Exosome_Microarray_Analysis.Rmd

---
title: "Exosome Microarray Analysis"
author: "Claire Levy"
date: "January 11, 2017"
output: github_document
---

```{r setup, include=FALSE, cache = TRUE}
knitr::opts_chunk$set(echo = TRUE, warning = FALSE, message = FALSE)


library(dplyr)
library(lumi)
library(limma)
library(pander)
library(stringr)
library(ggplot2)

##It turns out that you can just use this lumiR.batch function instead of all that stuff that I used in the HVE and explant MA read-in

#Even though I only have one MA file to read in, I am using lumiR.BATCH because I also want to use the argument where I can put in a pData file. The pData must be a tab delim txt file. I converted the .csv file that I already had from Lucia to a .txt file.

lumibatch<-lumiR.batch(fileList = "..\\raw_data\\FinalReport.txt",
            detectionTh = 0.05, sampleInfoFile = "..\\raw_data\\SampleInformation.txt")

#note that control data is here
#getControlData(lumibatch)

#and pData is here
#pData(lumibatch)

#and that the QC of non-normalized data has been done
#summary(lumibatch, 'QC')

```

```{r density plot of nonnorm data}

density(lumibatch)#number of probes for each sample that occur
#at a certain log2 intensity
```


```{r sample relations plot}
plot(lumibatch, what='sampleRelation',method="mds")
```

```{r boxplot}
#boxplot
boxplot(lumibatch)

```


```{r background correction, results = 'hide'}

#the data we got from the core had no background correction (I don't think it did anyway...) so I will do it here.
B_lumibatch<-lumiB(lumibatch, method = "bgAdjust")


# VST TRANSFORMATION 
#"Stabilizing the expression variance based on
#the bead level expression variance and mean relations"

TB_lumibatch <-lumiT (B_lumibatch, method = 'vst')
#can do this to look at a plot. plotting doesn't work if I do method = "log2"
#plotVST(TB_lumibatch)
```

This looks good, right?

```{r quantile normalization and plots}

#trying quantile normalization here (not robust spline as before)
NTB_lumibatch<-lumiN(TB_lumibatch,method="quantile")

#this seems to be fine with quantile normalization...
density(NTB_lumibatch)
```

```{r quantile norm boxplot}

boxplot(NTB_lumibatch)
```


This density plot looks weird but the boxplot looks ok I guess?
```{r RSN normalization and plot}

#trying quantile normalization here (not robust spline as before)
NrsnTB_lumibatch<-lumiN(TB_lumibatch,method="rsn")

density(NrsnTB_lumibatch)
```



```{r rns boxplot}
boxplot(NrsnTB_lumibatch)
```

```{r QC of normalized data}

#I'm going to use the quantile normalized data for now.

QNTB_lumibatch<-lumiQ(NTB_lumibatch)

```





```{r MDS plot}
#to get data into dataframe form for nice plotting...

dimensional_data <- plotMDS(QNTB_lumibatch)#make the plot, it invisibly returns a large object that has the actual dimensions data in matrices called x and y (one for each dimension)

#make a data frame where
d <- data.frame(Dim1 = dimensional_data$x, #Dim1 is a column containing the values from the x matrix
               Dim2 = dimensional_data$y, #Dim2 is a column containing the values from the y matrix
               sampleID = names(dimensional_data$x))#SampleNames is a column that holds the column names from the x matrix, which happen to be the microarray wells.



d <- as_data_frame(merge(d, pData(QNTB_lumibatch)), by = "sampleID")
```

Clustering appears to be primarily by cell type, then by Virus, then donor and Exosomes are sort of mixed.

```{r plot dims}


ggplot(d, aes(x=Dim1, y = Dim2))+
  geom_point(aes(color = Exosomes, shape = Virus), size = 3)

```

```{r plot dims facet donor}

ggplot(d, aes(x=Dim1, y = Dim2))+
  geom_point(aes(color = Exosomes, shape = Virus), size = 3)+
  facet_wrap(~Donor)
```