trick/computeCCF.R at master · ruping/trick · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
computeCCF <- function(f, A, S, pu, pa, sAGP, nt, nb, prior="unknown", overadj=1.6, sigTh=0.90) {

    ccf = 0
    ccf2 = 0
    sd = 0
    cc <- seq(0.02, 1, by = 0.01)
    evoType = "A1/A2/B/C"
    N = A + S
    nc = nt * pa + 2 * (1 - pa)
    nc2 = nt * sAGP + 2 * (1 - sAGP)

    #message(paste(c(f,A,S,pu,pa,sAGP,nt,nb),collapse=" "))
    if (nb == 1 & nt == 2) {   #normal diploid
        ccf = 2*(f/pu)
        ff = pu*cc/2
        Ms = computeSD(N, S, ff)
        ccf2 <- Ms$M1
        sd <- Ms$SD
    }
    else if (nt == 1) {        #het deletion
        ccf = (f/pu)*nc
        ff.C <- pu*cc/nc                       #dbinom
        Ms.C <- computeSD(N, S, ff.C)          #dbinom
        ccf2 <- Ms.C$M1                        #dbinom
        sd <- Ms.C$SD
        fh.ea <- (sAGP * (nt - nb) + 1 - sAGP)/nc2
        fl.ea <- (sAGP * (nt - nb))/nc2
        fh.t <- sAGP/nc2
        fh.e <- (1 - sAGP)/nc2
        pEarly.a <- pbeta(fh.ea, S+1, A+1) - pbeta(fl.ea, S+1, A+1)
        pLate <- pbeta(fh.t, S+1, A+1)
        pEuploid <- pbeta(fh.e, S+1, A+1)
        Ptot <- pEarly.a + pLate + pEuploid
        cp.A <- pEarly.a/Ptot
        cp.CD <- 1 - cp.A
        cp.C <- pLate/Ptot
        cp.D <- pEuploid/Ptot
        cp.AC <- 1 - cp.D
        cp.AD <- 1 - cp.C
        if (cp.A >= sigTh) {
            evoType <- "A1"
        } else if (cp.CD >= sigTh & cp.C < sigTh & cp.D < sigTh) {
            evoType <- "B/C"
        } else if (cp.C >= sigTh) {
            evoType <- "B"
        } else if (cp.D >= sigTh) {
            evoType <- "C"
        } else if (cp.AC >= sigTh & cp.A < sigTh & cp.C < sigTh) {
            evoType <- "A1/B"
        } else if (cp.AD >= sigTh & cp.A < sigTh & cp.D < sigTh) {
            evoType <- "A1/C"
        }
    } else if (nb == 0 & nt == 0) {        #homozygous deletion
        evoType <- "C"
        ccf = (f*nc2)/sAGP
        ff.C = cc[cc<=(1-sAGP)]/nc2
        Ms.C = computeSD(N, S, ff.C, cc=cc[cc<=(1-sAGP)])
        ccf2 <- (Ms.C$M1)/pu                    #dbinom
        sd <- Ms.C$SD
        #message(paste(c(ccf,ccf2),collapse=" "))
    } else if (nb == 0 | nt == 2 * nb) {   #NLOH or other balanced CNAs
        fh.ea <- (sAGP * (nt - nb) + 1 - sAGP)/nc2
        fl.ea <- (sAGP * (nt - nb))/nc2
        fh.t <- sAGP/nc2
        fh.e <- (1 - sAGP)/nc2
        pEarly.a <- pbeta(fh.ea, S+1, A+1) - pbeta(fl.ea, S+1, A+1)
        pLate <- pbeta(fh.t, S+1, A+1)
        pEuploid <- pbeta(fh.e, S+1, A+1)
        Ptot <- pEarly.a + pLate + pEuploid
        cpEarly.a <- pEarly.a/Ptot
        cpLate.eup <- 1 - cpEarly.a
        cpLate <- pLate/Ptot
        cpEup <- pEuploid/Ptot
        if (Ptot > 0) {
            if (cpEarly.a >= sigTh){
                evoType <- "A1"
            } else if (cpLate.eup >= sigTh){
                evoType <- "B/C"
            } else if (cpLate >= sigTh){
                evoType <- "B"
            } else if (cpEup >= sigTh){
                evoType <- "C"
            }
        }
        allprobs = c(pEarly.a, pLate, pEuploid)
        names(allprobs) = c("pEarly.a", "pLate", "pEuploid")
        maxType = names(allprobs[match(max(allprobs),allprobs)])
        #if (maxType == "pEarly.a" & prior != "late") {
        if (evoType == "A1" & prior != "late") {
            ccf = (f/pu)*nc - (nt - nb - 1)*pa
            ff.A <- pu*(cc - pa + (nt - nb) * pa)/nc    #dbinom
            Ms.A <- computeSD(N, S, ff.A)               #dbinom
            ccf2 <- Ms.A$M1                             #dbinom
            sd <- Ms.A$SD
        } else {
            ccf = (f/pu)*nc
            ff.C <- pu*cc/nc                        #dbinom
            Ms.C <- computeSD(N, S, ff.C)           #dbinom
            ccf2 <- Ms.C$M1                         #dbinom
            sd <- Ms.C$SD
        }
    } else if (nb >= 1 & nt > 2) {
        fh.ea <- (sAGP * (nt - nb) + 1 - sAGP)/nc2
        fl.ea <- (sAGP * (nt - nb))/nc2
        fh.eb <- (nb * sAGP + 1 - sAGP)/nc2
        fl.eb <- nb * sAGP/nc2
        fh.t <- sAGP/nc2
        fh.e <- (1 - sAGP)/nc2
        pEarly.a <- pbeta(fh.ea, S+1, A+1) - pbeta(fl.ea, S+1, A+1)
        pEarly.b <- pbeta(fh.eb, S+1, A+1) - pbeta(fl.eb, S+1, A+1)
        pLate <- pbeta(fh.t, S+1, A+1)
        pEuploid <- pbeta(fh.e, S+1, A+1)
        Ptot <- pEarly.a + pEarly.b + pLate + pEuploid
        cp.A <- pEarly.a/Ptot
        cp.B <- pEarly.b/Ptot
        cp.C <- pLate/Ptot
        cp.D <- pEuploid/Ptot
        #message(paste(c(cp.A, cp.B, cp.C, cp.D), collapse=" "))
        cp.AB <- 1 - cp.C - cp.D
        cp.AC <- 1 - cp.B - cp.D
        cp.AD <- 1 - cp.B - cp.D
        cp.BC <- 1 - cp.A - cp.D
        cp.BD <- 1 - cp.A - cp.C
        cp.CD <- 1 - cp.A - cp.B
        cp.ABC <- 1 - cp.D
        cp.ABD <- 1 - cp.C
        cp.ACD <- 1 - cp.B
        cp.BCD <- 1 - cp.A
        if (Ptot > 0) {
            if (cp.A >= sigTh) {                   # earl A
                evoType = "A1"
            } else if (cp.B >= sigTh){
                evoType <- "A2"
            } else if (cp.C >= sigTh){
                evoType <- "B"
            } else if (cp.D >= sigTh){
                evoType <- "C"
            } else if (cp.CD >= sigTh & cp.C < sigTh & cp.D < sigTh){
                evoType <- "B/C"
            } else if (cp.AB >= sigTh & cp.A < sigTh & cp.B < sigTh){
                evoType <- "A1/A2"
            } else if (cp.AC >= sigTh & cp.A < sigTh & cp.C < sigTh){
                evoType <- "A1/B"
            } else if (cp.AD >= sigTh & cp.A < sigTh & cp.D < sigTh){
                evoType <- "A1/C"
            } else if (cp.BC >= sigTh & cp.B < sigTh & cp.C < sigTh){
                evoType <- "A2/B"
            } else if (cp.BD >= sigTh & cp.B < sigTh & cp.D < sigTh){
                evoType <- "A2/C"
            } else if (cp.BCD >= sigTh & cp.BC < sigTh & cp.BD < sigTh & cp.CD < sigTh & cp.B < sigTh & cp.C < sigTh & cp.D < sigTh){
                evoType <- "A2/B/C"
            } else if (cp.ABC >= sigTh & cp.BC < sigTh & cp.AB < sigTh & cp.AC < sigTh & cp.B < sigTh & cp.C < sigTh & cp.A < sigTh){
                evoType <- "A1/A2/B"
            } else if (cp.ABD >= sigTh & cp.AB < sigTh & cp.AD < sigTh & cp.BD < sigTh & cp.B < sigTh & cp.D < sigTh & cp.A < sigTh){
                evoType <- "A1/A2/C"
            } else if (cp.ACD >= sigTh & cp.AC < sigTh & cp.AD < sigTh & cp.CD < sigTh & cp.A < sigTh & cp.D < sigTh & cp.C < sigTh){
                evoType <- "A1/B/C"
            }
        }
        allprobs = c(pEarly.a, pEarly.b, pLate, pEuploid)
        names(allprobs) = c("pEarly.a", "pEarly.b", "pLate", "pEuploid")
        maxType = names(allprobs[match(max(allprobs),allprobs)])
        #if (maxType == "pEarly.a" & prior != "late") {
        if (evoType == "A1" & prior != "late") {
            ccf = (f/pu)*nc - (nt - nb - 1)*pa          #early A1
            ff.A <- pu*(cc - pa + (nt - nb) * pa)/nc    #dbinom
            #ff.A <- (cc - sAGP + (nt - nb) * sAGP)/nc2
            Ms.A <- computeSD(N, S, ff.A)               #dbinom
            ccf2 <- Ms.A$M1                             #dbinom
            #ccf2 = ccf2/pu
            sd <- Ms.A$SD
        #} else if (maxType == "pEarly.b" & prior != "late") {
        } else if (evoType == "A2" & prior != "late") {
            ccf = (f/pu)*nc - (nb - 1)* pa         #early A2
            ff.B <- pu*(cc - pa + nb * pa)/nc      #dbinom
            Ms.B <- computeSD(N, S, ff.B)          #dbinom
            ccf2 <- Ms.B$M1                        #dbinom
            sd <- Ms.B$SD
        } else {
            ccf = (f/pu)*nc                        #other
            ff.C <- pu*cc/nc                       #dbinom
            Ms.C <- computeSD(N, S, ff.C)          #dbinom
            ccf2 <- Ms.C$M1                        #dbinom
            sd <- Ms.C$SD
        }
    }
    if ( f > 0.1 & ccf >= overadj ) {    #correct for over-adjustment
        if (evoType != "A1") {
            if ( (nt-nb) >= 3 ) {
                ccf = (f/pu)*2
            } else if ( nt >= 2 & (nt-nb) < 3 ) {
                ccf = (f/pu)*nc - (nt - nb - 1)*pa
            } else {
                ccf = (f/pu)*nc
            }
        } else {
            ccf = f*2
        }
    }
    return(c(ccf, evoType, ccf2, sd))
}


computeSD <- function(N, S, f, cc=seq(0.02, 1, by = 0.01)) {
    M1list <- c()
    M2list <- c()
    MLElist <- c()
    for (ii in 1:length(N)) {
        PF <- sum(dbinom(S[ii], N[ii], f), na.rm = TRUE)
        M1 <- sum(dbinom(S[ii], N[ii], f) * cc, na.rm = TRUE)/PF
        M2 <- sum(dbinom(S[ii], N[ii], f) * cc^2, na.rm = TRUE)/PF
        M1list <- c(M1list, M1)
        M2list <- c(M2list, M2)
        MLElist <- c(MLElist, cc[which.max(dbinom(S[ii], N[ii], f))])
    }
    return(list(M1 = MLElist, SD = sqrt(M2list - M1list^2)))
}

effectivePurity <- function(p, Nt) {
    pu = Nt*p/(Nt*p + 2*(1-p))
    return(pu)
}