TADmetric.Rmd

---
title: "TADs"
author: "M.Munn"
date: "July 5, 2016"
output: html_document
---

```{r}
rawESCdata <- read.csv("~/Dropbox/DS/wTrush/rawESCdata.csv", header=TRUE, stringsAsFactors = FALSE)
A = rawESCdata[2:length(rawESCdata[,2]), 2:4]
str(A)
dim(A)
```

Convert from character to numeric

```{r}
A = transform(A, X = as.numeric(X), X.1 = as.numeric(X.1), X.2 = as.numeric(X.2))
names(A) = c("startTAD", "endTAD", "interaction")
dim(A)
str(A)
```

Distance should be inversely proportional to interaction. There is room to change this along the way. Perhaps we can make this adjustable? some Shiny app.

```{r}
A$distance = 1/A$interaction
```

Also, we set up the distance metric.

```{r}
dist = matrix(, length(A[,1]), length(A[,1]))
dist = data.frame(dist)
for(i in 1:length(A[,1])){
    dist[A[i,1],A[i,2]] = A[i,4]
}

for(i in 1:dim(dist)[1]){
        for(j in 1:dim(dist)[2]){
                if(is.na(dist[i,j])){
                dist[i,j] = dist[j,i]      
                }       
        }
}

for(i in 1:dim(dist)[1]){dist[i,i] = 0}

sum(is.na(dist))
```

There are a LOT of NA values out of 393*393 - sum(is.na(dist)) = 384/2 = 192 actual values. We can just assume that distances beyond those values are irrelevant. That is, 

```{r}
min(dist, na.rm=TRUE)
max(dist,na.rm = TRUE)
```