2.23-15 2015-06-29
## knitr * 1.11 2015-08-14
## lambda.r 1.1.7 2015-03-20
## lattice * 0.20-33 2015-07-14
## latticeExtra 0.6-26 2013-08-15
## lazyeval 0.1.10 2015-01-02
## limma * 3.24.15 2015-08-06
## lme4 1.1-9 2015-08-20
## locfit 1.5-9.1 2013-04-20
## magrittr 1.5 2014-11-22
## MASS * 7.3-43 2015-07-16
## Matrix * 1.2-2 2015-07-08
## MatrixEQTL * 2.1.1 2015-02-03
## memoise 0.2.1 2014-04-22
## mgcv * 1.8-7 2015-07-23
## minqa 1.2.4 2014-10-09
## mnormt 1.5-3 2015-05-25
## munsell 0.4.2 2013-07-11
## nlme * 3.1-122 2015-08-19
## nloptr 1.0.4 2014-08-04
## nnet 7.3-10 2015-06-29
## org.Hs.eg.db * 3.1.2 2015-07-17
## plyr 1.8.3 2015-06-12
## preprocessCore * 1.30.0 2015-04-17
## proto 0.3-10 2012-12-22
## psych 1.5.6 2015-07-08
## qvalue * 2.0.0 2015-04-17
## R6 2.1.1 2015-08-19
## RColorBrewer 1.1-2 2014-12-07
## Rcpp * 0.12.0 2015-07-25
## RcppArmadillo * 0.5.400.2.0 2015-08-17
## RCurl 1.95-4.7 2015-06-30
## reshape2 1.4.1 2014-12-06
## rmarkdown 0.7 2015-06-13
## rpart 4.1-10 2015-06-29
## Rsamtools 1.20.4 2015-06-01
## RSkittleBrewer * 1.1 2015-09-05
## RSQLite * 1.0.0 2014-10-25
## rstudioapi 0.3.1 2015-04-07
## rtracklayer 1.28.9 2015-08-19
## rversions 1.0.2 2015-07-13
## S4Vectors * 0.6.5 2015-09-01
## scales 0.3.0 2015-08-25
## snm 1.16.0 2015-04-17
## snpStats * 1.18.0 2015-04-17
## stringi 0.5-5 2015-06-29
## stringr 1.0.0 2015-04-30
## survival * 2.38-3 2015-07-02
## sva * 3.14.0 2015-04-17
## tidyr 0.2.0 2014-12-05
## UsingR * 2.0-5 2015-08-06
## whisker 0.3-2 2013-04-28
## XML 3.98-1.3 2015-06-30
## xml2 0.1.2 2015-09-01
## xtable 1.7-4 2014-09-12
## XVector 0.8.0 2015-04-17
## yaml 2.1.13 2014-06-12
## zlibbioc 1.14.0 2015-04-17
## source
## CRAN (R 3.2.0)
## Bioconductor
## Bioconductor
## CRAN (R 3.2.0)
## CRAN (R 3.2.0)
## Bioconductor
## Bioconductor
## Bioconductor
## Bioconductor
## Bioconductor
## Bioconductor
## CRAN (R 3.2.0)
## Bioconductor
## CRAN (R 3.2.0)
## CRAN (R 3.2.0)
## CRAN (R 3.2.1)
## CRAN (R 3.2.0)
## CRAN (R 3.2.1)
## CRAN (R 3.2.1)
## CRAN (R 3.2.0)
## CRAN (R 3.2.1)
## Bioconductor
## CRAN (R 3.2.0)
## CRAN (R 3.2.0)
## CRAN (R 3.2.2)
## Github (jdstorey/edge@a1947b5)
## CRAN (R 3.2.2)
## CRAN (R 3.2.1)
## CRAN (R 3.2.0)
## CRAN (R 3.2.0)
## CRAN (R 3.2.0)
## CRAN (R 3.2.0)
## CRAN (R 3.2.1)
## Bioconductor
## Bioconductor
## Bioconductor
## Bioconductor
## Bioconductor
## Bioconductor
## Bioconductor
## local
## CRAN (R 3.2.0)
## CRAN (R 3.2.2)
## Bioconductor
## Bioconductor
## CRAN (R 3.2.0)
## CRAN (R 3.2.1)
## CRAN (R 3.2.0)
## CRAN (R 3.2.1)
## CRAN (R 3.2.0)
## CRAN (R 3.2.0)
## CRAN (R 3.2.0)
## CRAN (R 3.2.0)
## CRAN (R 3.2.1)
## Bioconductor
## CRAN (R 3.2.1)
## CRAN (R 3.2.2)
## CRAN (R 3.2.0)
## CRAN (R 3.2.1)
## CRAN (R 3.2.0)
## CRAN (R 3.2.0)
## Bioconductor
## CRAN (R 3.2.2)
## CRAN (R 3.2.0)
## CRAN (R 3.2.0)
## CRAN (R 3.2.1)
## CRAN (R 3.2.1)
## CRAN (R 3.2.0)
## CRAN (R 3.2.0)
## CRAN (R 3.2.1)
## CRAN (R 3.2.0)
## CRAN (R 3.2.1)
## CRAN (R 3.2.0)
## CRAN (R 3.2.1)
## CRAN (R 3.2.0)
## CRAN (R 3.2.1)
## Bioconductor
## CRAN (R 3.2.1)
## Bioconductor
## CRAN (R 3.2.0)
## CRAN (R 3.2.1)
## Bioconductor
## CRAN (R 3.2.1)
## CRAN (R 3.2.0)
## CRAN (R 3.2.1)
## CRAN (R 3.2.1)
## CRAN (R 3.2.1)
## CRAN (R 3.2.0)
## CRAN (R 3.2.1)
## CRAN (R 3.2.1)
## Bioconductor
## Github (alyssafrazee/RSkittleBrewer@0a96a20)
## CRAN (R 3.2.0)
## CRAN (R 3.2.0)
## Bioconductor
## CRAN (R 3.2.1)
## Bioconductor
## CRAN (R 3.2.2)
## Bioconductor
## Bioconductor
## CRAN (R 3.2.1)
## CRAN (R 3.2.0)
## CRAN (R 3.2.1)
## Bioconductor
## CRAN (R 3.2.0)
## CRAN (R 3.2.2)
## CRAN (R 3.2.0)
## CRAN (R 3.2.1)
## CRAN (R 3.2.2)
## CRAN (R 3.2.0)
## Bioconductor
## CRAN (R 3.2.0)
## BioconductorIt is also useful to compile the time the document was processed. This document was processed on: 2015-09-06.