_code/overseq2.r~ at master · cssmillie/_code · 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
library(optparse)
if(interactive()){
    args = list()
    args$map = '~/aviv/summary/test.map'
    args$ming = 1000
    args$num = 5
    args$out = 'test'
} else{
option_list = list(make_option('--dir', help='directory name'),
                   make_option('--ming', help='genes per cell cutoff', type='integer'),
                   make_option('--out', help='output prefix'),
                   make_option('--num', help='number of downsamples', default=10, type='integer')
                   )
args = parse_args(OptionParser(option_list=option_list))
}

library(vegan)

# subsample dge
subsample_dge = function(x1, y1, y2){
    # x1, x2 = number of umis (old, new)
    # y1, y2 = number of reads (old, new)
    r1 = rmultinom(1, y1-x1, rep(1, x1)) + 1
    r2 = rrarefy(r1, y2)
    x2 = sum(r2 > 0)
    return(x2)
}

# reads per collection
x.r = c()

# cells per collection
y.h = c()
y.m = c()
y.t = c()

# get folder name
folder = args$dir

# load human dge
print('loading human dge')
h.dge = read.table(paste('~/aviv/data/', folder, '/dge/human.dge.txt.gz', sep=''), header=T, row.names=1, sep='\t')

# load human reads
print('loading human reads')
h.reads = read.table(paste('~/aviv/data/', folder, '/dge/human.reads.dge.txt.gz', sep=''), header=T, row.names=1, sep='\t')

# load mouse dge
print('loading mouse dge')
m.dge = read.table(paste('~/aviv/data/', folder, '/dge/mouse.dge.txt.gz', sep=''), header=T, row.names=1, sep='\t')

# load mouse reads
print('loading mouse reads')
m.reads = read.table(paste('~/aviv/data/', folder, '/dge/mouse.reads.dge.txt.gz', sep=''), header=T, row.names=1, sep='\t')

# count total reads
total_reads = 0
if(ncol(h.dge) > 1){total_reads = total_reads + sum(h.reads)}
if(ncol(m.dge) > 1){total_reads = total_reads + sum(m.reads)}

# get read counts
max_reads = total_reads
min_reads = 10**6
x.i = seq(log10(min_reads), log10(max_reads), length.out=args$num)

# count collections
ncoll = 0

# subsample human reads
if(ncol(h.dge) > 1){
    h.coll = length(unique(sapply(strsplit(colnames(h.dge), '\\.'), '[', 1)))
    ncoll = h.coll
    for(xi in x.i){
        new_reads = rrarefy(h.reads, round((10**xi)*rowSums(h.reads)/sum(h.reads)))
        new_dge = matrix(0, nrow=nrow(h.dge), ncol=ncol(h.dge))
        subsample_dge = function(i,j){
            x1 = h.dge[i,j]
            y1 = h.reads[i,j]
            y2 = new_reads[i,j]
            r1 = rmultinom(1, y1-x1, rep(1,x1)) + 1
            r2 = rrarefy(r1, y2)
            x2 = sum(r2 > 0)
            return(x2)
        }
        subsample_dge = Vectorize(subsample_dge)
        ijs = which(new_reads > 0, arr.ind=T)
        new_dge[ijs] = apply(ijs, 1, function(a){subsample_dge(a[[1]], a[[2]])})
        h.cells = sum(colSums(new_dge > 0) >= args$ming)
        print(h.cells)
        y.h = c(y.h, h.cells/h.coll)
    }
} else {
    y.h = c(y.h, rep(0, length(x.i)))
}

# subsample mouse reads
if(ncol(m.dge) > 1){
    m.coll = length(unique(sapply(strsplit(colnames(m.dge), '\\.'), '[', 1)))
    ncoll = m.coll
    for(xi in x.i){
        new_reads = rrarefy(m.reads, round((10**xi)*rowSums(m.reads)/sum(m.reads)))
        new_dge = matrix(0, nrow=nrow(m.dge), ncol=ncol(m.dge))
        subsample_dge = function(i,j){
            x1 = m.dge[i,j]
            y1 = m.reads[i,j]
            y2 = new_reads[i,j]
            r1 = rmultinom(1, y1-x1, rep(1,x1)) + 1
            r2 = rrarefy(r1, y2)
            x2 = sum(r2 > 0)
            return(x2)
        }
        subsample_dge = Vectorize(subsample_dge)
        ijs= which(new_reads > 0, arr.ind=T)
        new_dge[ijs] = apply(ijs, 1, function(a){subsample_dge(a[[1]], a[[2]])})
        m.cells = sum(colSums(new_dge > 0) >= args$ming)
        print(m.cells)
        y.m = c(y.m, m.cells/m.coll)
    }
} else{
    y.m = c(y.m, rep(0, length(x.i)))
}

# reads per collection
x.r = c(x.r, x.i/ncoll)

# total genes per collection
y.t = y.h + y.m

# save output to file
z = list(x.r=x.r, y.h=y.h, y.m=y.m, y.t=y.t)
saveRDS(z, paste(args$out, 'overseq.rds', sep='.'))