stats_isomirs.md

title	author	date	output
isomiR annotation	Lorena Pantano	Fri Aug 5 16:26:15 2016	html_document toc toc_depth code_folding toc_float theme highlight true 2 hide true readable zenburn

data <- read.table("stats_isomirs_full", header = T)

Overview

The idea of this report is to show how well miraligner and srnabench are decting isomiRs. I simulated a bunch of isomirs (16000) that can have the folowing variation:

• starts at different position than the reference miRNA: t5
• ends at different position than the reference miRNA: t3
• have a mutation: muts
• have nt addition at the end: add

I used the file miRBase_isoAnnotation.txt from srnabench results. This file only have 1332 sequences, so I don't know what happened with the rest, but it is the only file I found with this information.

Detection

Correct isomir annotation for miraligner and srnabench.

library(ggplot2)
library(reshape)
data_gg <- melt(data, id.vars = c("name", "is_iso", "tool", "find"))
ggplot(data_gg, aes(tool, fill = is_iso)) + geom_bar() + scale_fill_brewer(palette = "Set1") + 
    theme_bw() + facet_wrap("find") + theme(axis.text.x = element_text(angle = 90, 
    hjust = 1))

#Accuracy

• is_mir is True when the isomiR is annotated to the correct miRNA
• info is True when the information of the changes are reported, is False when there are multiple changes but there is no annotation of the exactly variations. This happens because srnabench flag as mv when the sequence has variations at both sides.

data_gg_ann <- melt(subset(data, is_mir == "True"), id.vars = c("name", "is_iso", 
    "tool", "find"))
ggplot(subset(data_gg_ann, is_iso == "True" & find == "Yes"), aes(variable, 
    fill = value)) + geom_bar() + theme_bw() + scale_fill_brewer("correct", 
    palette = "Set1") + facet_wrap(~tool) + theme(axis.text.x = element_text(angle = 90, 
    hjust = 1))