edgeR vs. SAMSeq

A while ago I received a comment on comparing parametric methods against nonparametric for calling differential expression in count data. Here I compare SAMSeq (Jun Li and Robert Tibshirani (2011) Finding consistent patterns: a nonparametric approach for identifying differential expression in RNA-Seq data. Statistical Methods in Medical Research, in press.) with edgeR. For more information on SAMSeq refer to the paper. I intend to look at cuffdiff sometime later.

library(edgeR)
raw.data <- read.delim("pnas_expression_filtered.tsv",row.names=1)
d <- raw.data[,1:7]
rownames(d) <- rownames(raw.data)
head(d)
#                lane1 lane2 lane3 lane4 lane5 lane6 lane8
#ENSG00000124208   478   619   628   744   483   716   240
#ENSG00000182463    27    20    27    26    48    55    24
#ENSG00000124201   180   218   293   275   373   301    88
#ENSG00000124207    76    80    85    97    80    81    37
#ENSG00000125835   132   200   200   228   280   204    52
#ENSG00000125834    42    60    72    86   131    99    30
group <- c(rep("Control",4),rep("DHT",3))
d <- DGEList(counts = d, group=group)
#Calculating library sizes from column totals.
d <- calcNormFactors(d)
d <- estimateCommonDisp(d)
de.com <- exactTest(d)
#Comparison of groups:  DHT - Control
summary(decideTestsDGE(de.com,p.value=0.05))
#   [,1]
#-1  2368
#0  11715
#1   2411
d <- estimateTagwiseDisp(d, prop.used=0.5, grid.length=500)
de.tgw <- exactTest(d)
summary(decideTestsDGE(de.tgw, p.value=0.05))
#   [,1]
#-1  2119
#0  12048
#1   2327
#total 4446
de <- summary(decideTestsDGE(de.tgw,p.value=0.05))
total <- de[1] + de[3]
sink("edger.output")
topTags(de.tgw,n=total)
sink()
sessionInfo()
#R version 2.14.1 (2011-12-22)
#Platform: x86_64-unknown-linux-gnu (64-bit)
#
#locale:
#[1] C
#
#attached base packages:
#[1] stats     graphics  grDevices utils     datasets  methods   base
#
#other attached packages:
#[1] edgeR_2.4.1
#
#loaded via a namespace (and not attached):
#[1] limma_3.10.0

For SAMSeq:

Of the 4,408 significantly differentiated genes detected by SAMSeq using the settings above, 3,908 overlap the 4,446 detected by edgeR using a tagwise dispersion.

See also: edgeR vs. DESeq using this same dataset.

and

http://www.stanford.edu/~junli07/research.html#SAM