/
atlas_cnv.R
executable file
·480 lines (447 loc) · 26.9 KB
/
atlas_cnv.R
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
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
# --------------------------------------------------------------------------------
# atlas_cnv.R v0. called by main atlas_cnv.pl v0, Sep 11, 2018. Ted Chiang
# Copyright 2016-2018, Baylor College of Medicine Human Genome Sequencing Center.
# All rights reserved.
# --------------------------------------------------------------------------------
library(reshape2)
library("optparse")
library(plotrix)
option_list = list(
make_option(c("--rpkm_file"), type="character", default=NULL,
help="RPKM matrix data filename", metavar="character"),
make_option(c("--panel_file"), type="character", default="HG19_CAREseqv2_eMERGE3_LD_panel_design_3565_exons.tab",
help="panel design filename. 3 tab-delimited columns: 'Exon_Target', 'Gene_Exon', 'Call_CNV'
[default = %default]", metavar="character"),
make_option(c("--threshold_del"), type="numeric", default="-0.6",
help="log2 ratio cutoff to call a CNV deletion. [default= %default]", metavar="numeric"),
make_option(c("--threshold_dup"), type="numeric", default="0.4",
help="log2 ratio cutoff to call a CNV duplication. [default= %default]", metavar="numeric"),
# ???????#make_option(c("--threshold_exonQC_Zscore"), type="numeric", default="0.2",
# help="ExonQC cutoff to fail an exon target, ie. stddev > this value are flagged as failed. [default= %default]", metavar="numeric"),
make_option(c("--threshold_sampleQC"), type="numeric", default="0.2",
help="SampleQC cutoff to fail a sample, ie. stddev > this value are flagged as failed. [default= %default]", metavar="numeric"),
make_option(c("--threshold_sample_anova"), type="numeric", default="0.05",
help="In the coefficients of one way ANOVA of sample means, the Pr(>|t|) cutoff to flag a sample. [default= %default]", metavar="numeric")
);
opt_parser = OptionParser(option_list=option_list);
opt = parse_args(opt_parser);
if (is.null(opt$rpkm_file)) {
print_help(opt_parser)
stop("RPKM matrix datafile must be supplied (--rpkm_file).", call.=FALSE)
}
if (!file.exists(opt$panel_file)) {
print_help(opt_parser)
stop("Panel design file is not supplied, or default cannot be found. (--panel_file).", call.=FALSE)
}
# ccc: df of rpkms for all samples, all targets, ie. RPKM_matrix.IRC_MB-MID0075.CLEMRG_1pA (already 2066 columns, ie. only Call_CNV=Y)
rpkm_matrix_file = opt$rpkm_file
ccc = read.table(file=rpkm_matrix_file, header=T, check.names=F, row.names=1)
# Midpool summary result outputfile & midpool directory
midpool_summary_results = gsub('RPKM_matrix', 'atlas_cnv_summary', rpkm_matrix_file)
midpool_dir = gsub('/.*', '/', rpkm_matrix_file)
# ppp: df of the panel design file with 3 columns: ppp$Exon_Target, ppp$Gene_Exon, ppp$Call_CNV, ppp$RefSeq
panel_design_file = opt$panel_file
ppp = read.table(file=panel_design_file, header=T, check.names=F)
# =============================================================================
# CNV default cutoffs:
# 2^(0.5) = 1.414214.... so if RPKM ratio is >= 1.41 then, it's a DUP call.
# 2^(0.4692898) = 1.3844278
# 1-2^(-0.7) = 0.3844278
# 2^(-0.7) = 0.615572.... so if RPKM ratio is <= 0.61 then, it's a DEL call. <--OLD
# 2^(-0.415) = 0.75.... so if RPKM ratio is <= 0.75 then, it's a DEL call.
# =============================================================================
threshold_del = opt$threshold_del
threshold_dup = opt$threshold_dup
#threshold_exonQC = opt$threshold_exonQC
threshold_sampleQC = opt$threshold_sampleQC
threshold_sample_anova = opt$threshold_sample_anova
# some in-house functions:
find_median_sample = function (x) {
median_rpkm = mmm[x]
targ_coor = names(mmm[x]) # "1:1220087-1220186"
# ccc$"1:1220087-1220186" works but not as a variable ccc$"targ_coor".
# ccc[targ_coor] <-- column slice
# ccc[[targ_coor]] <-- column vector, same as ccc$"1:1220087-1220186"
median_rpkm = max(ccc[[targ_coor]][ which(ccc[[targ_coor]] <= median_rpkm) ]);
sample_idx = which(ccc[[targ_coor]] == median_rpkm)
sample_idx = sample_idx[1] # in case there are duplicate samples of 0 rpkm, then just pick one.
sample_id = rownames(ccc)[sample_idx]
med = c(targ_coor, median_rpkm, sample_id, sample_idx)
return (med)
}
my_row_sd = function(x) {
aaa = DDD[x,][FFF[x,]]
mysd = sd( aaa )
return (mysd)
}
my_row_count = function(x) {
count = sum(FFF[x,])
# cat ('n(for stddev cal)=', count, "\n")
return (count)
}
my_col_sd = function(x) {
aaa = DDD[,x][FFF[,x]]
mysd = sd( aaa )
return (mysd)
}
my_col_count = function(x) {
count = sum(FFF[,x])
# cat ('n(for stddev cal)=', count, "\n")
return (count)
}
my_col_sd_wo_outliers = function(x) {
aaa = DDD_wo_outliers[,x][FFF_wo_outliers[,x]]
mysd = sd( aaa )
return (mysd)
}
my_col_x_based_on_zscore_neg3_wo_outliers = function(x) {
aaa = DDD_wo_outliers[,x][FFF_wo_outliers[,x]]
mymean = mean( aaa )
mysd = sd( aaa )
#myx = (-4)*mysd + mymean
myx = (-2.576)*mysd + mymean
return (myx)
}
my_col_x_based_on_zscore_pos3_wo_outliers = function(x) {
aaa = DDD_wo_outliers[,x][FFF_wo_outliers[,x]]
mymean = mean( aaa )
mysd = sd( aaa )
#myx = (4)*mysd + mymean
myx = (2.576)*mysd + mymean
return (myx)
}
pass_fail = function(x) {
if (x==TRUE) { return ("Fail") }
else { return ("Pass") }
}
call_cnvs = function (x) { # calling cnv per sample via idx of ccc df
sample_id = rownames(fff)[x]
# --------------------------------------------------------------------------------------------------------------------------
# check to see if sample failed sample QC or ANOVA. If so, name the file: .cnv.FAILED_sampleQC or .cnv.FAILED_sampleANOVA
# --------------------------------------------------------------------------------------------------------------------------
anova_sample_id = paste('sample_id', sample_id, sep='')
if (is.element(sample_id, failed_samples) & !is.element(anova_sample_id, rownames(failed_samples_by_anova_pval_lt_5pct) )) {
outfile = paste(midpool_dir, sample_id, '.cnv.FAILED_sampleQC', sep='')
}
else if (!is.element(sample_id, failed_samples) & is.element(anova_sample_id, rownames(failed_samples_by_anova_pval_lt_5pct) )) {
outfile = paste(midpool_dir, sample_id, '.cnv.FAILED_sampleANOVA', sep='')
}
else if (is.element(sample_id, failed_samples) & is.element(anova_sample_id, rownames(failed_samples_by_anova_pval_lt_5pct) )) {
outfile = paste(midpool_dir, sample_id, '.cnv.FAILED_sampleQC_and_sampleANOVA', sep='')
}
else {
outfile = paste(midpool_dir, sample_id, '.cnv', sep='')
}
header = rbind(c('Exon_Target', 'Gene_Exon', 'cnv', 'log2R', 'rpkm', 'median_rpkm', 'Exon_Status', 'E_StDev', 'c_Score', 'RefSeq'))
write.table(header, file=outfile, append=T, quote=F, row.names=F, col.names=F, sep="\t")
# call dels
cat ("Calling dels on:", sample_id)
sample_dels=NULL
#dels = which(fff[x,1:(ncol(fff)-2)]<=threshold_del)
dels = which(fff[x,1:(ncol(fff)-2)]<=threshold_del & fff[x,1:(ncol(fff)-2)]<=threshold_del_soft)
if (length(dels)==0) { cat (" has no cnv dels.\n") }
else {
cat (" has cnv dels.\n")
Gene_Exon = ppp$Gene_Exon[ppp$Exon_Target %in% colnames(fff)[dels]]
RefSeq = ppp$RefSeq[ppp$Exon_Target %in% colnames(fff)[dels]]
cnv = rep('del', length(dels))
log2R = as.data.frame( t(fff[x, dels]) )
rpkm = as.data.frame( t( ccc[x, colnames(ccc) %in% colnames(fff)[dels]] ) )
median_rpkm = mmm$median_rpkm[mmm$targ_coor %in% colnames(fff)[dels]]
ExonQC_PF = is.element(colnames(fff)[dels], rownames(failed_exons_by_ExonQC))
ExonQC_PF = sapply(ExonQC_PF, pass_fail)
ExonQC_value = as.numeric(fff["ExonQC", dels])
Confidence_Score = round(log2R/ExonQC_value, 2)
sample_dels = data.frame( as.character(Gene_Exon), cnv, log2R, rpkm, median_rpkm, ExonQC_PF, ExonQC_value, Confidence_Score, RefSeq, stringsAsFactors=FALSE )
colnames(sample_dels) = c('Gene_Exon', 'cnv', 'log2R', 'rpkm', 'median_rpkm', 'Exon_Status', 'E_StDev', 'c_Score', 'RefSeq')
rownames(sample_dels) = colnames(fff)[dels]
write.table(sample_dels, file=outfile, append=T, quote=F, row.names=T, col.names=F, sep="\t")
}
# call dups
cat ("Calling dups on:", sample_id)
sample_dups=NULL
#dups = which(fff[x,1:(ncol(fff)-2)]>=threshold_dup)
dups = which(fff[x,1:(ncol(fff)-2)]>=threshold_dup & fff[x,1:(ncol(fff)-2)]>=threshold_dup_soft)
if (length(dups)==0) { cat (" has no cnv dups.\n") }
else {
cat (" has cnv dups.\n")
Gene_Exon = ppp$Gene_Exon[ppp$Exon_Target %in% colnames(fff)[dups]]
RefSeq = ppp$RefSeq[ppp$Exon_Target %in% colnames(fff)[dups]]
cnv = rep('dup', length(dups))
log2R = as.data.frame( t(fff[x, dups]) )
rpkm = as.data.frame( t( ccc[x, colnames(ccc) %in% colnames(fff)[dups]] ) )
median_rpkm = mmm$median_rpkm[mmm$targ_coor %in% colnames(fff)[dups]]
ExonQC_PF = is.element(colnames(fff)[dups], rownames(failed_exons_by_ExonQC))
ExonQC_PF = sapply(ExonQC_PF, pass_fail)
ExonQC_value = as.numeric(fff["ExonQC", dups])
Confidence_Score = round(log2R/ExonQC_value, 2)
sample_dups = data.frame( as.character(Gene_Exon), cnv, log2R, rpkm, median_rpkm, ExonQC_PF, ExonQC_value, Confidence_Score, RefSeq, stringsAsFactors=FALSE )
colnames(sample_dups) = c('Gene_Exon', 'cnv', 'log2R', 'rpkm', 'median_rpkm', 'Exon_Status', 'E_StDev', 'c_Score', 'RefSeq')
rownames(sample_dups) = colnames(fff)[dups]
write.table(sample_dups, file=outfile, append=T, quote=F, row.names=T, col.names=F, sep="\t")
}
# plots
plot_cnvs(rbind(sample_dels, sample_dups), sample_id)
}
plot_cnvs = function(x,y) { # x is df of sample_dels/dups, y is sample_id
pdffile = paste(midpool_dir, y, '.pdf', sep='')
pdf.options(bg='honeydew')
pdf(file=pdffile);
par(mfrow=c(2,1))
sample_id = gsub('[_A-Z]', '', y)
if (length(nrow(x))==0) {
plot.new()
text (0.5, 1, paste('No called CNVs for: ', sample_id, sep=''))
garbage = dev.off()
}
else {
# gene plots
cnv_genes = rle( gsub('-.*', '', x$Gene_Exon) ) # rle is like unix:uniq, rm the transcript & exon numbering of Gene_Exon: MTHFR-001_11, ie. '-001_11'.
#cnv_genes_2plot = cnv_genes$values[ which(cnv_genes$lengths >= 1) ] # Do gene plots w/ 1+ exons
cnv_genes_2plot = cnv_genes$values[ which(cnv_genes$lengths >= 2) ] # Do gene plots w/ 2+ exons
if (length(cnv_genes_2plot) > 0) { # are there any genes to plot?
for (g in 1:length(cnv_genes_2plot)) {
gene = cnv_genes_2plot[g]
gene_pattern = paste('^', cnv_genes_2plot[g], '-', sep='') # bug: grep 'NF2' gets 'RNF207'
cat ('Gene plot for sample: ', y, 'using gene pattern: ', gene_pattern, "\n")
gene_exon_all = ppp$Gene_Exon[ grep(gene_pattern, ppp$Gene_Exon) ] # ppp: df of the panel design file with 3 columns: ppp$Exon_Target, ppp$Gene_Exon, ppp$Call_CNV
gene_targ_coor_all = ppp$Exon_Target[ grep(gene_pattern, ppp$Gene_Exon) ] # ppp: df of the panel design file with 3 columns: ppp$Exon_Target, ppp$Gene_Exon, ppp$Call_CNV
gene_targ_coor_start = which(colnames(fff) == gene_targ_coor_all[1])
gene_targ_coor_stop = which(colnames(fff) == gene_targ_coor_all[length(gene_targ_coor_all)])
lll = fff[y, gene_targ_coor_start:gene_targ_coor_stop] # df: one row w/column slice of log2R for given gene targets.
lll = 2^(unlist(lll)) # convert log2R back to ratio for barplot
names(lll) = gene_exon_all
cnv_gene_exons = x$Gene_Exon[ grep(gene_pattern, x$Gene_Exon) ]
col_vector=rep('grey', length(gene_exon_all))
col_vector[ gene_exon_all %in% cnv_gene_exons ] = 'red'
mean_Confidence_Score = round( mean( x$c_Score[ grep(gene_pattern, x$Gene_Exon) ] ), 2)
barplot(lll, col=col_vector, ylim=c(0,2), ylab='sample / median' , las=2, cex.names=0.5, axis.lty=1, main='Gene plot', space=0)
mtext(gene, cex=0.7)
legend("top", c(paste('mean c_Score', mean_Confidence_Score, sep=' = ')), col=c('black'), pch=15, cex=0.5, bg="transparent")
legend("topright", c('c_Score: (-)loss, (+)gain', 'High: abs(c_Score) > 5', 'Med: abs(c_Score) = 3-5', 'Low: abs(c_Score) < 3'), col=c('black', 'grey', 'grey', 'grey'), pch=15, cex=0.5, bg="transparent")
#legend("topright", c('dip: S/M = 0.85-1.15', 'del(het): S/M = 0.4-0.6', 'del(hom): S/M = 0-0.1', 'dup: S/M = >1.2': ), col=c('black', 'black', 'black', 'black'), pch=15, cex=0.5, bg="transparent")
axis(1, at=match(cnv_gene_exons,names(lll))-0.5, labels=cnv_gene_exons, las=2, cex.axis=0.5, col.axis='red')
abline(h=0.5)
abline(h=1)
table_rows = grep(gene_pattern, x$Gene_Exon) # each plot.new() can only handle 12 rows per table, so make appropriate number of plots.
if (length(table_rows) <= 12) {
plot.new()
p_start = 1
p_stop = length(table_rows)
addtable2plot(x='top', table=x[ table_rows[p_start:p_stop], 1:8], cex=0.75, bty='o', hlines=T, vlines=T, bg='honeydew', xpad=0.25, ypad=1)
}
else if (length(table_rows) > 12) {
num_of_tables = floor(length(table_rows)/12)
p_stop = 0
for ( p in 1:num_of_tables ) {
plot.new()
p_start = p_stop + 1
p_stop = p*12
addtable2plot(x='top', table=x[ table_rows[p_start:p_stop], 1:8], cex=0.75, bty='o', hlines=T, vlines=T, bg='honeydew', xpad=0.25, ypad=1)
}
if ((length(table_rows) %% 12) > 0) { # a last table if needed.
plot.new() # last table
p_start = p_stop + 1
p_stop = length(table_rows)
addtable2plot(x='top', table=x[ table_rows[p_start:p_stop], 1:8], cex=0.75, bty='o', hlines=T, vlines=T, bg='honeydew', xpad=0.25, ypad=1)
}
}
}
}
# exon plots
for (i in 1:nrow(x)) {
cnv_targ_coor = rownames(x)[i]
cnv_gene_exon = x$Gene_Exon[i]
cnv_exon_qc_value = fff["ExonQC",cnv_targ_coor]
cnv_exon_qc_status = x$Exon_Status[i]
cnv_exon_confidence_score = x$c_Score[i]
lll = fff[1:(nrow(fff)-2), cnv_targ_coor , drop=FALSE] # df: column slice of log2R for given target. Note:drop=FALSE keeps dimnames when subsetting to one dim.
nnn = gsub('[_A-Z]', '', rownames(lll)) # remove the '_HMWCYBCXX' and 'IDMB', for x-axis labels in the plot.
j = match(cnv_targ_coor, mmm$targ_coor)
median_rpkm = mmm$median_rpkm[j]
median_sample_id = mmm$sample_id[j]
median_sample_id = gsub('[_A-Z]', '', median_sample_id)
median_sample_idx = mmm$sample_idx[j]
col_vector=rep('grey', dim(ccc)[1])
threshold_del_soft_this_exon = threshold_del_soft[[cnv_gene_exon]]
threshold_dup_soft_this_exon = threshold_dup_soft[[cnv_gene_exon]]
col_vector[ which(lll<=threshold_del) ] = 'darkgoldenrod3' # hom/het dels
col_vector[ which(lll>=threshold_dup) ] = 'darkgoldenrod3' # dups
col_vector[ which(lll<=threshold_del & lll<=threshold_del_soft_this_exon) ] = 'darkgoldenrod3' # hom/het dels
col_vector[ which(lll>=threshold_dup & lll>=threshold_dup_soft_this_exon) ] = 'darkgoldenrod3' # dups
col_vector[ which(rownames(ccc)==y)] = 'red' # this given sample
col_vector[ median_sample_idx ] = 'blue' # median sample used for reference
lll=as.vector(t(2^(lll))) # convert log2R back to ratio for barplot
names(lll)=nnn # as.vector conversion causes rownames to be lost
barplot(lll, col=col_vector, ylim=c(0,2), ylab='sample / median', las=2, cex.names=0.5, axis.lty=1, main='Exon plot', space=0)
mtext(paste(cnv_gene_exon, cnv_targ_coor, sep=', '), cex=0.7)
legend("top", c(paste('exon status', cnv_exon_qc_status, sep=': '), paste('E[StDev]', cnv_exon_qc_value, sep=' = '), paste('c_Score', cnv_exon_confidence_score, sep=' = ')), col=c('black', 'black', 'black'), pch=15, cex=0.5, bg="transparent")
legend("topright", c('sample', 'median'), col=c('red', 'blue'), pch=15, cex=0.5, bg="transparent")
axis(1, at=match(sample_id,names(lll))-0.5, labels=sample_id, las=2, cex.axis=0.5, col.axis='red')
axis(1, at=match(median_sample_id,names(lll))-0.5, labels=median_sample_id, las=2, cex.axis=0.5, col.axis='blue')
abline(h=0.5)
abline(h=1)
}
garbage = dev.off()
}
}
remove_5pc_outliers = function (aaa) {
for (i in seq_along(aaa)) {
a = which( aaa[[i]] < ( -1.96*sd(aaa[[i]]) + mean(aaa[[i]]) ) )
b = which( aaa[[i]] > ( 1.96*sd(aaa[[i]]) + mean(aaa[[i]]) ) )
aaa[a,i]=NA
aaa[b,i]=NA
}
return(aaa)
}
# mmm: df of medians for each target: rownames(targ_coor) ==> colnames(mmm$median_rpkm, mmm$sample_id, mmm$sample_idx)
# ie. the artificial sample of reference medians.
# mmm = apply(ccc, 2, median) # 2 for column, 1 for row.
ccc_wo_outliers = remove_5pc_outliers(ccc)
mmm = apply(ccc_wo_outliers, 2, median, na.rm=TRUE)
mmm = lapply(seq_along(mmm), find_median_sample)
mmm = data.frame(matrix(unlist(mmm), ncol=4, byrow=T), stringsAsFactors=FALSE) # suppresses data.frame()’s default behaviour which turns strings into factors.
colnames(mmm) = c('targ_coor', 'median_rpkm', 'sample_id', 'sample_idx')
mmm = data.frame(targ_coor=mmm$targ_coor, median_rpkm=as.integer(mmm$median_rpkm), sample_id=mmm$sample_id, sample_idx=as.integer(mmm$sample_idx), stringsAsFactors=FALSE)
# ddd: df of log2 ratios of: samples (rows) vs. targets
ddd = as.data.frame(t( log2( t(ccc)/mmm$median_rpkm) ), stringsAsFactors=FALSE) # a matrix has to be coerced into a dataframe!
DDD = as.matrix(ddd) # apply function requires matrix.
FFF = is.finite(DDD) # T/F matrix to rm the Inf, NaN.
# ---------------------------------------------------------------------------------------------
# Same as above 3, but on ccc_wo_outliers. THIS IS FOR COMPUTING THE ExonQC_wo_outliers vector.
# ---------------------------------------------------------------------------------------------
ddd_wo_outliers = as.data.frame(t( log2( t(ccc_wo_outliers)/mmm$median_rpkm) ), stringsAsFactors=FALSE) # a matrix has to be coerced into a dataframe!
DDD_wo_outliers = as.matrix(ddd_wo_outliers) # apply function requires matrix.
FFF_wo_outliers = is.finite(DDD_wo_outliers) # T/F matrix to rm the Inf, NaN.
# ---------------------------------------------------------------------------------------------
# Vector of exon stddev wo_outliers
ExonQC = sapply(seq_along(colnames(ddd)), my_col_sd)
ExonQC_wo_outliers = sapply(seq_along(colnames(ddd_wo_outliers)), my_col_sd_wo_outliers)
threshold_del_soft = sapply(seq_along(colnames(ddd_wo_outliers)), my_col_x_based_on_zscore_neg3_wo_outliers) # The 'x' log2Ratio threshold (soft and data-derived) to call deletions (wo/outliers)
threshold_dup_soft = sapply(seq_along(colnames(ddd_wo_outliers)), my_col_x_based_on_zscore_pos3_wo_outliers) # The 'x' log2Ratio threshold (soft and data-derived) to call deletions (wo/outliers)
names(ExonQC) = ppp$Gene_Exon[ which(ppp$Call_CNV=='Y') ]
names(ExonQC_wo_outliers) = ppp$Gene_Exon[ which(ppp$Call_CNV=='Y') ]
names(threshold_del_soft) = ppp$Gene_Exon[ which(ppp$Call_CNV=='Y') ]
names(threshold_dup_soft) = ppp$Gene_Exon[ which(ppp$Call_CNV=='Y') ]
exon_qc_outfile = gsub('RPKM_matrix', 'ExonQC', rpkm_matrix_file)
exon_qc_wo_outliers_outfile = gsub('RPKM_matrix', 'ExonQC_wo_outliers', rpkm_matrix_file)
threshold_del_soft_outfile = gsub('RPKM_matrix', 'Exon_threshold_del_soft_wo_outliers', rpkm_matrix_file)
threshold_dup_soft_outfile = gsub('RPKM_matrix', 'Exon_threshold_dup_soft_wo_outliers', rpkm_matrix_file)
suppressWarnings(write.table(ExonQC, file=exon_qc_outfile, append=T, quote=F, row.names=T, col.names=T, sep="\t"))
suppressWarnings(write.table(ExonQC_wo_outliers, file=exon_qc_wo_outliers_outfile, append=T, quote=F, row.names=T, col.names=T, sep="\t"))
suppressWarnings(write.table(threshold_del_soft, file=threshold_del_soft_outfile, append=T, quote=F, row.names=T, col.names=T, sep="\t"))
suppressWarnings(write.table(threshold_dup_soft, file=threshold_dup_soft_outfile, append=T, quote=F, row.names=T, col.names=T, sep="\t"))
# 95%, 97.5%, 99%, 99.95%, 99.99% => zscore: 1.645, 1.96, 2.576, 3.291, 4
#threshold_exonQC = 2.576 * sd(ExonQC_wo_outliers) + mean(ExonQC_wo_outliers)
threshold_exonQC = 3.291 * sd(ExonQC_wo_outliers) + mean(ExonQC_wo_outliers)
#threshold_exonQC = 100 # To disable ExonQC metric.
names(threshold_exonQC) = 'threshold_exonQC'
cat ("ExonQC threshold based on 99% of sd distribution is: ", threshold_exonQC, "\n")
cat ("CNV Exon threshold del/dup: ", threshold_del, ", ", threshold_dup, "\n")
#sample_size = round(0.1*length(ExonQC_wo_outliers_for_Call_CNV_Y)) # sample 10% of the data to compute a exonQC threshold.
#threshold_exonQC = NULL
#for (i in 1:1) {
# hhh = sample(ExonQC_wo_outliers_for_Call_CNV_Y, size=sample_size, replace=F)
# threshold_exonQC = c(threshold_exonQC, 1.96 * sd(hhh) + mean(hhh)) # the sd cutoff is 95% of this background sampling
#}
#threshold_exonQC = mean(threshold_exonQC)
#cat ("ExonQC [ave of 1 events, each sampling 10% of bkgrd exon SD thresholded at 95% is]:", threshold_exonQC, "\n")
# add SampleQC stddev of log2 ratios, and row counts of data w/o Inf, NaN.
SampleQC = sapply(seq_along(rownames(ddd)), my_row_sd)
Sample_count_wo_InfNaN = sapply(seq_along(rownames(ddd)), my_row_count)
# add ExonQC of target stddev of log2 ratios, and col counts of data w/o Inf, NaN.
Exon_count_wo_InfNaN = sapply(seq_along(colnames(ddd)), my_col_count)
# ==> as two columns at the right end of matrix.
fff = cbind( ddd, data.frame( SampleQC = SampleQC, Sample_count_wo_InfNaN = Sample_count_wo_InfNaN ) )
# ==> as two rows at bottom end of matrix.
eee = as.data.frame(rbind( c(ExonQC_wo_outliers, NA, NA), c(Exon_count_wo_InfNaN, NA, NA) ))
rownames(eee) = c('ExonQC', 'Exon_count_wo_InfNaN')
colnames(eee) = colnames(fff)
fff = rbind(fff, eee)
fff = round(fff, 2)
rm(DDD, FFF, eee)
# fff: df of ddd w/ removed panel targets not used for CNV calling ie. (rm N ==> ppp$Call_CNV)
# fff = ddd[, which(ppp$Call_CNV=='Y') ]
# fff = cbind( fff, SampleQC=ddd$SampleQC, Sample_count_wo_InfNaN=ddd$Sample_count_wo_InfNaN )
# rrr: df of sample rpkm means & stddev: rownames(sampleids) ==> colnames(rrr$rpkm_mean, rrr$rpkm_stddev)
rrr = data.frame(rpkm_mean = rowMeans(ccc), rpkm_stddev = apply(ccc, 1, sd))
rrr = round(rrr, 2)
sss = data.frame(rpkm_mean = rowMeans(ccc), rpkm_stddev = apply(ccc, 1, sd), SampleQC = SampleQC)
sss = round(sss, 2)
# write sample rpkm means & stddev results to outfile:
sample_mean_stddev_outfile = gsub('RPKM_matrix', 'Sample_RPKM-means-stddevs_log2-stddevs', rpkm_matrix_file)
suppressWarnings(write.table(sss, file=sample_mean_stddev_outfile, append=T, quote=F, row.names=T, col.names=T, sep="\t"))
# c_score matrix and pvalue matrix
cscore_outfile = gsub('RPKM_matrix', 'Cscore_matrix', rpkm_matrix_file)
pval_outfile = gsub('RPKM_matrix', 'Pval_matrix', rpkm_matrix_file)
c_scores = as.matrix( t( t(ddd) / ExonQC_wo_outliers) ) # apply function requires matrix.
#pvals = as.data.frame( apply(c_scores, 2, function(x) { 2*(1 - pnorm(abs(x))) } ), stringsAsFactors=FALSE ) # a matrix has to be coerced into a dataframe! two tailed pvalues
pvals = as.data.frame( apply(c_scores, 2, function(x) { 1 - pnorm(abs(x)) } ), stringsAsFactors=FALSE ) # a matrix has to be coerced into a dataframe! one tail pvalues
c_scores = as.data.frame(c_scores)
c_scores = round(c_scores, 2)
c_scores = cbind(Samples = rownames(c_scores), c_scores)
pvals = round(pvals, 2)
pvals = cbind(Samples = rownames(pvals), pvals)
suppressWarnings(write.table(c_scores, file=cscore_outfile, append=T, quote=F, row.names=F, col.names=T, sep="\t"))
suppressWarnings(write.table(pvals, file=pval_outfile, append=T, quote=F, row.names=F, col.names=T, sep="\t"))
# Failed exons by Exon QC > computed exon sd without outliers (row slice is a dataframe already)
fff_idx = which(fff["ExonQC",]>threshold_exonQC)
if (length(fff_idx)==1) { # when row slice is only 1 datapoint which is not automatically a row slice.
fff_idx = which(fff["ExonQC",]>threshold_exonQC)
failed_exons_by_ExonQC = as.data.frame(fff["ExonQC", fff_idx])
rownames(failed_exons_by_ExonQC) = colnames(fff)[fff_idx]
colnames(failed_exons_by_ExonQC) = 'ExonQC'
} else { # when row slice has 0 datapoints or 2 or more which would be a row slice.
failed_exons_by_ExonQC = as.data.frame( t( fff["ExonQC", which(fff["ExonQC",]>threshold_exonQC)] ) )
}
Exon_count_wo_InfNaN = as.data.frame( t( fff["Exon_count_wo_InfNaN", rownames(failed_exons_by_ExonQC)] ) )
failed_exons_by_ExonQC = data.frame(
Gene_Exon=as.character(ppp$Gene_Exon[ppp$Exon_Target %in% rownames(failed_exons_by_ExonQC)]),
failed_exons_by_ExonQC,
Exon_count_wo_InfNaN,
stringsAsFactors=FALSE
)
# Failed samples by SampleQC > 0.2
failed_samples = rownames(fff)[which(fff$SampleQC>threshold_sampleQC)]
failed_samples_by_SampleQC = data.frame(
failed_samples = failed_samples,
sd_SampleQC = fff$SampleQC[which(fff$SampleQC>threshold_sampleQC)],
Sample_count_wo_InfNaN = fff$Sample_count_wo_InfNaN[which(fff$SampleQC>threshold_sampleQC)],
rpkm_mean = rrr$rpkm_mean[rownames(rrr) %in% failed_samples],
stringsAsFactors=FALSE
)
# sample rpkm stats
sample_rpkm_stats = data.frame(sample_stats = c(min(rrr$rpkm_mean), max(rrr$rpkm_mean), median(rrr$rpkm_mean), min(rrr$rpkm_stddev), max(rrr$rpkm_stddev), median(rrr$rpkm_stddev)))
rownames(sample_rpkm_stats) = c('min_rpkm', 'max_rpkm', 'median_rpkm', 'min_stddev', 'max_stddev', 'median_stddev')
# Failed samples by anova, use rrr$rpkm_mean
aaa = melt(t(ccc)) # from reshape2 library to collapse df to data for lm.
colnames(aaa) = c('targ_coor', 'sample_id', 'rpkm')
fstatistic = anova(lm(rpkm ~ sample_id, data=aaa))$F[1]
pvalue = anova(lm(rpkm ~ sample_id, data=aaa))$P[1]
s = summary(lm(rpkm ~ sample_id, data=aaa))
anova = data.frame(fstatistic,pvalue)
failed_samples_by_anova_pval_lt_5pct = data.frame(Prob_gt_t = s$coefficients[ which(s$coefficients[,4]<threshold_sample_anova), 4])
# write midpool summary results of above 5 dfs:
suppressWarnings(write.table(threshold_exonQC, file=midpool_summary_results, append=T, quote=F, row.names=T, col.names=T, sep="\t"))
suppressWarnings(write.table(failed_exons_by_ExonQC, file=midpool_summary_results, append=T, quote=F, row.names=T, col.names=T, sep="\t"))
suppressWarnings(write.table(failed_samples_by_SampleQC, file=midpool_summary_results, append=T, quote=F, row.names=F, col.names=T, sep="\t"))
suppressWarnings(write.table(sample_rpkm_stats, file=midpool_summary_results, append=T, quote=F, row.names=T, col.names=T, sep="\t"))
suppressWarnings(write.table(anova, file=midpool_summary_results, append=T, quote=F, row.names=T, col.names=T, sep="\t"))
suppressWarnings(write.table(failed_samples_by_anova_pval_lt_5pct, file=midpool_summary_results, append=T, quote=F, row.names=T, col.names=T, sep="\t"))
# Calling CNVs on fff dataframe: call dels first since it creates the .cnv outputfile.
garbage = sapply(seq_along(rownames(ccc)), call_cnvs)
# -------under development, think about this-----------#
# dels_targs = cbind(colnames(fff)[dels], rep(1, length(dels)))
# dels_targs = split( dels_targs, seq(nrow(dels_targs)) )
# x = c("7:6013030-6013173", 1)
# targ_zscore = lapply( dels_targs, get_targ_zscore)
# get_targ_zscore = function (x) { # x is a list.
# i = x[1]
# j = as.integer(x[2])
# x1 = ccc[[i]][j]
# xbar = mean(ccc[[i]])
# mu = sd(ccc[[i]])
# zscore = (x1 - xbar)/mu
# return(zscore)
# }
# targ_pval = # write function to get the pval of the targ_zscore above.