Progress batman

From CSBLwiki

Revision as of 09:38, 7 April 2011 (view source) Igchoi (Talk | contribs) (→ADDA algorithm) ← Older edit		Revision as of 07:45, 14 April 2011 (view source) Batman (Talk | contribs) (→How to parse domains from the correlation matrix) Newer edit →
Line 68:		Line 68:
	===How to parse domains from the correlation matrix===		===How to parse domains from the correlation matrix===
		+	<pre style col="blue">
		+	</pre>
		+	chi = NULL
		+	for (i in 10:(ncol(count.que)-10)){
		+	A = i
		+	c1 = sum(corr[1:i,1:i])
		+	a = sum(corr[1:i ,(i+1):(ncol(count.que))])
		+	c2 = sum(corr[(i+1):(ncol(count.que)),(i+1):(ncol(count.que))])
		+	x =  (((c1*c2)-(a*a))^2)/(((c1+a)^2)*((c2+a)^2))
		+	# print(paste(A,c1,a,c2 ,x))
		+	chi = c(chi,x)
		+	p1 = which(chi==min(chi))+9
		+	# chi[i-9,1:2] = c(A,x)
		+	}
		+	chi1= NULL
		+	for (i in 10:p1){
		+	A = i
		+	c1 = sum(corr[1:i,1:i])
		+	a = sum(corr[1:i ,(i+1):(ncol(count.que))])
		+	c2 = sum(corr[(i+1):(ncol(count.que)),(i+1):(ncol(count.que))])
		+	x =  (((c1*c2)-(a*a))^2)/(((c1+a)^2)*((c2+a)^2))
		+	# print(paste(A,c1,a,c2 ,x))
		+	chi1 = c(chi1,x)
		+	p2 = which(chi1==min(chi1))+9
		+	# chi[i-9,1:2] = c(A,x)
		+	}
		+	chi2=NULL
		+	for (i in p1:(ncol(count.que)-10)){
		+	A = i
		+	c1 = sum(corr[1:i,1:i])
		+	a = sum(corr[1:i ,(i+1):(ncol(count.que))])
		+	c2 = sum(corr[(i+1):(ncol(count.que)),(i+1):(ncol(count.que))])
		+	x =  (((c1*c2)-(a*a))^2)/(((c1+a)^2)*((c2+a)^2))
		+	# print(paste(A,c1,a,c2 ,x))
		+	chi2 = c(chi2,x)
		+	p3 = which(chi2==min(chi2))+p1
		+	# chi[i-9,1:2] = c(A,x)
		+	}
		+
		+
	<pre>		<pre>
	</pre>		</pre>

---

Revision as of 07:45, 14 April 2011

Contents 1 2011 2 Schedule 3 ADDA algorithm 3.1 How to get residue correlation matrix 3.2 How to parse domains from the correlation matrix

2011

• Reference: PDF download

1. Error fetching PMID 12706730: [ADDA]

Schedule

ADDA algorithm

How to get residue correlation matrix

• sample data: sample.tab

# set working directory
setwd("/Users/igchoi/Downloads/")

# read blast table
tmp = read.table("sample.tab",sep="\t")

# check dimension of table
dim(tmp)

# change simpler GI number (skip this if you don't understand)
tmp[,1] = sapply(as.vector(tmp[,1]), function(x) {
  y = strsplit(x,"\\|"); return(unlist(y)[2]) })
tmp[,2] = sapply(as.vector(tmp[,2]), function(x) {
  y = strsplit(x,"\\|"); return(unlist(y)[2]) })

# check the result
tmp[1:5,]

# aligned region: V7 V8, V9 V10
# query length = 590 residues
count.que = matrix(0,nrow(tmp),590)  # <- build empty 'aligned' matrix
for(i in 1:nrow(tmp)) {
  res = unlist(tmp[i,7:8])
  print(res)
  count.que[i,c(res[1]:res[2])] = 1
}

# check aligned region (blue colored regions are aligned with other proteins - central region)
image(t(count.que),col=c("white","blue"))

# count number of neighbors occurring both position i and j (using 'aligned' matrix)
corr = matrix(0,590,590)  # <- set empty correlation matrix (590 x 590)

for (i in 1:ncol(count.que)) {
  for (j in i:ncol(count.que)) {
    print(paste(i,j))
    chk = count.que[,i]+count.que[,j]
    cnt = length(which(chk==2))
    corr[i,j] = cnt
  }
}

# range of number of neighbors (among 52 hits)
range(corr)

# check highly correlated region in the correlation matrix
image(corr,col=topo.colors(10))

How to parse domains from the correlation matrix

chi = NULL for (i in 10:(ncol(count.que)-10)){

       A = i
      c1 = sum(corr[1:i,1:i])
       a = sum(corr[1:i ,(i+1):(ncol(count.que))])
      c2 = sum(corr[(i+1):(ncol(count.que)),(i+1):(ncol(count.que))])
      x =  (((c1*c2)-(a*a))^2)/(((c1+a)^2)*((c2+a)^2))
     # print(paste(A,c1,a,c2 ,x))
      chi = c(chi,x)
      p1 = which(chi==min(chi))+9
     # chi[i-9,1:2] = c(A,x)

} chi1= NULL for (i in 10:p1){

       A = i
      c1 = sum(corr[1:i,1:i])
       a = sum(corr[1:i ,(i+1):(ncol(count.que))])
      c2 = sum(corr[(i+1):(ncol(count.que)),(i+1):(ncol(count.que))])
      x =  (((c1*c2)-(a*a))^2)/(((c1+a)^2)*((c2+a)^2))
     # print(paste(A,c1,a,c2 ,x))
      chi1 = c(chi1,x)
      p2 = which(chi1==min(chi1))+9
     # chi[i-9,1:2] = c(A,x)

} chi2=NULL for (i in p1:(ncol(count.que)-10)){

       A = i
      c1 = sum(corr[1:i,1:i])
       a = sum(corr[1:i ,(i+1):(ncol(count.que))])
      c2 = sum(corr[(i+1):(ncol(count.que)),(i+1):(ncol(count.que))])
      x =  (((c1*c2)-(a*a))^2)/(((c1+a)^2)*((c2+a)^2))
     # print(paste(A,c1,a,c2 ,x))
      chi2 = c(chi2,x)
      p3 = which(chi2==min(chi2))+p1
     # chi[i-9,1:2] = c(A,x)

}