Mercurial > repos > davidvanzessen > shm_csr

comparison baseline/Baseline_Functions.r @ 0:c33d93683a09 draft

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Uploaded
author davidvanzessen
date Thu, 13 Oct 2016 10:52:24 -0400
parents
children ba33b94637ca
comparison
equal deleted inserted replaced
-1:000000000000 0:c33d93683a09
1 #########################################################################################
2 # License Agreement
3 #
4 # THIS WORK IS PROVIDED UNDER THE TERMS OF THIS CREATIVE COMMONS PUBLIC LICENSE
5 # ("CCPL" OR "LICENSE"). THE WORK IS PROTECTED BY COPYRIGHT AND/OR OTHER
6 # APPLICABLE LAW. ANY USE OF THE WORK OTHER THAN AS AUTHORIZED UNDER THIS LICENSE
7 # OR COPYRIGHT LAW IS PROHIBITED.
8 #
9 # BY EXERCISING ANY RIGHTS TO THE WORK PROVIDED HERE, YOU ACCEPT AND AGREE TO BE
10 # BOUND BY THE TERMS OF THIS LICENSE. TO THE EXTENT THIS LICENSE MAY BE CONSIDERED
11 # TO BE A CONTRACT, THE LICENSOR GRANTS YOU THE RIGHTS CONTAINED HERE IN
12 # CONSIDERATION OF YOUR ACCEPTANCE OF SUCH TERMS AND CONDITIONS.
13 #
14 # BASELIne: Bayesian Estimation of Antigen-Driven Selection in Immunoglobulin Sequences
15 # Coded by: Mohamed Uduman & Gur Yaari
16 # Copyright 2012 Kleinstein Lab
17 # Version: 1.3 (01/23/2014)
18 #########################################################################################
19
20 # Global variables
21
22 FILTER_BY_MUTATIONS = 1000
23
24 # Nucleotides
25 NUCLEOTIDES = c("A","C","G","T")
26
27 # Amino Acids
28 AMINO_ACIDS <- c("F", "F", "L", "L", "S", "S", "S", "S", "Y", "Y", "*", "*", "C", "C", "*", "W", "L", "L", "L", "L", "P", "P", "P", "P", "H", "H", "Q", "Q", "R", "R", "R", "R", "I", "I", "I", "M", "T", "T", "T", "T", "N", "N", "K", "K", "S", "S", "R", "R", "V", "V", "V", "V", "A", "A", "A", "A", "D", "D", "E", "E", "G", "G", "G", "G")
29 names(AMINO_ACIDS) <- c("TTT", "TTC", "TTA", "TTG", "TCT", "TCC", "TCA", "TCG", "TAT", "TAC", "TAA", "TAG", "TGT", "TGC", "TGA", "TGG", "CTT", "CTC", "CTA", "CTG", "CCT", "CCC", "CCA", "CCG", "CAT", "CAC", "CAA", "CAG", "CGT", "CGC", "CGA", "CGG", "ATT", "ATC", "ATA", "ATG", "ACT", "ACC", "ACA", "ACG", "AAT", "AAC", "AAA", "AAG", "AGT", "AGC", "AGA", "AGG", "GTT", "GTC", "GTA", "GTG", "GCT", "GCC", "GCA", "GCG", "GAT", "GAC", "GAA", "GAG", "GGT", "GGC", "GGA", "GGG")
30 names(AMINO_ACIDS) <- names(AMINO_ACIDS)
31
32 #Amino Acid Traits
33 #"*" "A" "C" "D" "E" "F" "G" "H" "I" "K" "L" "M" "N" "P" "Q" "R" "S" "T" "V" "W" "Y"
34 #B = "Hydrophobic/Burried" N = "Intermediate/Neutral" S="Hydrophilic/Surface")
35 TRAITS_AMINO_ACIDS_CHOTHIA98 <- c("*","N","B","S","S","B","N","N","B","S","B","B","S","N","S","S","N","N","B","B","N")
36 names(TRAITS_AMINO_ACIDS_CHOTHIA98) <- sort(unique(AMINO_ACIDS))
37 TRAITS_AMINO_ACIDS <- array(NA,21)
38
39 # Codon Table
40 CODON_TABLE <- as.data.frame(matrix(NA,ncol=64,nrow=12))
41
42 # Substitution Model: Smith DS et al. 1996
43 substitution_Literature_Mouse <- matrix(c(0, 0.156222928, 0.601501588, 0.242275484, 0.172506739, 0, 0.241239892, 0.586253369, 0.54636291, 0.255795364, 0, 0.197841727, 0.290240811, 0.467680608, 0.24207858, 0),nrow=4,byrow=T,dimnames=list(NUCLEOTIDES,NUCLEOTIDES))
44 substitution_Flu_Human <- matrix(c(0,0.2795596,0.5026927,0.2177477,0.1693210,0,0.3264723,0.5042067,0.4983549,0.3328321,0,0.1688130,0.2021079,0.4696077,0.3282844,0),4,4,byrow=T,dimnames=list(NUCLEOTIDES,NUCLEOTIDES))
45 substitution_Flu25_Human <- matrix(c(0,0.2580641,0.5163685,0.2255674,0.1541125,0,0.3210224,0.5248651,0.5239281,0.3101292,0,0.1659427,0.1997207,0.4579444,0.3423350,0),4,4,byrow=T,dimnames=list(NUCLEOTIDES,NUCLEOTIDES))
46 load("FiveS_Substitution.RData")
47
48 # Mutability Models: Shapiro GS et al. 2002
49 triMutability_Literature_Human <- matrix(c(0.24, 1.2, 0.96, 0.43, 2.14, 2, 1.11, 1.9, 0.85, 1.83, 2.36, 1.31, 0.82, 0.52, 0.89, 1.33, 1.4, 0.82, 1.83, 0.73, 1.83, 1.62, 1.53, 0.57, 0.92, 0.42, 0.42, 1.47, 3.44, 2.58, 1.18, 0.47, 0.39, 1.12, 1.8, 0.68, 0.47, 2.19, 2.35, 2.19, 1.05, 1.84, 1.26, 0.28, 0.98, 2.37, 0.66, 1.58, 0.67, 0.92, 1.76, 0.83, 0.97, 0.56, 0.75, 0.62, 2.26, 0.62, 0.74, 1.11, 1.16, 0.61, 0.88, 0.67, 0.37, 0.07, 1.08, 0.46, 0.31, 0.94, 0.62, 0.57, 0.29, NA, 1.44, 0.46, 0.69, 0.57, 0.24, 0.37, 1.1, 0.99, 1.39, 0.6, 2.26, 1.24, 1.36, 0.52, 0.33, 0.26, 1.25, 0.37, 0.58, 1.03, 1.2, 0.34, 0.49, 0.33, 2.62, 0.16, 0.4, 0.16, 0.35, 0.75, 1.85, 0.94, 1.61, 0.85, 2.09, 1.39, 0.3, 0.52, 1.33, 0.29, 0.51, 0.26, 0.51, 3.83, 2.01, 0.71, 0.58, 0.62, 1.07, 0.28, 1.2, 0.74, 0.25, 0.59, 1.09, 0.91, 1.36, 0.45, 2.89, 1.27, 3.7, 0.69, 0.28, 0.41, 1.17, 0.56, 0.93, 3.41, 1, 1, NA, 5.9, 0.74, 2.51, 2.24, 2.24, 1.95, 3.32, 2.34, 1.3, 2.3, 1, 0.66, 0.73, 0.93, 0.41, 0.65, 0.89, 0.65, 0.32, NA, 0.43, 0.85, 0.43, 0.31, 0.31, 0.23, 0.29, 0.57, 0.71, 0.48, 0.44, 0.76, 0.51, 1.7, 0.85, 0.74, 2.23, 2.08, 1.16, 0.51, 0.51, 1, 0.5, NA, NA, 0.71, 2.14), nrow=64,byrow=T)
50 triMutability_Literature_Mouse <- matrix(c(1.31, 1.35, 1.42, 1.18, 2.02, 2.02, 1.02, 1.61, 1.99, 1.42, 2.01, 1.03, 2.02, 0.97, 0.53, 0.71, 1.19, 0.83, 0.96, 0.96, 0, 1.7, 2.22, 0.59, 1.24, 1.07, 0.51, 1.68, 3.36, 3.36, 1.14, 0.29, 0.33, 0.9, 1.11, 0.63, 1.08, 2.07, 2.27, 1.74, 0.22, 1.19, 2.37, 1.15, 1.15, 1.56, 0.81, 0.34, 0.87, 0.79, 2.13, 0.49, 0.85, 0.97, 0.36, 0.82, 0.66, 0.63, 1.15, 0.94, 0.85, 0.25, 0.93, 1.19, 0.4, 0.2, 0.44, 0.44, 0.88, 1.06, 0.77, 0.39, 0, 0, 0, 0, 0, 0, 0.43, 0.43, 0.86, 0.59, 0.59, 0, 1.18, 0.86, 2.9, 1.66, 0.4, 0.2, 1.54, 0.43, 0.69, 1.71, 0.68, 0.55, 0.91, 0.7, 1.71, 0.09, 0.27, 0.63, 0.2, 0.45, 1.01, 1.63, 0.96, 1.48, 2.18, 1.2, 1.31, 0.66, 2.13, 0.49, 0, 0, 0, 2.97, 2.8, 0.79, 0.4, 0.5, 0.4, 0.11, 1.68, 0.42, 0.13, 0.44, 0.93, 0.71, 1.11, 1.19, 2.71, 1.08, 3.43, 0.4, 0.67, 0.47, 1.02, 0.14, 1.56, 1.98, 0.53, 0.33, 0.63, 2.06, 1.77, 1.46, 3.74, 2.93, 2.1, 2.18, 0.78, 0.73, 2.93, 0.63, 0.57, 0.17, 0.85, 0.52, 0.31, 0.31, 0, 0, 0.51, 0.29, 0.83, 0.54, 0.28, 0.47, 0.9, 0.99, 1.24, 2.47, 0.73, 0.23, 1.13, 0.24, 2.12, 0.24, 0.33, 0.83, 1.41, 0.62, 0.28, 0.35, 0.77, 0.17, 0.72, 0.58, 0.45, 0.41), nrow=64,byrow=T)
51 triMutability_Names <- c("AAA", "AAC", "AAG", "AAT", "ACA", "ACC", "ACG", "ACT", "AGA", "AGC", "AGG", "AGT", "ATA", "ATC", "ATG", "ATT", "CAA", "CAC", "CAG", "CAT", "CCA", "CCC", "CCG", "CCT", "CGA", "CGC", "CGG", "CGT", "CTA", "CTC", "CTG", "CTT", "GAA", "GAC", "GAG", "GAT", "GCA", "GCC", "GCG", "GCT", "GGA", "GGC", "GGG", "GGT", "GTA", "GTC", "GTG", "GTT", "TAA", "TAC", "TAG", "TAT", "TCA", "TCC", "TCG", "TCT", "TGA", "TGC", "TGG", "TGT", "TTA", "TTC", "TTG", "TTT")
52 load("FiveS_Mutability.RData")
53
54 # Functions
55
56 # Translate codon to amino acid
57 translateCodonToAminoAcid<-function(Codon){
58 return(AMINO_ACIDS[Codon])
59 }
60
61 # Translate amino acid to trait change
62 translateAminoAcidToTraitChange<-function(AminoAcid){
63 return(TRAITS_AMINO_ACIDS[AminoAcid])
64 }
65
66 # Initialize Amino Acid Trait Changes
67 initializeTraitChange <- function(traitChangeModel=1,species=1,traitChangeFileName=NULL){
68 if(!is.null(traitChangeFileName)){
69 tryCatch(
70 traitChange <- read.delim(traitChangeFileName,sep="\t",header=T)
71 , error = function(ex){
72 cat("Error|Error reading trait changes. Please check file name/path and format.\n")
73 q()
74 }
75 )
76 }else{
77 traitChange <- TRAITS_AMINO_ACIDS_CHOTHIA98
78 }
79 TRAITS_AMINO_ACIDS <<- traitChange
80 }
81
82 # Read in formatted nucleotide substitution matrix
83 initializeSubstitutionMatrix <- function(substitutionModel,species,subsMatFileName=NULL){
84 if(!is.null(subsMatFileName)){
85 tryCatch(
86 subsMat <- read.delim(subsMatFileName,sep="\t",header=T)
87 , error = function(ex){
88 cat("Error|Error reading substitution matrix. Please check file name/path and format.\n")
89 q()
90 }
91 )
92 if(sum(apply(subsMat,1,sum)==1)!=4) subsMat = t(apply(subsMat,1,function(x)x/sum(x)))
93 }else{
94 if(substitutionModel==1)subsMat <- substitution_Literature_Mouse
95 if(substitutionModel==2)subsMat <- substitution_Flu_Human
96 if(substitutionModel==3)subsMat <- substitution_Flu25_Human
97
98 }
99
100 if(substitutionModel==0){
101 subsMat <- matrix(1,4,4)
102 subsMat[,] = 1/3
103 subsMat[1,1] = 0
104 subsMat[2,2] = 0
105 subsMat[3,3] = 0
106 subsMat[4,4] = 0
107 }
108
109
110 NUCLEOTIDESN = c(NUCLEOTIDES,"N", "-")
111 if(substitutionModel==5){
112 subsMat <- FiveS_Substitution
113 return(subsMat)
114 }else{
115 subsMat <- rbind(subsMat,rep(NA,4),rep(NA,4))
116 return( matrix(data.matrix(subsMat),6,4,dimnames=list(NUCLEOTIDESN,NUCLEOTIDES) ) )
117 }
118 }
119
120
121 # Read in formatted Mutability file
122 initializeMutabilityMatrix <- function(mutabilityModel=1, species=1,mutabilityMatFileName=NULL){
123 if(!is.null(mutabilityMatFileName)){
124 tryCatch(
125 mutabilityMat <- read.delim(mutabilityMatFileName,sep="\t",header=T)
126 , error = function(ex){
127 cat("Error|Error reading mutability matrix. Please check file name/path and format.\n")
128 q()
129 }
130 )
131 }else{
132 mutabilityMat <- triMutability_Literature_Human
133 if(species==2) mutabilityMat <- triMutability_Literature_Mouse
134 }
135
136 if(mutabilityModel==0){ mutabilityMat <- matrix(1,64,3)}
137
138 if(mutabilityModel==5){
139 mutabilityMat <- FiveS_Mutability
140 return(mutabilityMat)
141 }else{
142 return( matrix( data.matrix(mutabilityMat), 64, 3, dimnames=list(triMutability_Names,1:3)) )
143 }
144 }
145
146 # Read FASTA file formats
147 # Modified from read.fasta from the seqinR package
148 baseline.read.fasta <-
149 function (file = system.file("sequences/sample.fasta", package = "seqinr"),
150 seqtype = c("DNA", "AA"), as.string = FALSE, forceDNAtolower = TRUE,
151 set.attributes = TRUE, legacy.mode = TRUE, seqonly = FALSE,
152 strip.desc = FALSE, sizeof.longlong = .Machine$sizeof.longlong,
153 endian = .Platform$endian, apply.mask = TRUE)
154 {
155 seqtype <- match.arg(seqtype)
156
157 lines <- readLines(file)
158
159 if (legacy.mode) {
160 comments <- grep("^;", lines)
161 if (length(comments) > 0)
162 lines <- lines[-comments]
163 }
164
165
166 ind_groups<-which(substr(lines, 1L, 3L) == ">>>")
167 lines_mod<-lines
168
169 if(!length(ind_groups)){
170 lines_mod<-c(">>>All sequences combined",lines)
171 }
172
173 ind_groups<-which(substr(lines_mod, 1L, 3L) == ">>>")
174
175 lines <- array("BLA",dim=(length(ind_groups)+length(lines_mod)))
176 id<-sapply(1:length(ind_groups),function(i)ind_groups[i]+i-1)+1
177 lines[id] <- "THIS IS A FAKE SEQUENCE"
178 lines[-id] <- lines_mod
179 rm(lines_mod)
180
181 ind <- which(substr(lines, 1L, 1L) == ">")
182 nseq <- length(ind)
183 if (nseq == 0) {
184 stop("no line starting with a > character found")
185 }
186 start <- ind + 1
187 end <- ind - 1
188
189 while( any(which(ind%in%end)) ){
190 ind=ind[-which(ind%in%end)]
191 nseq <- length(ind)
192 if (nseq == 0) {
193 stop("no line starting with a > character found")
194 }
195 start <- ind + 1
196 end <- ind - 1
197 }
198
199 end <- c(end[-1], length(lines))
200 sequences <- lapply(seq_len(nseq), function(i) paste(lines[start[i]:end[i]], collapse = ""))
201 if (seqonly)
202 return(sequences)
203 nomseq <- lapply(seq_len(nseq), function(i) {
204
205 #firstword <- strsplit(lines[ind[i]], " ")[[1]][1]
206 substr(lines[ind[i]], 2, nchar(lines[ind[i]]))
207
208 })
209 if (seqtype == "DNA") {
210 if (forceDNAtolower) {
211 sequences <- as.list(tolower(chartr(".","-",sequences)))
212 }else{
213 sequences <- as.list(toupper(chartr(".","-",sequences)))
214 }
215 }
216 if (as.string == FALSE)
217 sequences <- lapply(sequences, s2c)
218 if (set.attributes) {
219 for (i in seq_len(nseq)) {
220 Annot <- lines[ind[i]]
221 if (strip.desc)
222 Annot <- substr(Annot, 2L, nchar(Annot))
223 attributes(sequences[[i]]) <- list(name = nomseq[[i]],
224 Annot = Annot, class = switch(seqtype, AA = "SeqFastaAA",
225 DNA = "SeqFastadna"))
226 }
227 }
228 names(sequences) <- nomseq
229 return(sequences)
230 }
231
232
233 # Replaces non FASTA characters in input files with N
234 replaceNonFASTAChars <-function(inSeq="ACGTN-AApA"){
235 gsub('[^ACGTNacgt[:punct:]-[:punct:].]','N',inSeq,perl=TRUE)
236 }
237
238 # Find the germlines in the FASTA list
239 germlinesInFile <- function(seqIDs){
240 firstChar = sapply(seqIDs,function(x){substr(x,1,1)})
241 secondChar = sapply(seqIDs,function(x){substr(x,2,2)})
242 return(firstChar==">" & secondChar!=">")
243 }
244
245 # Find the groups in the FASTA list
246 groupsInFile <- function(seqIDs){
247 sapply(seqIDs,function(x){substr(x,1,2)})==">>"
248 }
249
250 # In the process of finding germlines/groups, expand from the start to end of the group
251 expandTillNext <- function(vecPosToID){
252 IDs = names(vecPosToID)
253 posOfInterests = which(vecPosToID)
254
255 expandedID = rep(NA,length(IDs))
256 expandedIDNames = gsub(">","",IDs[posOfInterests])
257 startIndexes = c(1,posOfInterests[-1])
258 stopIndexes = c(posOfInterests[-1]-1,length(IDs))
259 expandedID = unlist(sapply(1:length(startIndexes),function(i){
260 rep(i,stopIndexes[i]-startIndexes[i]+1)
261 }))
262 names(expandedID) = unlist(sapply(1:length(startIndexes),function(i){
263 rep(expandedIDNames[i],stopIndexes[i]-startIndexes[i]+1)
264 }))
265 return(expandedID)
266 }
267
268 # Process FASTA (list) to return a matrix[input, germline)
269 processInputAdvanced <- function(inputFASTA){
270
271 seqIDs = names(inputFASTA)
272 numbSeqs = length(seqIDs)
273 posGermlines1 = germlinesInFile(seqIDs)
274 numbGermlines = sum(posGermlines1)
275 posGroups1 = groupsInFile(seqIDs)
276 numbGroups = sum(posGroups1)
277 consDef = NA
278
279 if(numbGermlines==0){
280 posGermlines = 2
281 numbGermlines = 1
282 }
283
284 glPositionsSum = cumsum(posGermlines1)
285 glPositions = table(glPositionsSum)
286 #Find the position of the conservation row
287 consDefPos = as.numeric(names(glPositions[names(glPositions)!=0 & glPositions==1]))+1
288 if( length(consDefPos)> 0 ){
289 consDefID = match(consDefPos, glPositionsSum)
290 #The coservation rows need to be pulled out and stores seperately
291 consDef = inputFASTA[consDefID]
292 inputFASTA = inputFASTA[-consDefID]
293
294 seqIDs = names(inputFASTA)
295 numbSeqs = length(seqIDs)
296 posGermlines1 = germlinesInFile(seqIDs)
297 numbGermlines = sum(posGermlines1)
298 posGroups1 = groupsInFile(seqIDs)
299 numbGroups = sum(posGroups1)
300 if(numbGermlines==0){
301 posGermlines = 2
302 numbGermlines = 1
303 }
304 }
305
306 posGroups <- expandTillNext(posGroups1)
307 posGermlines <- expandTillNext(posGermlines1)
308 posGermlines[posGroups1] = 0
309 names(posGermlines)[posGroups1] = names(posGroups)[posGroups1]
310 posInput = rep(TRUE,numbSeqs)
311 posInput[posGroups1 | posGermlines1] = FALSE
312
313 matInput = matrix(NA, nrow=sum(posInput), ncol=2)
314 rownames(matInput) = seqIDs[posInput]
315 colnames(matInput) = c("Input","Germline")
316
317 vecInputFASTA = unlist(inputFASTA)
318 matInput[,1] = vecInputFASTA[posInput]
319 matInput[,2] = vecInputFASTA[ which( names(inputFASTA)%in%paste(">",names(posGermlines)[posInput],sep="") )[ posGermlines[posInput]] ]
320
321 germlines = posGermlines[posInput]
322 groups = posGroups[posInput]
323
324 return( list("matInput"=matInput, "germlines"=germlines, "groups"=groups, "conservationDefinition"=consDef ))
325 }
326
327
328 # Replace leading and trailing dashes in the sequence
329 replaceLeadingTrailingDashes <- function(x,readEnd){
330 iiGap = unlist(gregexpr("-",x[1]))
331 ggGap = unlist(gregexpr("-",x[2]))
332 #posToChange = intersect(iiGap,ggGap)
333
334
335 seqIn = replaceLeadingTrailingDashesHelper(x[1])
336 seqGL = replaceLeadingTrailingDashesHelper(x[2])
337 seqTemplate = rep('N',readEnd)
338 seqIn <- c(seqIn,seqTemplate[(length(seqIn)+1):readEnd])
339 seqGL <- c(seqGL,seqTemplate[(length(seqGL)+1):readEnd])
340 # if(posToChange!=-1){
341 # seqIn[posToChange] = "-"
342 # seqGL[posToChange] = "-"
343 # }
344
345 seqIn = c2s(seqIn[1:readEnd])
346 seqGL = c2s(seqGL[1:readEnd])
347
348 lenGL = nchar(seqGL)
349 if(lenGL<readEnd){
350 seqGL = paste(seqGL,c2s(rep("N",readEnd-lenGL)),sep="")
351 }
352
353 lenInput = nchar(seqIn)
354 if(lenInput<readEnd){
355 seqIn = paste(seqIn,c2s(rep("N",readEnd-lenInput)),sep="")
356 }
357 return( c(seqIn,seqGL) )
358 }
359
360 replaceLeadingTrailingDashesHelper <- function(x){
361 grepResults = gregexpr("-*",x)
362 grepResultsPos = unlist(grepResults)
363 grepResultsLen = attr(grepResults[[1]],"match.length")
364 #print(paste("x = '", x, "'", sep=""))
365 x = s2c(x)
366 if(x[1]=="-"){
367 x[1:grepResultsLen[1]] = "N"
368 }
369 if(x[length(x)]=="-"){
370 x[(length(x)-grepResultsLen[length(grepResultsLen)]+1):length(x)] = "N"
371 }
372 return(x)
373 }
374
375
376
377
378 # Check sequences for indels
379 checkForInDels <- function(matInputP){
380 insPos <- checkInsertion(matInputP)
381 delPos <- checkDeletions(matInputP)
382 return(list("Insertions"=insPos, "Deletions"=delPos))
383 }
384
385 # Check sequences for insertions
386 checkInsertion <- function(matInputP){
387 insertionCheck = apply( matInputP,1, function(x){
388 inputGaps <- as.vector( gregexpr("-",x[1])[[1]] )
389 glGaps <- as.vector( gregexpr("-",x[2])[[1]] )
390 return( is.finite( match(FALSE, glGaps%in%inputGaps ) ) )
391 })
392 return(as.vector(insertionCheck))
393 }
394 # Fix inserstions
395 fixInsertions <- function(matInputP){
396 insPos <- checkInsertion(matInputP)
397 sapply((1:nrow(matInputP))[insPos],function(rowIndex){
398 x <- matInputP[rowIndex,]
399 inputGaps <- gregexpr("-",x[1])[[1]]
400 glGaps <- gregexpr("-",x[2])[[1]]
401 posInsertions <- glGaps[!(glGaps%in%inputGaps)]
402 inputInsertionToN <- s2c(x[2])
403 inputInsertionToN[posInsertions]!="-"
404 inputInsertionToN[posInsertions] <- "N"
405 inputInsertionToN <- c2s(inputInsertionToN)
406 matInput[rowIndex,2] <<- inputInsertionToN
407 })
408 return(insPos)
409 }
410
411 # Check sequences for deletions
412 checkDeletions <-function(matInputP){
413 deletionCheck = apply( matInputP,1, function(x){
414 inputGaps <- as.vector( gregexpr("-",x[1])[[1]] )
415 glGaps <- as.vector( gregexpr("-",x[2])[[1]] )
416 return( is.finite( match(FALSE, inputGaps%in%glGaps ) ) )
417 })
418 return(as.vector(deletionCheck))
419 }
420 # Fix sequences with deletions
421 fixDeletions <- function(matInputP){
422 delPos <- checkDeletions(matInputP)
423 sapply((1:nrow(matInputP))[delPos],function(rowIndex){
424 x <- matInputP[rowIndex,]
425 inputGaps <- gregexpr("-",x[1])[[1]]
426 glGaps <- gregexpr("-",x[2])[[1]]
427 posDeletions <- inputGaps[!(inputGaps%in%glGaps)]
428 inputDeletionToN <- s2c(x[1])
429 inputDeletionToN[posDeletions] <- "N"
430 inputDeletionToN <- c2s(inputDeletionToN)
431 matInput[rowIndex,1] <<- inputDeletionToN
432 })
433 return(delPos)
434 }
435
436
437 # Trim DNA sequence to the last codon
438 trimToLastCodon <- function(seqToTrim){
439 seqLen = nchar(seqToTrim)
440 trimmedSeq = s2c(seqToTrim)
441 poi = seqLen
442 tailLen = 0
443
444 while(trimmedSeq[poi]=="-" || trimmedSeq[poi]=="."){
445 tailLen = tailLen + 1
446 poi = poi - 1
447 }
448
449 trimmedSeq = c2s(trimmedSeq[1:(seqLen-tailLen)])
450 seqLen = nchar(trimmedSeq)
451 # Trim sequence to last codon
452 if( getCodonPos(seqLen)[3] > seqLen )
453 trimmedSeq = substr(seqToTrim,1, ( (getCodonPos(seqLen)[1])-1 ) )
454
455 return(trimmedSeq)
456 }
457
458 # Given a nuclotide position, returns the pos of the 3 nucs that made the codon
459 # e.g. nuc 86 is part of nucs 85,86,87
460 getCodonPos <- function(nucPos){
461 codonNum = (ceiling(nucPos/3))*3
462 return( (codonNum-2):codonNum)
463 }
464
465 # Given a nuclotide position, returns the codon number
466 # e.g. nuc 86 = codon 29
467 getCodonNumb <- function(nucPos){
468 return( ceiling(nucPos/3) )
469 }
470
471 # Given a codon, returns all the nuc positions that make the codon
472 getCodonNucs <- function(codonNumb){
473 getCodonPos(codonNumb*3)
474 }
475
476 computeCodonTable <- function(testID=1){
477
478 if(testID<=4){
479 # Pre-compute every codons
480 intCounter = 1
481 for(pOne in NUCLEOTIDES){
482 for(pTwo in NUCLEOTIDES){
483 for(pThree in NUCLEOTIDES){
484 codon = paste(pOne,pTwo,pThree,sep="")
485 colnames(CODON_TABLE)[intCounter] = codon
486 intCounter = intCounter + 1
487 CODON_TABLE[,codon] = mutationTypeOptimized(cbind(permutateAllCodon(codon),rep(codon,12)))
488 }
489 }
490 }
491 chars = c("N","A","C","G","T", "-")
492 for(a in chars){
493 for(b in chars){
494 for(c in chars){
495 if(a=="N" | b=="N" | c=="N"){
496 #cat(paste(a,b,c),sep="","\n")
497 CODON_TABLE[,paste(a,b,c,sep="")] = rep(NA,12)
498 }
499 }
500 }
501 }
502
503 chars = c("-","A","C","G","T")
504 for(a in chars){
505 for(b in chars){
506 for(c in chars){
507 if(a=="-" | b=="-" | c=="-"){
508 #cat(paste(a,b,c),sep="","\n")
509 CODON_TABLE[,paste(a,b,c,sep="")] = rep(NA,12)
510 }
511 }
512 }
513 }
514 CODON_TABLE <<- as.matrix(CODON_TABLE)
515 }
516 }
517
518 collapseClone <- function(vecInputSeqs,glSeq,readEnd,nonTerminalOnly=0){
519 #print(length(vecInputSeqs))
520 vecInputSeqs = unique(vecInputSeqs)
521 if(length(vecInputSeqs)==1){
522 return( list( c(vecInputSeqs,glSeq), F) )
523 }else{
524 charInputSeqs <- sapply(vecInputSeqs, function(x){
525 s2c(x)[1:readEnd]
526 })
527 charGLSeq <- s2c(glSeq)
528 matClone <- sapply(1:readEnd, function(i){
529 posNucs = unique(charInputSeqs[i,])
530 posGL = charGLSeq[i]
531 error = FALSE
532 if(posGL=="-" & sum(!(posNucs%in%c("-","N")))==0 ){
533 return(c("-",error))
534 }
535 if(length(posNucs)==1)
536 return(c(posNucs[1],error))
537 else{
538 if("N"%in%posNucs){
539 error=TRUE
540 }
541 if(sum(!posNucs[posNucs!="N"]%in%posGL)==0){
542 return( c(posGL,error) )
543 }else{
544 #return( c(sample(posNucs[posNucs!="N"],1),error) )
545 if(nonTerminalOnly==0){
546 return( c(sample(charInputSeqs[i,charInputSeqs[i,]!="N" & charInputSeqs[i,]!=posGL],1),error) )
547 }else{
548 posNucs = charInputSeqs[i,charInputSeqs[i,]!="N" & charInputSeqs[i,]!=posGL]
549 posNucsTable = table(posNucs)
550 if(sum(posNucsTable>1)==0){
551 return( c(posGL,error) )
552 }else{
553 return( c(sample( posNucs[posNucs%in%names(posNucsTable)[posNucsTable>1]],1),error) )
554 }
555 }
556
557 }
558 }
559 })
560
561
562 #print(length(vecInputSeqs))
563 return(list(c(c2s(matClone[1,]),glSeq),"TRUE"%in%matClone[2,]))
564 }
565 }
566
567 # Compute the expected for each sequence-germline pair
568 getExpectedIndividual <- function(matInput){
569 if( any(grep("multicore",search())) ){
570 facGL <- factor(matInput[,2])
571 facLevels = levels(facGL)
572 LisGLs_MutabilityU = mclapply(1:length(facLevels), function(x){
573 computeMutabilities(facLevels[x])
574 })
575 facIndex = match(facGL,facLevels)
576
577 LisGLs_Mutability = mclapply(1:nrow(matInput), function(x){
578 cInput = rep(NA,nchar(matInput[x,1]))
579 cInput[s2c(matInput[x,1])!="N"] = 1
580 LisGLs_MutabilityU[[facIndex[x]]] * cInput
581 })
582
583 LisGLs_Targeting = mclapply(1:dim(matInput)[1], function(x){
584 computeTargeting(matInput[x,2],LisGLs_Mutability[[x]])
585 })
586
587 LisGLs_MutationTypes = mclapply(1:length(matInput[,2]),function(x){
588 #print(x)
589 computeMutationTypes(matInput[x,2])
590 })
591
592 LisGLs_Exp = mclapply(1:dim(matInput)[1], function(x){
593 computeExpected(LisGLs_Targeting[[x]],LisGLs_MutationTypes[[x]])
594 })
595
596 ul_LisGLs_Exp = unlist(LisGLs_Exp)
597 return(matrix(ul_LisGLs_Exp,ncol=4,nrow=(length(ul_LisGLs_Exp)/4),byrow=T))
598 }else{
599 facGL <- factor(matInput[,2])
600 facLevels = levels(facGL)
601 LisGLs_MutabilityU = lapply(1:length(facLevels), function(x){
602 computeMutabilities(facLevels[x])
603 })
604 facIndex = match(facGL,facLevels)
605
606 LisGLs_Mutability = lapply(1:nrow(matInput), function(x){
607 cInput = rep(NA,nchar(matInput[x,1]))
608 cInput[s2c(matInput[x,1])!="N"] = 1
609 LisGLs_MutabilityU[[facIndex[x]]] * cInput
610 })
611
612 LisGLs_Targeting = lapply(1:dim(matInput)[1], function(x){
613 computeTargeting(matInput[x,2],LisGLs_Mutability[[x]])
614 })
615
616 LisGLs_MutationTypes = lapply(1:length(matInput[,2]),function(x){
617 #print(x)
618 computeMutationTypes(matInput[x,2])
619 })
620
621 LisGLs_Exp = lapply(1:dim(matInput)[1], function(x){
622 computeExpected(LisGLs_Targeting[[x]],LisGLs_MutationTypes[[x]])
623 })
624
625 ul_LisGLs_Exp = unlist(LisGLs_Exp)
626 return(matrix(ul_LisGLs_Exp,ncol=4,nrow=(length(ul_LisGLs_Exp)/4),byrow=T))
627
628 }
629 }
630
631 # Compute mutabilities of sequence based on the tri-nucleotide model
632 computeMutabilities <- function(paramSeq){
633 seqLen = nchar(paramSeq)
634 seqMutabilites = rep(NA,seqLen)
635
636 gaplessSeq = gsub("-", "", paramSeq)
637 gaplessSeqLen = nchar(gaplessSeq)
638 gaplessSeqMutabilites = rep(NA,gaplessSeqLen)
639
640 if(mutabilityModel!=5){
641 pos<- 3:(gaplessSeqLen)
642 subSeq = substr(rep(gaplessSeq,gaplessSeqLen-2),(pos-2),(pos+2))
643 gaplessSeqMutabilites[pos] =
644 tapply( c(
645 getMutability( substr(subSeq,1,3), 3) ,
646 getMutability( substr(subSeq,2,4), 2),
647 getMutability( substr(subSeq,3,5), 1)
648 ),rep(1:(gaplessSeqLen-2),3),mean,na.rm=TRUE
649 )
650 #Pos 1
651 subSeq = substr(gaplessSeq,1,3)
652 gaplessSeqMutabilites[1] = getMutability(subSeq , 1)
653 #Pos 2
654 subSeq = substr(gaplessSeq,1,4)
655 gaplessSeqMutabilites[2] = mean( c(
656 getMutability( substr(subSeq,1,3), 2) ,
657 getMutability( substr(subSeq,2,4), 1)
658 ),na.rm=T
659 )
660 seqMutabilites[which(s2c(paramSeq)!="-")]<- gaplessSeqMutabilites
661 return(seqMutabilites)
662 }else{
663
664 pos<- 3:(gaplessSeqLen)
665 subSeq = substr(rep(gaplessSeq,gaplessSeqLen-2),(pos-2),(pos+2))
666 gaplessSeqMutabilites[pos] = sapply(subSeq,function(x){ getMutability5(x) }, simplify=T)
667 seqMutabilites[which(s2c(paramSeq)!="-")]<- gaplessSeqMutabilites
668 return(seqMutabilites)
669 }
670
671 }
672
673 # Returns the mutability of a triplet at a given position
674 getMutability <- function(codon, pos=1:3){
675 triplets <- rownames(mutability)
676 mutability[ match(codon,triplets) ,pos]
677 }
678
679 getMutability5 <- function(fivemer){
680 return(mutability[fivemer])
681 }
682
683 # Returns the substitution probabilty
684 getTransistionProb <- function(nuc){
685 substitution[nuc,]
686 }
687
688 getTransistionProb5 <- function(fivemer){
689 if(any(which(fivemer==colnames(substitution)))){
690 return(substitution[,fivemer])
691 }else{
692 return(array(NA,4))
693 }
694 }
695
696 # Given a nuc, returns the other 3 nucs it can mutate to
697 canMutateTo <- function(nuc){
698 NUCLEOTIDES[- which(NUCLEOTIDES==nuc)]
699 }
700
701 # Given a nucleotide, returns the probabilty of other nucleotide it can mutate to
702 canMutateToProb <- function(nuc){
703 substitution[nuc,canMutateTo(nuc)]
704 }
705
706 # Compute targeting, based on precomputed mutatbility & substitution
707 computeTargeting <- function(param_strSeq,param_vecMutabilities){
708
709 if(substitutionModel!=5){
710 vecSeq = s2c(param_strSeq)
711 matTargeting = sapply( 1:length(vecSeq), function(x) { param_vecMutabilities[x] * getTransistionProb(vecSeq[x]) } )
712 #matTargeting = apply( rbind(vecSeq,param_vecMutabilities),2, function(x) { as.vector(as.numeric(x[2]) * getTransistionProb(x[1])) } )
713 dimnames( matTargeting ) = list(NUCLEOTIDES,1:(length(vecSeq)))
714 return (matTargeting)
715 }else{
716
717 seqLen = nchar(param_strSeq)
718 seqsubstitution = matrix(NA,ncol=seqLen,nrow=4)
719 paramSeq <- param_strSeq
720 gaplessSeq = gsub("-", "", paramSeq)
721 gaplessSeqLen = nchar(gaplessSeq)
722 gaplessSeqSubstitution = matrix(NA,ncol=gaplessSeqLen,nrow=4)
723
724 pos<- 3:(gaplessSeqLen)
725 subSeq = substr(rep(gaplessSeq,gaplessSeqLen-2),(pos-2),(pos+2))
726 gaplessSeqSubstitution[,pos] = sapply(subSeq,function(x){ getTransistionProb5(x) }, simplify=T)
727 seqsubstitution[,which(s2c(paramSeq)!="-")]<- gaplessSeqSubstitution
728 #matTargeting <- param_vecMutabilities %*% seqsubstitution
729 matTargeting <- sweep(seqsubstitution,2,param_vecMutabilities,`*`)
730 dimnames( matTargeting ) = list(NUCLEOTIDES,1:(seqLen))
731 return (matTargeting)
732 }
733 }
734
735 # Compute the mutations types
736 computeMutationTypes <- function(param_strSeq){
737 #cat(param_strSeq,"\n")
738 #vecSeq = trimToLastCodon(param_strSeq)
739 lenSeq = nchar(param_strSeq)
740 vecCodons = sapply({1:(lenSeq/3)}*3-2,function(x){substr(param_strSeq,x,x+2)})
741 matMutationTypes = matrix( unlist(CODON_TABLE[,vecCodons]) ,ncol=lenSeq,nrow=4, byrow=F)
742 dimnames( matMutationTypes ) = list(NUCLEOTIDES,1:(ncol(matMutationTypes)))
743 return(matMutationTypes)
744 }
745 computeMutationTypesFast <- function(param_strSeq){
746 matMutationTypes = matrix( CODON_TABLE[,param_strSeq] ,ncol=3,nrow=4, byrow=F)
747 #dimnames( matMutationTypes ) = list(NUCLEOTIDES,1:(length(vecSeq)))
748 return(matMutationTypes)
749 }
750 mutationTypeOptimized <- function( matOfCodons ){
751 apply( matOfCodons,1,function(x){ mutationType(x[2],x[1]) } )
752 }
753
754 # Returns a vector of codons 1 mutation away from the given codon
755 permutateAllCodon <- function(codon){
756 cCodon = s2c(codon)
757 matCodons = t(array(cCodon,dim=c(3,12)))
758 matCodons[1:4,1] = NUCLEOTIDES
759 matCodons[5:8,2] = NUCLEOTIDES
760 matCodons[9:12,3] = NUCLEOTIDES
761 apply(matCodons,1,c2s)
762 }
763
764 # Given two codons, tells you if the mutation is R or S (based on your definition)
765 mutationType <- function(codonFrom,codonTo){
766 if(testID==4){
767 if( is.na(codonFrom) | is.na(codonTo) | is.na(translateCodonToAminoAcid(codonFrom)) | is.na(translateCodonToAminoAcid(codonTo)) ){
768 return(NA)
769 }else{
770 mutationType = "S"
771 if( translateAminoAcidToTraitChange(translateCodonToAminoAcid(codonFrom)) != translateAminoAcidToTraitChange(translateCodonToAminoAcid(codonTo)) ){
772 mutationType = "R"
773 }
774 if(translateCodonToAminoAcid(codonTo)=="*" | translateCodonToAminoAcid(codonFrom)=="*"){
775 mutationType = "Stop"
776 }
777 return(mutationType)
778 }
779 }else if(testID==5){
780 if( is.na(codonFrom) | is.na(codonTo) | is.na(translateCodonToAminoAcid(codonFrom)) | is.na(translateCodonToAminoAcid(codonTo)) ){
781 return(NA)
782 }else{
783 if(codonFrom==codonTo){
784 mutationType = "S"
785 }else{
786 codonFrom = s2c(codonFrom)
787 codonTo = s2c(codonTo)
788 mutationType = "Stop"
789 nucOfI = codonFrom[which(codonTo!=codonFrom)]
790 if(nucOfI=="C"){
791 mutationType = "R"
792 }else if(nucOfI=="G"){
793 mutationType = "S"
794 }
795 }
796 return(mutationType)
797 }
798 }else{
799 if( is.na(codonFrom) | is.na(codonTo) | is.na(translateCodonToAminoAcid(codonFrom)) | is.na(translateCodonToAminoAcid(codonTo)) ){
800 return(NA)
801 }else{
802 mutationType = "S"
803 if( translateCodonToAminoAcid(codonFrom) != translateCodonToAminoAcid(codonTo) ){
804 mutationType = "R"
805 }
806 if(translateCodonToAminoAcid(codonTo)=="*" | translateCodonToAminoAcid(codonFrom)=="*"){
807 mutationType = "Stop"
808 }
809 return(mutationType)
810 }
811 }
812 }
813
814
815 #given a mat of targeting & it's corresponding mutationtypes returns
816 #a vector of Exp_RCDR,Exp_SCDR,Exp_RFWR,Exp_RFWR
817 computeExpected <- function(paramTargeting,paramMutationTypes){
818 # Replacements
819 RPos = which(paramMutationTypes=="R")
820 #FWR
821 Exp_R_FWR = sum(paramTargeting[ RPos[which(FWR_Nuc_Mat[RPos]==T)] ],na.rm=T)
822 #CDR
823 Exp_R_CDR = sum(paramTargeting[ RPos[which(CDR_Nuc_Mat[RPos]==T)] ],na.rm=T)
824 # Silents
825 SPos = which(paramMutationTypes=="S")
826 #FWR
827 Exp_S_FWR = sum(paramTargeting[ SPos[which(FWR_Nuc_Mat[SPos]==T)] ],na.rm=T)
828 #CDR
829 Exp_S_CDR = sum(paramTargeting[ SPos[which(CDR_Nuc_Mat[SPos]==T)] ],na.rm=T)
830
831 return(c(Exp_R_CDR,Exp_S_CDR,Exp_R_FWR,Exp_S_FWR))
832 }
833
834 # Count the mutations in a sequence
835 # each mutation is treated independently
836 analyzeMutations2NucUri_website <- function( rev_in_matrix ){
837 paramGL = rev_in_matrix[2,]
838 paramSeq = rev_in_matrix[1,]
839
840 #Fill seq with GL seq if gapped
841 #if( any(paramSeq=="-") ){
842 # gapPos_Seq = which(paramSeq=="-")
843 # gapPos_Seq_ToReplace = gapPos_Seq[paramGL[gapPos_Seq] != "-"]
844 # paramSeq[gapPos_Seq_ToReplace] = paramGL[gapPos_Seq_ToReplace]
845 #}
846
847
848 #if( any(paramSeq=="N") ){
849 # gapPos_Seq = which(paramSeq=="N")
850 # gapPos_Seq_ToReplace = gapPos_Seq[paramGL[gapPos_Seq] != "N"]
851 # paramSeq[gapPos_Seq_ToReplace] = paramGL[gapPos_Seq_ToReplace]
852 #}
853
854 analyzeMutations2NucUri( matrix(c( paramGL, paramSeq ),2,length(paramGL),byrow=T) )
855
856 }
857
858 #1 = GL
859 #2 = Seq
860 analyzeMutations2NucUri <- function( in_matrix=matrix(c(c("A","A","A","C","C","C"),c("A","G","G","C","C","A")),2,6,byrow=T) ){
861 paramGL = in_matrix[2,]
862 paramSeq = in_matrix[1,]
863 paramSeqUri = paramGL
864 #mutations = apply(rbind(paramGL,paramSeq), 2, function(x){!x[1]==x[2]})
865 mutations_val = paramGL != paramSeq
866 if(any(mutations_val)){
867 mutationPos = {1:length(mutations_val)}[mutations_val]
868 mutationPos = mutationPos[sapply(mutationPos, function(x){!any(paramSeq[getCodonPos(x)]=="N")})]
869 length_mutations =length(mutationPos)
870 mutationInfo = rep(NA,length_mutations)
871 if(any(mutationPos)){
872
873 pos<- mutationPos
874 pos_array<-array(sapply(pos,getCodonPos))
875 codonGL = paramGL[pos_array]
876
877 codonSeq = sapply(pos,function(x){
878 seqP = paramGL[getCodonPos(x)]
879 muCodonPos = {x-1}%%3+1
880 seqP[muCodonPos] = paramSeq[x]
881 return(seqP)
882 })
883 GLcodons = apply(matrix(codonGL,length_mutations,3,byrow=TRUE),1,c2s)
884 Seqcodons = apply(codonSeq,2,c2s)
885 mutationInfo = apply(rbind(GLcodons , Seqcodons),2,function(x){mutationType(c2s(x[1]),c2s(x[2]))})
886 names(mutationInfo) = mutationPos
887 }
888 if(any(!is.na(mutationInfo))){
889 return(mutationInfo[!is.na(mutationInfo)])
890 }else{
891 return(NA)
892 }
893
894
895 }else{
896 return (NA)
897 }
898 }
899
900 processNucMutations2 <- function(mu){
901 if(!is.na(mu)){
902 #R
903 if(any(mu=="R")){
904 Rs = mu[mu=="R"]
905 nucNumbs = as.numeric(names(Rs))
906 R_CDR = sum(as.integer(CDR_Nuc[nucNumbs]),na.rm=T)
907 R_FWR = sum(as.integer(FWR_Nuc[nucNumbs]),na.rm=T)
908 }else{
909 R_CDR = 0
910 R_FWR = 0
911 }
912
913 #S
914 if(any(mu=="S")){
915 Ss = mu[mu=="S"]
916 nucNumbs = as.numeric(names(Ss))
917 S_CDR = sum(as.integer(CDR_Nuc[nucNumbs]),na.rm=T)
918 S_FWR = sum(as.integer(FWR_Nuc[nucNumbs]),na.rm=T)
919 }else{
920 S_CDR = 0
921 S_FWR = 0
922 }
923
924
925 retVec = c(R_CDR,S_CDR,R_FWR,S_FWR)
926 retVec[is.na(retVec)]=0
927 return(retVec)
928 }else{
929 return(rep(0,4))
930 }
931 }
932
933
934 ## Z-score Test
935 computeZScore <- function(mat, test="Focused"){
936 matRes <- matrix(NA,ncol=2,nrow=(nrow(mat)))
937 if(test=="Focused"){
938 #Z_Focused_CDR
939 #P_Denom = sum( mat[1,c(5,6,8)], na.rm=T )
940 P = apply(mat[,c(5,6,8)],1,function(x){(x[1]/sum(x))})
941 R_mean = apply(cbind(mat[,c(1,2,4)],P),1,function(x){x[4]*(sum(x[1:3]))})
942 R_sd=sqrt(R_mean*(1-P))
943 matRes[,1] = (mat[,1]-R_mean)/R_sd
944
945 #Z_Focused_FWR
946 #P_Denom = sum( mat[1,c(7,6,8)], na.rm=T )
947 P = apply(mat[,c(7,6,8)],1,function(x){(x[1]/sum(x))})
948 R_mean = apply(cbind(mat[,c(3,2,4)],P),1,function(x){x[4]*(sum(x[1:3]))})
949 R_sd=sqrt(R_mean*(1-P))
950 matRes[,2] = (mat[,3]-R_mean)/R_sd
951 }
952
953 if(test=="Local"){
954 #Z_Focused_CDR
955 #P_Denom = sum( mat[1,c(5,6,8)], na.rm=T )
956 P = apply(mat[,c(5,6)],1,function(x){(x[1]/sum(x))})
957 R_mean = apply(cbind(mat[,c(1,2)],P),1,function(x){x[3]*(sum(x[1:2]))})
958 R_sd=sqrt(R_mean*(1-P))
959 matRes[,1] = (mat[,1]-R_mean)/R_sd
960
961 #Z_Focused_FWR
962 #P_Denom = sum( mat[1,c(7,6,8)], na.rm=T )
963 P = apply(mat[,c(7,8)],1,function(x){(x[1]/sum(x))})
964 R_mean = apply(cbind(mat[,c(3,4)],P),1,function(x){x[3]*(sum(x[1:2]))})
965 R_sd=sqrt(R_mean*(1-P))
966 matRes[,2] = (mat[,3]-R_mean)/R_sd
967 }
968
969 if(test=="Imbalanced"){
970 #Z_Focused_CDR
971 #P_Denom = sum( mat[1,c(5,6,8)], na.rm=T )
972 P = apply(mat[,5:8],1,function(x){((x[1]+x[2])/sum(x))})
973 R_mean = apply(cbind(mat[,1:4],P),1,function(x){x[5]*(sum(x[1:4]))})
974 R_sd=sqrt(R_mean*(1-P))
975 matRes[,1] = (mat[,1]-R_mean)/R_sd
976
977 #Z_Focused_FWR
978 #P_Denom = sum( mat[1,c(7,6,8)], na.rm=T )
979 P = apply(mat[,5:8],1,function(x){((x[3]+x[4])/sum(x))})
980 R_mean = apply(cbind(mat[,1:4],P),1,function(x){x[5]*(sum(x[1:4]))})
981 R_sd=sqrt(R_mean*(1-P))
982 matRes[,2] = (mat[,3]-R_mean)/R_sd
983 }
984
985 matRes[is.nan(matRes)] = NA
986 return(matRes)
987 }
988
989 # Return a p-value for a z-score
990 z2p <- function(z){
991 p=NA
992 if( !is.nan(z) && !is.na(z)){
993 if(z>0){
994 p = (1 - pnorm(z,0,1))
995 } else if(z<0){
996 p = (-1 * pnorm(z,0,1))
997 } else{
998 p = 0.5
999 }
1000 }else{
1001 p = NA
1002 }
1003 return(p)
1004 }
1005
1006
1007 ## Bayesian Test
1008
1009 # Fitted parameter for the bayesian framework
1010 BAYESIAN_FITTED<-c(0.407277142798302, 0.554007336744485, 0.63777155771234, 0.693989162719009, 0.735450014674917, 0.767972534429806, 0.794557287143399, 0.816906816601605, 0.83606796225341, 0.852729446430296, 0.867370424541641, 0.880339760590323, 0.891900995024999, 0.902259181289864, 0.911577919359,0.919990301665853, 0.927606458124537, 0.934518806350661, 0.940805863754375, 0.946534836475715, 0.951763691199255, 0.95654428191308, 0.960920179487397, 0.964930893680829, 0.968611312149038, 0.971992459313836, 0.975102110004818, 0.977964943023096, 0.980603428208439, 0.983037660179428, 0.985285800977406, 0.987364285326685, 0.989288037855441, 0.991070478823525, 0.992723699729969, 0.994259575477392, 0.995687688867975, 0.997017365051493, 0.998257085153047, 0.999414558305388, 1.00049681357804, 1.00151036237481, 1.00246080204981, 1.00335370751909, 1.0041939329768, 1.0049859393417, 1.00573382091263, 1.00644127217376, 1.00711179729107, 1.00774845526417, 1.00835412715854, 1.00893143010366, 1.00948275846309, 1.01001030293661, 1.01051606798079, 1.01100188771288, 1.01146944044216, 1.01192026195449, 1.01235575766094, 1.01277721370986)
1011 CONST_i <- sort(c(((2^(seq(-39,0,length.out=201)))/2)[1:200],(c(0:11,13:99)+0.5)/100,1-(2^(seq(-39,0,length.out=201)))/2))
1012
1013 # Given x, M & p, returns a pdf
1014 calculate_bayes <- function ( x=3, N=10, p=0.33,
1015 i=CONST_i,
1016 max_sigma=20,length_sigma=4001
1017 ){
1018 if(!0%in%N){
1019 G <- max(length(x),length(N),length(p))
1020 x=array(x,dim=G)
1021 N=array(N,dim=G)
1022 p=array(p,dim=G)
1023 sigma_s<-seq(-max_sigma,max_sigma,length.out=length_sigma)
1024 sigma_1<-log({i/{1-i}}/{p/{1-p}})
1025 index<-min(N,60)
1026 y<-dbeta(i,x+BAYESIAN_FITTED[index],N+BAYESIAN_FITTED[index]-x)*(1-p)*p*exp(sigma_1)/({1-p}^2+2*p*{1-p}*exp(sigma_1)+{p^2}*exp(2*sigma_1))
1027 if(!sum(is.na(y))){
1028 tmp<-approx(sigma_1,y,sigma_s)$y
1029 tmp/sum(tmp)/{2*max_sigma/{length_sigma-1}}
1030 }else{
1031 return(NA)
1032 }
1033 }else{
1034 return(NA)
1035 }
1036 }
1037 # Given a mat of observed & expected, return a list of CDR & FWR pdf for selection
1038 computeBayesianScore <- function(mat, test="Focused", max_sigma=20,length_sigma=4001){
1039 flagOneSeq = F
1040 if(nrow(mat)==1){
1041 mat=rbind(mat,mat)
1042 flagOneSeq = T
1043 }
1044 if(test=="Focused"){
1045 #CDR
1046 P = c(apply(mat[,c(5,6,8)],1,function(x){(x[1]/sum(x))}),0.5)
1047 N = c(apply(mat[,c(1,2,4)],1,function(x){(sum(x))}),0)
1048 X = c(mat[,1],0)
1049 bayesCDR = apply(cbind(X,N,P),1,function(x){calculate_bayes(x=x[1],N=x[2],p=x[3],max_sigma=max_sigma,length_sigma=length_sigma)})
1050 bayesCDR = bayesCDR[-length(bayesCDR)]
1051
1052 #FWR
1053 P = c(apply(mat[,c(7,6,8)],1,function(x){(x[1]/sum(x))}),0.5)
1054 N = c(apply(mat[,c(3,2,4)],1,function(x){(sum(x))}),0)
1055 X = c(mat[,3],0)
1056 bayesFWR = apply(cbind(X,N,P),1,function(x){calculate_bayes(x=x[1],N=x[2],p=x[3],max_sigma=max_sigma,length_sigma=length_sigma)})
1057 bayesFWR = bayesFWR[-length(bayesFWR)]
1058 }
1059
1060 if(test=="Local"){
1061 #CDR
1062 P = c(apply(mat[,c(5,6)],1,function(x){(x[1]/sum(x))}),0.5)
1063 N = c(apply(mat[,c(1,2)],1,function(x){(sum(x))}),0)
1064 X = c(mat[,1],0)
1065 bayesCDR = apply(cbind(X,N,P),1,function(x){calculate_bayes(x=x[1],N=x[2],p=x[3],max_sigma=max_sigma,length_sigma=length_sigma)})
1066 bayesCDR = bayesCDR[-length(bayesCDR)]
1067
1068 #FWR
1069 P = c(apply(mat[,c(7,8)],1,function(x){(x[1]/sum(x))}),0.5)
1070 N = c(apply(mat[,c(3,4)],1,function(x){(sum(x))}),0)
1071 X = c(mat[,3],0)
1072 bayesFWR = apply(cbind(X,N,P),1,function(x){calculate_bayes(x=x[1],N=x[2],p=x[3],max_sigma=max_sigma,length_sigma=length_sigma)})
1073 bayesFWR = bayesFWR[-length(bayesFWR)]
1074 }
1075
1076 if(test=="Imbalanced"){
1077 #CDR
1078 P = c(apply(mat[,c(5:8)],1,function(x){((x[1]+x[2])/sum(x))}),0.5)
1079 N = c(apply(mat[,c(1:4)],1,function(x){(sum(x))}),0)
1080 X = c(apply(mat[,c(1:2)],1,function(x){(sum(x))}),0)
1081 bayesCDR = apply(cbind(X,N,P),1,function(x){calculate_bayes(x=x[1],N=x[2],p=x[3],max_sigma=max_sigma,length_sigma=length_sigma)})
1082 bayesCDR = bayesCDR[-length(bayesCDR)]
1083
1084 #FWR
1085 P = c(apply(mat[,c(5:8)],1,function(x){((x[3]+x[4])/sum(x))}),0.5)
1086 N = c(apply(mat[,c(1:4)],1,function(x){(sum(x))}),0)
1087 X = c(apply(mat[,c(3:4)],1,function(x){(sum(x))}),0)
1088 bayesFWR = apply(cbind(X,N,P),1,function(x){calculate_bayes(x=x[1],N=x[2],p=x[3],max_sigma=max_sigma,length_sigma=length_sigma)})
1089 bayesFWR = bayesFWR[-length(bayesFWR)]
1090 }
1091
1092 if(test=="ImbalancedSilent"){
1093 #CDR
1094 P = c(apply(mat[,c(6,8)],1,function(x){((x[1])/sum(x))}),0.5)
1095 N = c(apply(mat[,c(2,4)],1,function(x){(sum(x))}),0)
1096 X = c(apply(mat[,c(2,4)],1,function(x){(x[1])}),0)
1097 bayesCDR = apply(cbind(X,N,P),1,function(x){calculate_bayes(x=x[1],N=x[2],p=x[3],max_sigma=max_sigma,length_sigma=length_sigma)})
1098 bayesCDR = bayesCDR[-length(bayesCDR)]
1099
1100 #FWR
1101 P = c(apply(mat[,c(6,8)],1,function(x){((x[2])/sum(x))}),0.5)
1102 N = c(apply(mat[,c(2,4)],1,function(x){(sum(x))}),0)
1103 X = c(apply(mat[,c(2,4)],1,function(x){(x[2])}),0)
1104 bayesFWR = apply(cbind(X,N,P),1,function(x){calculate_bayes(x=x[1],N=x[2],p=x[3],max_sigma=max_sigma,length_sigma=length_sigma)})
1105 bayesFWR = bayesFWR[-length(bayesFWR)]
1106 }
1107
1108 if(flagOneSeq==T){
1109 bayesCDR = bayesCDR[1]
1110 bayesFWR = bayesFWR[1]
1111 }
1112 return( list("CDR"=bayesCDR, "FWR"=bayesFWR) )
1113 }
1114
1115 ##Covolution
1116 break2chunks<-function(G=1000){
1117 base<-2^round(log(sqrt(G),2),0)
1118 return(c(rep(base,floor(G/base)-1),base+G-(floor(G/base)*base)))
1119 }
1120
1121 PowersOfTwo <- function(G=100){
1122 exponents <- array()
1123 i = 0
1124 while(G > 0){
1125 i=i+1
1126 exponents[i] <- floor( log2(G) )
1127 G <- G-2^exponents[i]
1128 }
1129 return(exponents)
1130 }
1131
1132 convolutionPowersOfTwo <- function( cons, length_sigma=4001 ){
1133 G = ncol(cons)
1134 if(G>1){
1135 for(gen in log(G,2):1){
1136 ll<-seq(from=2,to=2^gen,by=2)
1137 sapply(ll,function(l){cons[,l/2]<<-weighted_conv(cons[,l],cons[,l-1],length_sigma=length_sigma)})
1138 }
1139 }
1140 return( cons[,1] )
1141 }
1142
1143 convolutionPowersOfTwoByTwos <- function( cons, length_sigma=4001,G=1 ){
1144 if(length(ncol(cons))) G<-ncol(cons)
1145 groups <- PowersOfTwo(G)
1146 matG <- matrix(NA, ncol=length(groups), nrow=length(cons)/G )
1147 startIndex = 1
1148 for( i in 1:length(groups) ){
1149 stopIndex <- 2^groups[i] + startIndex - 1
1150 if(stopIndex!=startIndex){
1151 matG[,i] <- convolutionPowersOfTwo( cons[,startIndex:stopIndex], length_sigma=length_sigma )
1152 startIndex = stopIndex + 1
1153 }
1154 else {
1155 if(G>1) matG[,i] <- cons[,startIndex:stopIndex]
1156 else matG[,i] <- cons
1157 #startIndex = stopIndex + 1
1158 }
1159 }
1160 return( list( matG, groups ) )
1161 }
1162
1163 weighted_conv<-function(x,y,w=1,m=100,length_sigma=4001){
1164 lx<-length(x)
1165 ly<-length(y)
1166 if({lx<m}| {{lx*w}<m}| {{ly}<m}| {{ly*w}<m}){
1167 if(w<1){
1168 y1<-approx(1:ly,y,seq(1,ly,length.out=m))$y
1169 x1<-approx(1:lx,x,seq(1,lx,length.out=m/w))$y
1170 lx<-length(x1)
1171 ly<-length(y1)
1172 }
1173 else {
1174 y1<-approx(1:ly,y,seq(1,ly,length.out=m*w))$y
1175 x1<-approx(1:lx,x,seq(1,lx,length.out=m))$y
1176 lx<-length(x1)
1177 ly<-length(y1)
1178 }
1179 }
1180 else{
1181 x1<-x
1182 y1<-approx(1:ly,y,seq(1,ly,length.out=floor(lx*w)))$y
1183 ly<-length(y1)
1184 }
1185 tmp<-approx(x=1:(lx+ly-1),y=convolve(x1,rev(y1),type="open"),xout=seq(1,lx+ly-1,length.out=length_sigma))$y
1186 tmp[tmp<=0] = 0
1187 return(tmp/sum(tmp))
1188 }
1189
1190 calculate_bayesGHelper <- function( listMatG,length_sigma=4001 ){
1191 matG <- listMatG[[1]]
1192 groups <- listMatG[[2]]
1193 i = 1
1194 resConv <- matG[,i]
1195 denom <- 2^groups[i]
1196 if(length(groups)>1){
1197 while( i<length(groups) ){
1198 i = i + 1
1199 resConv <- weighted_conv(resConv, matG[,i], w= {{2^groups[i]}/denom} ,length_sigma=length_sigma)
1200 #cat({{2^groups[i]}/denom},"\n")
1201 denom <- denom + 2^groups[i]
1202 }
1203 }
1204 return(resConv)
1205 }
1206
1207 # Given a list of PDFs, returns a convoluted PDF
1208 groupPosteriors <- function( listPosteriors, max_sigma=20, length_sigma=4001 ,Threshold=2 ){
1209 listPosteriors = listPosteriors[ !is.na(listPosteriors) ]
1210 Length_Postrior<-length(listPosteriors)
1211 if(Length_Postrior>1 & Length_Postrior<=Threshold){
1212 cons = matrix(unlist(listPosteriors),length(listPosteriors[[1]]),length(listPosteriors))
1213 listMatG <- convolutionPowersOfTwoByTwos(cons,length_sigma=length_sigma)
1214 y<-calculate_bayesGHelper(listMatG,length_sigma=length_sigma)
1215 return( y/sum(y)/(2*max_sigma/(length_sigma-1)) )
1216 }else if(Length_Postrior==1) return(listPosteriors[[1]])
1217 else if(Length_Postrior==0) return(NA)
1218 else {
1219 cons = matrix(unlist(listPosteriors),length(listPosteriors[[1]]),length(listPosteriors))
1220 y = fastConv(cons,max_sigma=max_sigma, length_sigma=length_sigma )
1221 return( y/sum(y)/(2*max_sigma/(length_sigma-1)) )
1222 }
1223 }
1224
1225 fastConv<-function(cons, max_sigma=20, length_sigma=4001){
1226 chunks<-break2chunks(G=ncol(cons))
1227 if(ncol(cons)==3) chunks<-2:1
1228 index_chunks_end <- cumsum(chunks)
1229 index_chunks_start <- c(1,index_chunks_end[-length(index_chunks_end)]+1)
1230 index_chunks <- cbind(index_chunks_start,index_chunks_end)
1231
1232 case <- sum(chunks!=chunks[1])
1233 if(case==1) End <- max(1,((length(index_chunks)/2)-1))
1234 else End <- max(1,((length(index_chunks)/2)))
1235
1236 firsts <- sapply(1:End,function(i){
1237 indexes<-index_chunks[i,1]:index_chunks[i,2]
1238 convolutionPowersOfTwoByTwos(cons[ ,indexes])[[1]]
1239 })
1240 if(case==0){
1241 result<-calculate_bayesGHelper( convolutionPowersOfTwoByTwos(firsts) )
1242 }else if(case==1){
1243 last<-list(calculate_bayesGHelper(
1244 convolutionPowersOfTwoByTwos( cons[ ,index_chunks[length(index_chunks)/2,1]:index_chunks[length(index_chunks)/2,2]] )
1245 ),0)
1246 result_first<-calculate_bayesGHelper(convolutionPowersOfTwoByTwos(firsts))
1247 result<-calculate_bayesGHelper(
1248 list(
1249 cbind(
1250 result_first,last[[1]]),
1251 c(log(index_chunks_end[length(index_chunks)/2-1],2),log(index_chunks[length(index_chunks)/2,2]-index_chunks[length(index_chunks)/2,1]+1,2))
1252 )
1253 )
1254 }
1255 return(as.vector(result))
1256 }
1257
1258 # Computes the 95% CI for a pdf
1259 calcBayesCI <- function(Pdf,low=0.025,up=0.975,max_sigma=20, length_sigma=4001){
1260 if(length(Pdf)!=length_sigma) return(NA)
1261 sigma_s=seq(-max_sigma,max_sigma,length.out=length_sigma)
1262 cdf = cumsum(Pdf)
1263 cdf = cdf/cdf[length(cdf)]
1264 return( c(sigma_s[findInterval(low,cdf)-1] , sigma_s[findInterval(up,cdf)]) )
1265 }
1266
1267 # Computes a mean for a pdf
1268 calcBayesMean <- function(Pdf,max_sigma=20,length_sigma=4001){
1269 if(length(Pdf)!=length_sigma) return(NA)
1270 sigma_s=seq(-max_sigma,max_sigma,length.out=length_sigma)
1271 norm = {length_sigma-1}/2/max_sigma
1272 return( (Pdf%*%sigma_s/norm) )
1273 }
1274
1275 # Returns the mean, and the 95% CI for a pdf
1276 calcBayesOutputInfo <- function(Pdf,low=0.025,up=0.975,max_sigma=20, length_sigma=4001){
1277 if(is.na(Pdf))
1278 return(rep(NA,3))
1279 bCI = calcBayesCI(Pdf=Pdf,low=low,up=up,max_sigma=max_sigma,length_sigma=length_sigma)
1280 bMean = calcBayesMean(Pdf=Pdf,max_sigma=max_sigma,length_sigma=length_sigma)
1281 return(c(bMean, bCI))
1282 }
1283
1284 # Computes the p-value of a pdf
1285 computeSigmaP <- function(Pdf, length_sigma=4001, max_sigma=20){
1286 if(length(Pdf)>1){
1287 norm = {length_sigma-1}/2/max_sigma
1288 pVal = {sum(Pdf[1:{{length_sigma-1}/2}]) + Pdf[{{length_sigma+1}/2}]/2}/norm
1289 if(pVal>0.5){
1290 pVal = pVal-1
1291 }
1292 return(pVal)
1293 }else{
1294 return(NA)
1295 }
1296 }
1297
1298 # Compute p-value of two distributions
1299 compareTwoDistsFaster <-function(sigma_S=seq(-20,20,length.out=4001), N=10000, dens1=runif(4001,0,1), dens2=runif(4001,0,1)){
1300 #print(c(length(dens1),length(dens2)))
1301 if(length(dens1)>1 & length(dens2)>1 ){
1302 dens1<-dens1/sum(dens1)
1303 dens2<-dens2/sum(dens2)
1304 cum2 <- cumsum(dens2)-dens2/2
1305 tmp<- sum(sapply(1:length(dens1),function(i)return(dens1[i]*cum2[i])))
1306 #print(tmp)
1307 if(tmp>0.5)tmp<-tmp-1
1308 return( tmp )
1309 }
1310 else {
1311 return(NA)
1312 }
1313 #return (sum(sapply(1:N,function(i)(sample(sigma_S,1,prob=dens1)>sample(sigma_S,1,prob=dens2))))/N)
1314 }
1315
1316 # get number of seqeunces contributing to the sigma (i.e. seqeunces with mutations)
1317 numberOfSeqsWithMutations <- function(matMutations,test=1){
1318 if(test==4)test=2
1319 cdrSeqs <- 0
1320 fwrSeqs <- 0
1321 if(test==1){#focused
1322 cdrMutations <- apply(matMutations, 1, function(x){ sum(x[c(1,2,4)]) })
1323 fwrMutations <- apply(matMutations, 1, function(x){ sum(x[c(3,4,2)]) })
1324 if( any(which(cdrMutations>0)) ) cdrSeqs <- sum(cdrMutations>0)
1325 if( any(which(fwrMutations>0)) ) fwrSeqs <- sum(fwrMutations>0)
1326 }
1327 if(test==2){#local
1328 cdrMutations <- apply(matMutations, 1, function(x){ sum(x[c(1,2)]) })
1329 fwrMutations <- apply(matMutations, 1, function(x){ sum(x[c(3,4)]) })
1330 if( any(which(cdrMutations>0)) ) cdrSeqs <- sum(cdrMutations>0)
1331 if( any(which(fwrMutations>0)) ) fwrSeqs <- sum(fwrMutations>0)
1332 }
1333 return(c("CDR"=cdrSeqs, "FWR"=fwrSeqs))
1334 }
1335
1336
1337
1338 shadeColor <- function(sigmaVal=NA,pVal=NA){
1339 if(is.na(sigmaVal) & is.na(pVal)) return(NA)
1340 if(is.na(sigmaVal) & !is.na(pVal)) sigmaVal=sign(pVal)
1341 if(is.na(pVal) || pVal==1 || pVal==0){
1342 returnColor = "#FFFFFF";
1343 }else{
1344 colVal=abs(pVal);
1345
1346 if(sigmaVal<0){
1347 if(colVal>0.1)
1348 returnColor = "#CCFFCC";
1349 if(colVal<=0.1)
1350 returnColor = "#99FF99";
1351 if(colVal<=0.050)
1352 returnColor = "#66FF66";
1353 if(colVal<=0.010)
1354 returnColor = "#33FF33";
1355 if(colVal<=0.005)
1356 returnColor = "#00FF00";
1357
1358 }else{
1359 if(colVal>0.1)
1360 returnColor = "#FFCCCC";
1361 if(colVal<=0.1)
1362 returnColor = "#FF9999";
1363 if(colVal<=0.05)
1364 returnColor = "#FF6666";
1365 if(colVal<=0.01)
1366 returnColor = "#FF3333";
1367 if(colVal<0.005)
1368 returnColor = "#FF0000";
1369 }
1370 }
1371
1372 return(returnColor)
1373 }
1374
1375
1376
1377 plotHelp <- function(xfrac=0.05,yfrac=0.05,log=FALSE){
1378 if(!log){
1379 x = par()$usr[1]-(par()$usr[2]-par()$usr[1])*xfrac
1380 y = par()$usr[4]+(par()$usr[4]-par()$usr[3])*yfrac
1381 }else {
1382 if(log==2){
1383 x = par()$usr[1]-(par()$usr[2]-par()$usr[1])*xfrac
1384 y = 10^((par()$usr[4])+((par()$usr[4])-(par()$usr[3]))*yfrac)
1385 }
1386 if(log==1){
1387 x = 10^((par()$usr[1])-((par()$usr[2])-(par()$usr[1]))*xfrac)
1388 y = par()$usr[4]+(par()$usr[4]-par()$usr[3])*yfrac
1389 }
1390 if(log==3){
1391 x = 10^((par()$usr[1])-((par()$usr[2])-(par()$usr[1]))*xfrac)
1392 y = 10^((par()$usr[4])+((par()$usr[4])-(par()$usr[3]))*yfrac)
1393 }
1394 }
1395 return(c("x"=x,"y"=y))
1396 }
1397
1398 # SHMulation
1399
1400 # Based on targeting, introduce a single mutation & then update the targeting
1401 oneMutation <- function(){
1402 # Pick a postion + mutation
1403 posMutation = sample(1:(seqGermlineLen*4),1,replace=F,prob=as.vector(seqTargeting))
1404 posNucNumb = ceiling(posMutation/4) # Nucleotide number
1405 posNucKind = 4 - ( (posNucNumb*4) - posMutation ) # Nuc the position mutates to
1406
1407 #mutate the simulation sequence
1408 seqSimVec <- s2c(seqSim)
1409 seqSimVec[posNucNumb] <- NUCLEOTIDES[posNucKind]
1410 seqSim <<- c2s(seqSimVec)
1411
1412 #update Mutability, Targeting & MutationsTypes
1413 updateMutabilityNTargeting(posNucNumb)
1414
1415 #return(c(posNucNumb,NUCLEOTIDES[posNucKind]))
1416 return(posNucNumb)
1417 }
1418
1419 updateMutabilityNTargeting <- function(position){
1420 min_i<-max((position-2),1)
1421 max_i<-min((position+2),nchar(seqSim))
1422 min_ii<-min(min_i,3)
1423
1424 #mutability - update locally
1425 seqMutability[(min_i):(max_i)] <<- computeMutabilities(substr(seqSim,position-4,position+4))[(min_ii):(max_i-min_i+min_ii)]
1426
1427
1428 #targeting - compute locally
1429 seqTargeting[,min_i:max_i] <<- computeTargeting(substr(seqSim,min_i,max_i),seqMutability[min_i:max_i])
1430 seqTargeting[is.na(seqTargeting)] <<- 0
1431 #mutCodonPos = getCodonPos(position)
1432 mutCodonPos = seq(getCodonPos(min_i)[1],getCodonPos(max_i)[3])
1433 #cat(mutCodonPos,"\n")
1434 mutTypeCodon = getCodonPos(position)
1435 seqMutationTypes[,mutTypeCodon] <<- computeMutationTypesFast( substr(seqSim,mutTypeCodon[1],mutTypeCodon[3]) )
1436 # Stop = 0
1437 if(any(seqMutationTypes[,mutCodonPos]=="Stop",na.rm=T )){
1438 seqTargeting[,mutCodonPos][seqMutationTypes[,mutCodonPos]=="Stop"] <<- 0
1439 }
1440
1441
1442 #Selection
1443 selectedPos = (min_i*4-4)+(which(seqMutationTypes[,min_i:max_i]=="R"))
1444 # CDR
1445 selectedCDR = selectedPos[which(matCDR[selectedPos]==T)]
1446 seqTargeting[selectedCDR] <<- seqTargeting[selectedCDR] * exp(selCDR)
1447 seqTargeting[selectedCDR] <<- seqTargeting[selectedCDR]/baseLineCDR_K
1448
1449 # FWR
1450 selectedFWR = selectedPos[which(matFWR[selectedPos]==T)]
1451 seqTargeting[selectedFWR] <<- seqTargeting[selectedFWR] * exp(selFWR)
1452 seqTargeting[selectedFWR] <<- seqTargeting[selectedFWR]/baseLineFWR_K
1453
1454 }
1455
1456
1457
1458 # Validate the mutation: if the mutation has not been sampled before validate it, else discard it.
1459 validateMutation <- function(){
1460 if( !(mutatedPos%in%mutatedPositions) ){ # if it's a new mutation
1461 uniqueMutationsIntroduced <<- uniqueMutationsIntroduced + 1
1462 mutatedPositions[uniqueMutationsIntroduced] <<- mutatedPos
1463 }else{
1464 if(substr(seqSim,mutatedPos,mutatedPos)==substr(seqGermline,mutatedPos,mutatedPos)){ # back to germline mutation
1465 mutatedPositions <<- mutatedPositions[-which(mutatedPositions==mutatedPos)]
1466 uniqueMutationsIntroduced <<- uniqueMutationsIntroduced - 1
1467 }
1468 }
1469 }
1470
1471
1472
1473 # Places text (labels) at normalized coordinates
1474 myaxis <- function(xfrac=0.05,yfrac=0.05,log=FALSE,w="text",cex=1,adj=1,thecol="black"){
1475 par(xpd=TRUE)
1476 if(!log)
1477 text(par()$usr[1]-(par()$usr[2]-par()$usr[1])*xfrac,par()$usr[4]+(par()$usr[4]-par()$usr[3])*yfrac,w,cex=cex,adj=adj,col=thecol)
1478 else {
1479 if(log==2)
1480 text(
1481 par()$usr[1]-(par()$usr[2]-par()$usr[1])*xfrac,
1482 10^((par()$usr[4])+((par()$usr[4])-(par()$usr[3]))*yfrac),
1483 w,cex=cex,adj=adj,col=thecol)
1484 if(log==1)
1485 text(
1486 10^((par()$usr[1])-((par()$usr[2])-(par()$usr[1]))*xfrac),
1487 par()$usr[4]+(par()$usr[4]-par()$usr[3])*yfrac,
1488 w,cex=cex,adj=adj,col=thecol)
1489 if(log==3)
1490 text(
1491 10^((par()$usr[1])-((par()$usr[2])-(par()$usr[1]))*xfrac),
1492 10^((par()$usr[4])+((par()$usr[4])-(par()$usr[3]))*yfrac),
1493 w,cex=cex,adj=adj,col=thecol)
1494 }
1495 par(xpd=FALSE)
1496 }
1497
1498
1499
1500 # Count the mutations in a sequence
1501 analyzeMutations <- function( inputMatrixIndex, model = 0 , multipleMutation=0, seqWithStops=0){
1502
1503 paramGL = s2c(matInput[inputMatrixIndex,2])
1504 paramSeq = s2c(matInput[inputMatrixIndex,1])
1505
1506 #if( any(paramSeq=="N") ){
1507 # gapPos_Seq = which(paramSeq=="N")
1508 # gapPos_Seq_ToReplace = gapPos_Seq[paramGL[gapPos_Seq] != "N"]
1509 # paramSeq[gapPos_Seq_ToReplace] = paramGL[gapPos_Seq_ToReplace]
1510 #}
1511 mutations_val = paramGL != paramSeq
1512
1513 if(any(mutations_val)){
1514 mutationPos = which(mutations_val)#{1:length(mutations_val)}[mutations_val]
1515 length_mutations =length(mutationPos)
1516 mutationInfo = rep(NA,length_mutations)
1517
1518 pos<- mutationPos
1519 pos_array<-array(sapply(pos,getCodonPos))
1520 codonGL = paramGL[pos_array]
1521 codonSeqWhole = paramSeq[pos_array]
1522 codonSeq = sapply(pos,function(x){
1523 seqP = paramGL[getCodonPos(x)]
1524 muCodonPos = {x-1}%%3+1
1525 seqP[muCodonPos] = paramSeq[x]
1526 return(seqP)
1527 })
1528 GLcodons = apply(matrix(codonGL,length_mutations,3,byrow=TRUE),1,c2s)
1529 SeqcodonsWhole = apply(matrix(codonSeqWhole,length_mutations,3,byrow=TRUE),1,c2s)
1530 Seqcodons = apply(codonSeq,2,c2s)
1531
1532 mutationInfo = apply(rbind(GLcodons , Seqcodons),2,function(x){mutationType(c2s(x[1]),c2s(x[2]))})
1533 names(mutationInfo) = mutationPos
1534
1535 mutationInfoWhole = apply(rbind(GLcodons , SeqcodonsWhole),2,function(x){mutationType(c2s(x[1]),c2s(x[2]))})
1536 names(mutationInfoWhole) = mutationPos
1537
1538 mutationInfo <- mutationInfo[!is.na(mutationInfo)]
1539 mutationInfoWhole <- mutationInfoWhole[!is.na(mutationInfoWhole)]
1540
1541 if(any(!is.na(mutationInfo))){
1542
1543 #Filter based on Stop (at the codon level)
1544 if(seqWithStops==1){
1545 nucleotidesAtStopCodons = names(mutationInfoWhole[mutationInfoWhole!="Stop"])
1546 mutationInfo = mutationInfo[nucleotidesAtStopCodons]
1547 mutationInfoWhole = mutationInfo[nucleotidesAtStopCodons]
1548 }else{
1549 countStops = sum(mutationInfoWhole=="Stop")
1550 if(seqWithStops==2 & countStops==0) mutationInfo = NA
1551 if(seqWithStops==3 & countStops>0) mutationInfo = NA
1552 }
1553
1554 if(any(!is.na(mutationInfo))){
1555 #Filter mutations based on multipleMutation
1556 if(multipleMutation==1 & !is.na(mutationInfo)){
1557 mutationCodons = getCodonNumb(as.numeric(names(mutationInfoWhole)))
1558 tableMutationCodons <- table(mutationCodons)
1559 codonsWithMultipleMutations <- as.numeric(names(tableMutationCodons[tableMutationCodons>1]))
1560 if(any(codonsWithMultipleMutations)){
1561 #remove the nucleotide mutations in the codons with multiple mutations
1562 mutationInfo <- mutationInfo[!(mutationCodons %in% codonsWithMultipleMutations)]
1563 #replace those codons with Ns in the input sequence
1564 paramSeq[unlist(lapply(codonsWithMultipleMutations, getCodonNucs))] = "N"
1565 matInput[inputMatrixIndex,1] <<- c2s(paramSeq)
1566 }
1567 }
1568
1569 #Filter mutations based on the model
1570 if(any(mutationInfo)==T | is.na(any(mutationInfo))){
1571
1572 if(model==1 & !is.na(mutationInfo)){
1573 mutationInfo <- mutationInfo[mutationInfo=="S"]
1574 }
1575 if(any(mutationInfo)==T | is.na(any(mutationInfo))) return(mutationInfo)
1576 else return(NA)
1577 }else{
1578 return(NA)
1579 }
1580 }else{
1581 return(NA)
1582 }
1583
1584
1585 }else{
1586 return(NA)
1587 }
1588
1589
1590 }else{
1591 return (NA)
1592 }
1593 }
1594
1595 analyzeMutationsFixed <- function( inputArray, model = 0 , multipleMutation=0, seqWithStops=0){
1596
1597 paramGL = s2c(inputArray[2])
1598 paramSeq = s2c(inputArray[1])
1599 inputSeq <- inputArray[1]
1600 #if( any(paramSeq=="N") ){
1601 # gapPos_Seq = which(paramSeq=="N")
1602 # gapPos_Seq_ToReplace = gapPos_Seq[paramGL[gapPos_Seq] != "N"]
1603 # paramSeq[gapPos_Seq_ToReplace] = paramGL[gapPos_Seq_ToReplace]
1604 #}
1605 mutations_val = paramGL != paramSeq
1606
1607 if(any(mutations_val)){
1608 mutationPos = which(mutations_val)#{1:length(mutations_val)}[mutations_val]
1609 length_mutations =length(mutationPos)
1610 mutationInfo = rep(NA,length_mutations)
1611
1612 pos<- mutationPos
1613 pos_array<-array(sapply(pos,getCodonPos))
1614 codonGL = paramGL[pos_array]
1615 codonSeqWhole = paramSeq[pos_array]
1616 codonSeq = sapply(pos,function(x){
1617 seqP = paramGL[getCodonPos(x)]
1618 muCodonPos = {x-1}%%3+1
1619 seqP[muCodonPos] = paramSeq[x]
1620 return(seqP)
1621 })
1622 GLcodons = apply(matrix(codonGL,length_mutations,3,byrow=TRUE),1,c2s)
1623 SeqcodonsWhole = apply(matrix(codonSeqWhole,length_mutations,3,byrow=TRUE),1,c2s)
1624 Seqcodons = apply(codonSeq,2,c2s)
1625
1626 mutationInfo = apply(rbind(GLcodons , Seqcodons),2,function(x){mutationType(c2s(x[1]),c2s(x[2]))})
1627 names(mutationInfo) = mutationPos
1628
1629 mutationInfoWhole = apply(rbind(GLcodons , SeqcodonsWhole),2,function(x){mutationType(c2s(x[1]),c2s(x[2]))})
1630 names(mutationInfoWhole) = mutationPos
1631
1632 mutationInfo <- mutationInfo[!is.na(mutationInfo)]
1633 mutationInfoWhole <- mutationInfoWhole[!is.na(mutationInfoWhole)]
1634
1635 if(any(!is.na(mutationInfo))){
1636
1637 #Filter based on Stop (at the codon level)
1638 if(seqWithStops==1){
1639 nucleotidesAtStopCodons = names(mutationInfoWhole[mutationInfoWhole!="Stop"])
1640 mutationInfo = mutationInfo[nucleotidesAtStopCodons]
1641 mutationInfoWhole = mutationInfo[nucleotidesAtStopCodons]
1642 }else{
1643 countStops = sum(mutationInfoWhole=="Stop")
1644 if(seqWithStops==2 & countStops==0) mutationInfo = NA
1645 if(seqWithStops==3 & countStops>0) mutationInfo = NA
1646 }
1647
1648 if(any(!is.na(mutationInfo))){
1649 #Filter mutations based on multipleMutation
1650 if(multipleMutation==1 & !is.na(mutationInfo)){
1651 mutationCodons = getCodonNumb(as.numeric(names(mutationInfoWhole)))
1652 tableMutationCodons <- table(mutationCodons)
1653 codonsWithMultipleMutations <- as.numeric(names(tableMutationCodons[tableMutationCodons>1]))
1654 if(any(codonsWithMultipleMutations)){
1655 #remove the nucleotide mutations in the codons with multiple mutations
1656 mutationInfo <- mutationInfo[!(mutationCodons %in% codonsWithMultipleMutations)]
1657 #replace those codons with Ns in the input sequence
1658 paramSeq[unlist(lapply(codonsWithMultipleMutations, getCodonNucs))] = "N"
1659 #matInput[inputMatrixIndex,1] <<- c2s(paramSeq)
1660 inputSeq <- c2s(paramSeq)
1661 }
1662 }
1663
1664 #Filter mutations based on the model
1665 if(any(mutationInfo)==T | is.na(any(mutationInfo))){
1666
1667 if(model==1 & !is.na(mutationInfo)){
1668 mutationInfo <- mutationInfo[mutationInfo=="S"]
1669 }
1670 if(any(mutationInfo)==T | is.na(any(mutationInfo))) return(list(mutationInfo,inputSeq))
1671 else return(list(NA,inputSeq))
1672 }else{
1673 return(list(NA,inputSeq))
1674 }
1675 }else{
1676 return(list(NA,inputSeq))
1677 }
1678
1679
1680 }else{
1681 return(list(NA,inputSeq))
1682 }
1683
1684
1685 }else{
1686 return (list(NA,inputSeq))
1687 }
1688 }
1689
1690 # triMutability Background Count
1691 buildMutabilityModel <- function( inputMatrixIndex, model=0 , multipleMutation=0, seqWithStops=0, stopMutations=0){
1692
1693 #rowOrigMatInput = matInput[inputMatrixIndex,]
1694 seqGL = gsub("-", "", matInput[inputMatrixIndex,2])
1695 seqInput = gsub("-", "", matInput[inputMatrixIndex,1])
1696 #matInput[inputMatrixIndex,] <<- cbind(seqInput,seqGL)
1697 tempInput <- cbind(seqInput,seqGL)
1698 seqLength = nchar(seqGL)
1699 list_analyzeMutationsFixed<- analyzeMutationsFixed(tempInput, model, multipleMutation, seqWithStops)
1700 mutationCount <- list_analyzeMutationsFixed[[1]]
1701 seqInput <- list_analyzeMutationsFixed[[2]]
1702 BackgroundMatrix = mutabilityMatrix
1703 MutationMatrix = mutabilityMatrix
1704 MutationCountMatrix = mutabilityMatrix
1705 if(!is.na(mutationCount)){
1706 if((stopMutations==0 & model==0) | (stopMutations==1 & (sum(mutationCount=="Stop")<length(mutationCount))) | (model==1 & (sum(mutationCount=="S")>0)) ){
1707
1708 fivermerStartPos = 1:(seqLength-4)
1709 fivemerLength <- length(fivermerStartPos)
1710 fivemerGL <- substr(rep(seqGL,length(fivermerStartPos)),(fivermerStartPos),(fivermerStartPos+4))
1711 fivemerSeq <- substr(rep(seqInput,length(fivermerStartPos)),(fivermerStartPos),(fivermerStartPos+4))
1712
1713 #Background
1714 for(fivemerIndex in 1:fivemerLength){
1715 fivemer = fivemerGL[fivemerIndex]
1716 if(!any(grep("N",fivemer))){
1717 fivemerCodonPos = fivemerCodon(fivemerIndex)
1718 fivemerReadingFrameCodon = substr(fivemer,fivemerCodonPos[1],fivemerCodonPos[3])
1719 fivemerReadingFrameCodonInputSeq = substr(fivemerSeq[fivemerIndex],fivemerCodonPos[1],fivemerCodonPos[3])
1720
1721 # All mutations model
1722 #if(!any(grep("N",fivemerReadingFrameCodon))){
1723 if(model==0){
1724 if(stopMutations==0){
1725 if(!any(grep("N",fivemerReadingFrameCodonInputSeq)))
1726 BackgroundMatrix[fivemer] <- (BackgroundMatrix[fivemer] + 1)
1727 }else{
1728 if( !any(grep("N",fivemerReadingFrameCodonInputSeq)) & translateCodonToAminoAcid(fivemerReadingFrameCodon)!="*" ){
1729 positionWithinCodon = which(fivemerCodonPos==3)#positionsWithinCodon[(fivemerCodonPos[1]%%3)+1]
1730 BackgroundMatrix[fivemer] <- (BackgroundMatrix[fivemer] + probNonStopMutations[fivemerReadingFrameCodon,positionWithinCodon])
1731 }
1732 }
1733 }else{ # Only silent mutations
1734 if( !any(grep("N",fivemerReadingFrameCodonInputSeq)) & translateCodonToAminoAcid(fivemerReadingFrameCodon)!="*" & translateCodonToAminoAcid(fivemerReadingFrameCodonInputSeq)==translateCodonToAminoAcid(fivemerReadingFrameCodon)){
1735 positionWithinCodon = which(fivemerCodonPos==3)
1736 BackgroundMatrix[fivemer] <- (BackgroundMatrix[fivemer] + probSMutations[fivemerReadingFrameCodon,positionWithinCodon])
1737 }
1738 }
1739 #}
1740 }
1741 }
1742
1743 #Mutations
1744 if(stopMutations==1) mutationCount = mutationCount[mutationCount!="Stop"]
1745 if(model==1) mutationCount = mutationCount[mutationCount=="S"]
1746 mutationPositions = as.numeric(names(mutationCount))
1747 mutationCount = mutationCount[mutationPositions>2 & mutationPositions<(seqLength-1)]
1748 mutationPositions = mutationPositions[mutationPositions>2 & mutationPositions<(seqLength-1)]
1749 countMutations = 0
1750 for(mutationPosition in mutationPositions){
1751 fivemerIndex = mutationPosition-2
1752 fivemer = fivemerSeq[fivemerIndex]
1753 GLfivemer = fivemerGL[fivemerIndex]
1754 fivemerCodonPos = fivemerCodon(fivemerIndex)
1755 fivemerReadingFrameCodon = substr(fivemer,fivemerCodonPos[1],fivemerCodonPos[3])
1756 fivemerReadingFrameCodonGL = substr(GLfivemer,fivemerCodonPos[1],fivemerCodonPos[3])
1757 if(!any(grep("N",fivemer)) & !any(grep("N",GLfivemer))){
1758 if(model==0){
1759 countMutations = countMutations + 1
1760 MutationMatrix[GLfivemer] <- (MutationMatrix[GLfivemer] + 1)
1761 MutationCountMatrix[GLfivemer] <- (MutationCountMatrix[GLfivemer] + 1)
1762 }else{
1763 if( translateCodonToAminoAcid(fivemerReadingFrameCodonGL)!="*" ){
1764 countMutations = countMutations + 1
1765 positionWithinCodon = which(fivemerCodonPos==3)
1766 glNuc = substr(fivemerReadingFrameCodonGL,positionWithinCodon,positionWithinCodon)
1767 inputNuc = substr(fivemerReadingFrameCodon,positionWithinCodon,positionWithinCodon)
1768 MutationMatrix[GLfivemer] <- (MutationMatrix[GLfivemer] + substitution[glNuc,inputNuc])
1769 MutationCountMatrix[GLfivemer] <- (MutationCountMatrix[GLfivemer] + 1)
1770 }
1771 }
1772 }
1773 }
1774
1775 seqMutability = MutationMatrix/BackgroundMatrix
1776 seqMutability = seqMutability/sum(seqMutability,na.rm=TRUE)
1777 #cat(inputMatrixIndex,"\t",countMutations,"\n")
1778 return(list("seqMutability" = seqMutability,"numbMutations" = countMutations,"seqMutabilityCount" = MutationCountMatrix, "BackgroundMatrix"=BackgroundMatrix))
1779
1780 }
1781 }
1782
1783 }
1784
1785 #Returns the codon position containing the middle nucleotide
1786 fivemerCodon <- function(fivemerIndex){
1787 codonPos = list(2:4,1:3,3:5)
1788 fivemerType = fivemerIndex%%3
1789 return(codonPos[[fivemerType+1]])
1790 }
1791
1792 #returns probability values for one mutation in codons resulting in R, S or Stop
1793 probMutations <- function(typeOfMutation){
1794 matMutationProb <- matrix(0,ncol=3,nrow=125,dimnames=list(words(alphabet = c(NUCLEOTIDES,"N"), length=3),c(1:3)))
1795 for(codon in rownames(matMutationProb)){
1796 if( !any(grep("N",codon)) ){
1797 for(muPos in 1:3){
1798 matCodon = matrix(rep(s2c(codon),3),nrow=3,ncol=3,byrow=T)
1799 glNuc = matCodon[1,muPos]
1800 matCodon[,muPos] = canMutateTo(glNuc)
1801 substitutionRate = substitution[glNuc,matCodon[,muPos]]
1802 typeOfMutations = apply(rbind(rep(codon,3),apply(matCodon,1,c2s)),2,function(x){mutationType(c2s(x[1]),c2s(x[2]))})
1803 matMutationProb[codon,muPos] <- sum(substitutionRate[typeOfMutations==typeOfMutation])
1804 }
1805 }
1806 }
1807
1808 return(matMutationProb)
1809 }
1810
1811
1812
1813
1814 #Mapping Trinucleotides to fivemers
1815 mapTriToFivemer <- function(triMutability=triMutability_Literature_Human){
1816 rownames(triMutability) <- triMutability_Names
1817 Fivemer<-rep(NA,1024)
1818 names(Fivemer)<-words(alphabet=NUCLEOTIDES,length=5)
1819 Fivemer<-sapply(names(Fivemer),function(Word)return(sum( c(triMutability[substring(Word,3,5),1],triMutability[substring(Word,2,4),2],triMutability[substring(Word,1,3),3]),na.rm=TRUE)))
1820 Fivemer<-Fivemer/sum(Fivemer)
1821 return(Fivemer)
1822 }
1823
1824 collapseFivemerToTri<-function(Fivemer,Weights=MutabilityWeights,position=1,NUC="A"){
1825 Indices<-substring(names(Fivemer),3,3)==NUC
1826 Factors<-substring(names(Fivemer[Indices]),(4-position),(6-position))
1827 tapply(which(Indices),Factors,function(i)weighted.mean(Fivemer[i],Weights[i],na.rm=TRUE))
1828 }
1829
1830
1831
1832 CountFivemerToTri<-function(Fivemer,Weights=MutabilityWeights,position=1,NUC="A"){
1833 Indices<-substring(names(Fivemer),3,3)==NUC
1834 Factors<-substring(names(Fivemer[Indices]),(4-position),(6-position))
1835 tapply(which(Indices),Factors,function(i)sum(Weights[i],na.rm=TRUE))
1836 }
1837
1838 #Uses the real counts of the mutated fivemers
1839 CountFivemerToTri2<-function(Fivemer,Counts=MutabilityCounts,position=1,NUC="A"){
1840 Indices<-substring(names(Fivemer),3,3)==NUC
1841 Factors<-substring(names(Fivemer[Indices]),(4-position),(6-position))
1842 tapply(which(Indices),Factors,function(i)sum(Counts[i],na.rm=TRUE))
1843 }
1844
1845 bootstrap<-function(x=c(33,12,21),M=10000,alpha=0.05){
1846 N<-sum(x)
1847 if(N){
1848 p<-x/N
1849 k<-length(x)-1
1850 tmp<-rmultinom(M, size = N, prob=p)
1851 tmp_p<-apply(tmp,2,function(y)y/N)
1852 (apply(tmp_p,1,function(y)quantile(y,c(alpha/2/k,1-alpha/2/k))))
1853 }
1854 else return(matrix(0,2,length(x)))
1855 }
1856
1857
1858
1859
1860 bootstrap2<-function(x=c(33,12,21),n=10,M=10000,alpha=0.05){
1861
1862 N<-sum(x)
1863 k<-length(x)
1864 y<-rep(1:k,x)
1865 tmp<-sapply(1:M,function(i)sample(y,n))
1866 if(n>1)tmp_p<-sapply(1:M,function(j)sapply(1:k,function(i)sum(tmp[,j]==i)))/n
1867 if(n==1)tmp_p<-sapply(1:M,function(j)sapply(1:k,function(i)sum(tmp[j]==i)))/n
1868 (apply(tmp_p,1,function(z)quantile(z,c(alpha/2/(k-1),1-alpha/2/(k-1)))))
1869 }
1870
1871
1872
1873 p_value<-function(x=c(33,12,21),M=100000,x_obs=c(2,5,3)){
1874 n=sum(x_obs)
1875 N<-sum(x)
1876 k<-length(x)
1877 y<-rep(1:k,x)
1878 tmp<-sapply(1:M,function(i)sample(y,n))
1879 if(n>1)tmp_p<-sapply(1:M,function(j)sapply(1:k,function(i)sum(tmp[,j]==i)))
1880 if(n==1)tmp_p<-sapply(1:M,function(j)sapply(1:k,function(i)sum(tmp[j]==i)))
1881 tmp<-rbind(sapply(1:3,function(i)sum(tmp_p[i,]>=x_obs[i])/M),
1882 sapply(1:3,function(i)sum(tmp_p[i,]<=x_obs[i])/M))
1883 sapply(1:3,function(i){if(tmp[1,i]>=tmp[2,i])return(-tmp[2,i])else return(tmp[1,i])})
1884 }
1885
1886 #"D:\\Sequences\\IMGT Germlines\\Human_SNPless_IGHJ.FASTA"
1887 # Remove SNPs from IMGT germline segment alleles
1888 generateUnambiguousRepertoire <- function(repertoireInFile,repertoireOutFile){
1889 repertoireIn <- read.fasta(repertoireInFile, seqtype="DNA",as.string=T,set.attributes=F,forceDNAtolower=F)
1890 alleleNames <- sapply(names(repertoireIn),function(x)strsplit(x,"|",fixed=TRUE)[[1]][2])
1891 SNPs <- tapply(repertoireIn,sapply(alleleNames,function(x)strsplit(x,"*",fixed=TRUE)[[1]][1]),function(x){
1892 Indices<-NULL
1893 for(i in 1:length(x)){
1894 firstSeq = s2c(x[[1]])
1895 iSeq = s2c(x[[i]])
1896 Indices<-c(Indices,which(firstSeq[1:320]!=iSeq[1:320] & firstSeq[1:320]!="." & iSeq[1:320]!="." ))
1897 }
1898 return(sort(unique(Indices)))
1899 })
1900 repertoireOut <- repertoireIn
1901 repertoireOut <- lapply(names(repertoireOut), function(repertoireName){
1902 alleleName <- strsplit(repertoireName,"|",fixed=TRUE)[[1]][2]
1903 geneSegmentName <- strsplit(alleleName,"*",fixed=TRUE)[[1]][1]
1904 alleleSeq <- s2c(repertoireOut[[repertoireName]])
1905 alleleSeq[as.numeric(unlist(SNPs[geneSegmentName]))] <- "N"
1906 alleleSeq <- c2s(alleleSeq)
1907 repertoireOut[[repertoireName]] <- alleleSeq
1908 })
1909 names(repertoireOut) <- names(repertoireIn)
1910 write.fasta(repertoireOut,names(repertoireOut),file.out=repertoireOutFile)
1911
1912 }
1913
1914
1915
1916
1917
1918
1919 ############
1920 groupBayes2 = function(indexes, param_resultMat){
1921
1922 BayesGDist_Focused_CDR = calculate_bayesG( x=param_resultMat[indexes,1], N=apply(param_resultMat[indexes,c(1,2,4)],1,sum,na.rm=T), p=apply(param_resultMat[indexes,5:8],1,function(x){x[1]/(x[1]+x[2]+x[4])}))
1923 BayesGDist_Focused_FWR = calculate_bayesG( x=param_resultMat[indexes,3], N=apply(param_resultMat[indexes,c(3,2,4)],1,sum,na.rm=T), p=apply(param_resultMat[indexes,5:8],1,function(x){x[3]/(x[3]+x[2]+x[4])}))
1924 #BayesGDist_Local_CDR = calculate_bayesG( x=param_resultMat[indexes,1], N=apply(param_resultMat[indexes,c(1,2)],1,sum,na.rm=T), p=apply(param_resultMat[indexes,5:8],1,function(x){x[1]/(x[1]+x[2])}))
1925 #BayesGDist_Local_FWR = calculate_bayesG( x=param_resultMat[indexes,3], N=apply(param_resultMat[indexes,c(3,4)],1,sum,na.rm=T), p=apply(param_resultMat[indexes,5:8],1,function(x){x[3]/(x[3]+x[4])}))
1926 #BayesGDist_Global_CDR = calculate_bayesG( x=param_resultMat[indexes,1], N=apply(param_resultMat[indexes,c(1,2,3,4)],1,sum,na.rm=T), p=apply(param_resultMat[indexes,5:8],1,function(x){x[1]/(x[1]+x[2]+x[3]+x[4])}))
1927 #BayesGDist_Global_FWR = calculate_bayesG( x=param_resultMat[indexes,3], N=apply(param_resultMat[indexes,c(1,2,3,4)],1,sum,na.rm=T), p=apply(param_resultMat[indexes,5:8],1,function(x){x[3]/(x[1]+x[2]+x[3]+x[4])}))
1928 return ( list("BayesGDist_Focused_CDR"=BayesGDist_Focused_CDR,
1929 "BayesGDist_Focused_FWR"=BayesGDist_Focused_FWR) )
1930 #"BayesGDist_Local_CDR"=BayesGDist_Local_CDR,
1931 #"BayesGDist_Local_FWR" = BayesGDist_Local_FWR))
1932 # "BayesGDist_Global_CDR" = BayesGDist_Global_CDR,
1933 # "BayesGDist_Global_FWR" = BayesGDist_Global_FWR) )
1934
1935
1936 }
1937
1938
1939 calculate_bayesG <- function( x=array(), N=array(), p=array(), max_sigma=20, length_sigma=4001){
1940 G <- max(length(x),length(N),length(p))
1941 x=array(x,dim=G)
1942 N=array(N,dim=G)
1943 p=array(p,dim=G)
1944
1945 indexOfZero = N>0 & p>0
1946 N = N[indexOfZero]
1947 x = x[indexOfZero]
1948 p = p[indexOfZero]
1949 G <- length(x)
1950
1951 if(G){
1952
1953 cons<-array( dim=c(length_sigma,G) )
1954 if(G==1) {
1955 return(calculate_bayes(x=x[G],N=N[G],p=p[G],max_sigma=max_sigma,length_sigma=length_sigma))
1956 }
1957 else {
1958 for(g in 1:G) cons[,g] <- calculate_bayes(x=x[g],N=N[g],p=p[g],max_sigma=max_sigma,length_sigma=length_sigma)
1959 listMatG <- convolutionPowersOfTwoByTwos(cons,length_sigma=length_sigma)
1960 y<-calculate_bayesGHelper(listMatG,length_sigma=length_sigma)
1961 return( y/sum(y)/(2*max_sigma/(length_sigma-1)) )
1962 }
1963 }else{
1964 return(NA)
1965 }
1966 }
1967
1968
1969 calculate_bayesGHelper <- function( listMatG,length_sigma=4001 ){
1970 matG <- listMatG[[1]]
1971 groups <- listMatG[[2]]
1972 i = 1
1973 resConv <- matG[,i]
1974 denom <- 2^groups[i]
1975 if(length(groups)>1){
1976 while( i<length(groups) ){
1977 i = i + 1
1978 resConv <- weighted_conv(resConv, matG[,i], w= {{2^groups[i]}/denom} ,length_sigma=length_sigma)
1979 #cat({{2^groups[i]}/denom},"\n")
1980 denom <- denom + 2^groups[i]
1981 }
1982 }
1983 return(resConv)
1984 }
1985
1986 weighted_conv<-function(x,y,w=1,m=100,length_sigma=4001){
1987 lx<-length(x)
1988 ly<-length(y)
1989 if({lx<m}| {{lx*w}<m}| {{ly}<m}| {{ly*w}<m}){
1990 if(w<1){
1991 y1<-approx(1:ly,y,seq(1,ly,length.out=m))$y
1992 x1<-approx(1:lx,x,seq(1,lx,length.out=m/w))$y
1993 lx<-length(x1)
1994 ly<-length(y1)
1995 }
1996 else {
1997 y1<-approx(1:ly,y,seq(1,ly,length.out=m*w))$y
1998 x1<-approx(1:lx,x,seq(1,lx,length.out=m))$y
1999 lx<-length(x1)
2000 ly<-length(y1)
2001 }
2002 }
2003 else{
2004 x1<-x
2005 y1<-approx(1:ly,y,seq(1,ly,length.out=floor(lx*w)))$y
2006 ly<-length(y1)
2007 }
2008 tmp<-approx(x=1:(lx+ly-1),y=convolve(x1,rev(y1),type="open"),xout=seq(1,lx+ly-1,length.out=length_sigma))$y
2009 tmp[tmp<=0] = 0
2010 return(tmp/sum(tmp))
2011 }
2012
2013 ########################
2014
2015
2016
2017
2018 mutabilityMatrixONE<-rep(0,4)
2019 names(mutabilityMatrixONE)<-NUCLEOTIDES
2020
2021 # triMutability Background Count
2022 buildMutabilityModelONE <- function( inputMatrixIndex, model=0 , multipleMutation=0, seqWithStops=0, stopMutations=0){
2023
2024 #rowOrigMatInput = matInput[inputMatrixIndex,]
2025 seqGL = gsub("-", "", matInput[inputMatrixIndex,2])
2026 seqInput = gsub("-", "", matInput[inputMatrixIndex,1])
2027 matInput[inputMatrixIndex,] <<- c(seqInput,seqGL)
2028 seqLength = nchar(seqGL)
2029 mutationCount <- analyzeMutations(inputMatrixIndex, model, multipleMutation, seqWithStops)
2030 BackgroundMatrix = mutabilityMatrixONE
2031 MutationMatrix = mutabilityMatrixONE
2032 MutationCountMatrix = mutabilityMatrixONE
2033 if(!is.na(mutationCount)){
2034 if((stopMutations==0 & model==0) | (stopMutations==1 & (sum(mutationCount=="Stop")<length(mutationCount))) | (model==1 & (sum(mutationCount=="S")>0)) ){
2035
2036 # ONEmerStartPos = 1:(seqLength)
2037 # ONEmerLength <- length(ONEmerStartPos)
2038 ONEmerGL <- s2c(seqGL)
2039 ONEmerSeq <- s2c(seqInput)
2040
2041 #Background
2042 for(ONEmerIndex in 1:seqLength){
2043 ONEmer = ONEmerGL[ONEmerIndex]
2044 if(ONEmer!="N"){
2045 ONEmerCodonPos = getCodonPos(ONEmerIndex)
2046 ONEmerReadingFrameCodon = c2s(ONEmerGL[ONEmerCodonPos])
2047 ONEmerReadingFrameCodonInputSeq = c2s(ONEmerSeq[ONEmerCodonPos] )
2048
2049 # All mutations model
2050 #if(!any(grep("N",ONEmerReadingFrameCodon))){
2051 if(model==0){
2052 if(stopMutations==0){
2053 if(!any(grep("N",ONEmerReadingFrameCodonInputSeq)))
2054 BackgroundMatrix[ONEmer] <- (BackgroundMatrix[ONEmer] + 1)
2055 }else{
2056 if( !any(grep("N",ONEmerReadingFrameCodonInputSeq)) & translateCodonToAminoAcid(ONEmerReadingFrameCodonInputSeq)!="*"){
2057 positionWithinCodon = which(ONEmerCodonPos==ONEmerIndex)#positionsWithinCodon[(ONEmerCodonPos[1]%%3)+1]
2058 BackgroundMatrix[ONEmer] <- (BackgroundMatrix[ONEmer] + probNonStopMutations[ONEmerReadingFrameCodon,positionWithinCodon])
2059 }
2060 }
2061 }else{ # Only silent mutations
2062 if( !any(grep("N",ONEmerReadingFrameCodonInputSeq)) & translateCodonToAminoAcid(ONEmerReadingFrameCodonInputSeq)!="*" & translateCodonToAminoAcid(ONEmerReadingFrameCodonInputSeq)==translateCodonToAminoAcid(ONEmerReadingFrameCodon) ){
2063 positionWithinCodon = which(ONEmerCodonPos==ONEmerIndex)
2064 BackgroundMatrix[ONEmer] <- (BackgroundMatrix[ONEmer] + probSMutations[ONEmerReadingFrameCodon,positionWithinCodon])
2065 }
2066 }
2067 }
2068 }
2069 }
2070
2071 #Mutations
2072 if(stopMutations==1) mutationCount = mutationCount[mutationCount!="Stop"]
2073 if(model==1) mutationCount = mutationCount[mutationCount=="S"]
2074 mutationPositions = as.numeric(names(mutationCount))
2075 mutationCount = mutationCount[mutationPositions>2 & mutationPositions<(seqLength-1)]
2076 mutationPositions = mutationPositions[mutationPositions>2 & mutationPositions<(seqLength-1)]
2077 countMutations = 0
2078 for(mutationPosition in mutationPositions){
2079 ONEmerIndex = mutationPosition
2080 ONEmer = ONEmerSeq[ONEmerIndex]
2081 GLONEmer = ONEmerGL[ONEmerIndex]
2082 ONEmerCodonPos = getCodonPos(ONEmerIndex)
2083 ONEmerReadingFrameCodon = c2s(ONEmerSeq[ONEmerCodonPos])
2084 ONEmerReadingFrameCodonGL =c2s(ONEmerGL[ONEmerCodonPos])
2085 if(!any(grep("N",ONEmer)) & !any(grep("N",GLONEmer))){
2086 if(model==0){
2087 countMutations = countMutations + 1
2088 MutationMatrix[GLONEmer] <- (MutationMatrix[GLONEmer] + 1)
2089 MutationCountMatrix[GLONEmer] <- (MutationCountMatrix[GLONEmer] + 1)
2090 }else{
2091 if( translateCodonToAminoAcid(ONEmerReadingFrameCodonGL)!="*" ){
2092 countMutations = countMutations + 1
2093 positionWithinCodon = which(ONEmerCodonPos==ONEmerIndex)
2094 glNuc = substr(ONEmerReadingFrameCodonGL,positionWithinCodon,positionWithinCodon)
2095 inputNuc = substr(ONEmerReadingFrameCodon,positionWithinCodon,positionWithinCodon)
2096 MutationMatrix[GLONEmer] <- (MutationMatrix[GLONEmer] + substitution[glNuc,inputNuc])
2097 MutationCountMatrix[GLONEmer] <- (MutationCountMatrix[GLONEmer] + 1)
2098 }
2099 }
2100 }
2101 }
2102
2103 seqMutability = MutationMatrix/BackgroundMatrix
2104 seqMutability = seqMutability/sum(seqMutability,na.rm=TRUE)
2105 #cat(inputMatrixIndex,"\t",countMutations,"\n")
2106 return(list("seqMutability" = seqMutability,"numbMutations" = countMutations,"seqMutabilityCount" = MutationCountMatrix, "BackgroundMatrix"=BackgroundMatrix))
2107 # tmp<-list("seqMutability" = seqMutability,"numbMutations" = countMutations,"seqMutabilityCount" = MutationCountMatrix)
2108 }
2109 }
2110
2111 ################
2112 # $Id: trim.R 989 2006-10-29 15:28:26Z ggorjan $
2113
2114 trim <- function(s, recode.factor=TRUE, ...)
2115 UseMethod("trim", s)
2116
2117 trim.default <- function(s, recode.factor=TRUE, ...)
2118 s
2119
2120 trim.character <- function(s, recode.factor=TRUE, ...)
2121 {
2122 s <- sub(pattern="^ +", replacement="", x=s)
2123 s <- sub(pattern=" +$", replacement="", x=s)
2124 s
2125 }
2126
2127 trim.factor <- function(s, recode.factor=TRUE, ...)
2128 {
2129 levels(s) <- trim(levels(s))
2130 if(recode.factor) {
2131 dots <- list(x=s, ...)
2132 if(is.null(dots$sort)) dots$sort <- sort
2133 s <- do.call(what=reorder.factor, args=dots)
2134 }
2135 s
2136 }
2137
2138 trim.list <- function(s, recode.factor=TRUE, ...)
2139 lapply(s, trim, recode.factor=recode.factor, ...)
2140
2141 trim.data.frame <- function(s, recode.factor=TRUE, ...)
2142 {
2143 s[] <- trim.list(s, recode.factor=recode.factor, ...)
2144 s
2145 }
2146 #######################################
2147 # Compute the expected for each sequence-germline pair by codon
2148 getExpectedIndividualByCodon <- function(matInput){
2149 if( any(grep("multicore",search())) ){
2150 facGL <- factor(matInput[,2])
2151 facLevels = levels(facGL)
2152 LisGLs_MutabilityU = mclapply(1:length(facLevels), function(x){
2153 computeMutabilities(facLevels[x])
2154 })
2155 facIndex = match(facGL,facLevels)
2156
2157 LisGLs_Mutability = mclapply(1:nrow(matInput), function(x){
2158 cInput = rep(NA,nchar(matInput[x,1]))
2159 cInput[s2c(matInput[x,1])!="N"] = 1
2160 LisGLs_MutabilityU[[facIndex[x]]] * cInput
2161 })
2162
2163 LisGLs_Targeting = mclapply(1:dim(matInput)[1], function(x){
2164 computeTargeting(matInput[x,2],LisGLs_Mutability[[x]])
2165 })
2166
2167 LisGLs_MutationTypes = mclapply(1:length(matInput[,2]),function(x){
2168 #print(x)
2169 computeMutationTypes(matInput[x,2])
2170 })
2171
2172 LisGLs_R_Exp = mclapply(1:nrow(matInput), function(x){
2173 Exp_R <- rollapply(as.zoo(1:readEnd),width=3,by=3,
2174 function(codonNucs){
2175 RPos = which(LisGLs_MutationTypes[[x]][,codonNucs]=="R")
2176 sum( LisGLs_Targeting[[x]][,codonNucs][RPos], na.rm=T )
2177 }
2178 )
2179 })
2180
2181 LisGLs_S_Exp = mclapply(1:nrow(matInput), function(x){
2182 Exp_S <- rollapply(as.zoo(1:readEnd),width=3,by=3,
2183 function(codonNucs){
2184 SPos = which(LisGLs_MutationTypes[[x]][,codonNucs]=="S")
2185 sum( LisGLs_Targeting[[x]][,codonNucs][SPos], na.rm=T )
2186 }
2187 )
2188 })
2189
2190 Exp_R = matrix(unlist(LisGLs_R_Exp),nrow=nrow(matInput),ncol=readEnd/3,T)
2191 Exp_S = matrix(unlist(LisGLs_S_Exp),nrow=nrow(matInput),ncol=readEnd/3,T)
2192 return( list( "Expected_R"=Exp_R, "Expected_S"=Exp_S) )
2193 }else{
2194 facGL <- factor(matInput[,2])
2195 facLevels = levels(facGL)
2196 LisGLs_MutabilityU = lapply(1:length(facLevels), function(x){
2197 computeMutabilities(facLevels[x])
2198 })
2199 facIndex = match(facGL,facLevels)
2200
2201 LisGLs_Mutability = lapply(1:nrow(matInput), function(x){
2202 cInput = rep(NA,nchar(matInput[x,1]))
2203 cInput[s2c(matInput[x,1])!="N"] = 1
2204 LisGLs_MutabilityU[[facIndex[x]]] * cInput
2205 })
2206
2207 LisGLs_Targeting = lapply(1:dim(matInput)[1], function(x){
2208 computeTargeting(matInput[x,2],LisGLs_Mutability[[x]])
2209 })
2210
2211 LisGLs_MutationTypes = lapply(1:length(matInput[,2]),function(x){
2212 #print(x)
2213 computeMutationTypes(matInput[x,2])
2214 })
2215
2216 LisGLs_R_Exp = lapply(1:nrow(matInput), function(x){
2217 Exp_R <- rollapply(as.zoo(1:readEnd),width=3,by=3,
2218 function(codonNucs){
2219 RPos = which(LisGLs_MutationTypes[[x]][,codonNucs]=="R")
2220 sum( LisGLs_Targeting[[x]][,codonNucs][RPos], na.rm=T )
2221 }
2222 )
2223 })
2224
2225 LisGLs_S_Exp = lapply(1:nrow(matInput), function(x){
2226 Exp_S <- rollapply(as.zoo(1:readEnd),width=3,by=3,
2227 function(codonNucs){
2228 SPos = which(LisGLs_MutationTypes[[x]][,codonNucs]=="S")
2229 sum( LisGLs_Targeting[[x]][,codonNucs][SPos], na.rm=T )
2230 }
2231 )
2232 })
2233
2234 Exp_R = matrix(unlist(LisGLs_R_Exp),nrow=nrow(matInput),ncol=readEnd/3,T)
2235 Exp_S = matrix(unlist(LisGLs_S_Exp),nrow=nrow(matInput),ncol=readEnd/3,T)
2236 return( list( "Expected_R"=Exp_R, "Expected_S"=Exp_S) )
2237 }
2238 }
2239
2240 # getObservedMutationsByCodon <- function(listMutations){
2241 # numbSeqs <- length(listMutations)
2242 # obsMu_R <- matrix(0,nrow=numbSeqs,ncol=readEnd/3,dimnames=list(c(1:numbSeqs),c(1:(readEnd/3))))
2243 # obsMu_S <- obsMu_R
2244 # temp <- mclapply(1:length(listMutations), function(i){
2245 # arrMutations = listMutations[[i]]
2246 # RPos = as.numeric(names(arrMutations)[arrMutations=="R"])
2247 # RPos <- sapply(RPos,getCodonNumb)
2248 # if(any(RPos)){
2249 # tabR <- table(RPos)
2250 # obsMu_R[i,as.numeric(names(tabR))] <<- tabR
2251 # }
2252 #
2253 # SPos = as.numeric(names(arrMutations)[arrMutations=="S"])
2254 # SPos <- sapply(SPos,getCodonNumb)
2255 # if(any(SPos)){
2256 # tabS <- table(SPos)
2257 # obsMu_S[i,names(tabS)] <<- tabS
2258 # }
2259 # }
2260 # )
2261 # return( list( "Observed_R"=obsMu_R, "Observed_S"=obsMu_S) )
2262 # }
2263
2264 getObservedMutationsByCodon <- function(listMutations){
2265 numbSeqs <- length(listMutations)
2266 obsMu_R <- matrix(0,nrow=numbSeqs,ncol=readEnd/3,dimnames=list(c(1:numbSeqs),c(1:(readEnd/3))))
2267 obsMu_S <- obsMu_R
2268 temp <- lapply(1:length(listMutations), function(i){
2269 arrMutations = listMutations[[i]]
2270 RPos = as.numeric(names(arrMutations)[arrMutations=="R"])
2271 RPos <- sapply(RPos,getCodonNumb)
2272 if(any(RPos)){
2273 tabR <- table(RPos)
2274 obsMu_R[i,as.numeric(names(tabR))] <<- tabR
2275 }
2276
2277 SPos = as.numeric(names(arrMutations)[arrMutations=="S"])
2278 SPos <- sapply(SPos,getCodonNumb)
2279 if(any(SPos)){
2280 tabS <- table(SPos)
2281 obsMu_S[i,names(tabS)] <<- tabS
2282 }
2283 }
2284 )
2285 return( list( "Observed_R"=obsMu_R, "Observed_S"=obsMu_S) )
2286 }
2287