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
|
from __future__ import absolute_import
import numpy as np
import bisect
def rightmostBinSearch(vec, val):
"""
Return the rightmost position in the vector vec of val. If val is
absent then we return the leftmost position of the value:
min(vec[vec > val]). If val is greater than all elements in vec we
return len(vec).
"""
assert(len(vec) > 0)
i = bisect.bisect_left(vec, val)
if (len(vec) == i):
return(i)
while (i + 1 < len(vec) and vec[i + 1] == val):
i += 1
return(i)
def leftmostBinSearch(vec, val):
"""
Return the leftmost position in the vector vec of val. If val is
absent then we return the lefternmost position for the value:
max(vec[vec < val]). The time complexity here is potentially worse
than log(n) because of the extra step of walking backwards.
"""
assert(len(vec) > 0)
i = bisect.bisect_left(vec, val)
if (i == 0):
return(i)
elif (i == len(vec)):
v = vec[i-1]
i -= 1
else:
v = vec[i]
if (v > val):
i -= 1
while (i > 0 and vec[i-1] == vec[i]):
i -= 1
return(i)
def getOverlappingRanges(tStart, tEnd, nBack, nOverlap, rangeStart, rangeEnd):
"""
Return indices overlapping the range defined by [rangeStart,
rangeEnd). Here tStart, tEnd, nBack, nOverlap are vectors of
length n sorted according to tStart and tEnd. The vectors nBack
and nOverlap are typically produced by computeIndices[DP].
"""
assert(rangeEnd > rangeStart and
len(tStart) == len(tEnd) == len(nBack) == len(nOverlap))
lM = leftmostBinSearch(tStart, rangeStart)
lM = lM - nBack[lM]
rM = rightmostBinSearch(tStart, rangeEnd - .5)
assert(rM >= lM and rM >= 0 and lM >= 0)
if (lM == rM):
return(np.array([], dtype = "uint32"))
else:
# We only keep the reads in the range lM .. rM that
# actually overlap the range, as determined by
# tEnd > rangeStart
idxs = np.arange(lM, rM, dtype = "uint32") # lM .. rM
toKeep = tEnd[idxs] > rangeStart
return(idxs[toKeep])
def projectIntoRange(tStart, tEnd, winStart, winEnd):
"""
Find coverage in the range [winStart, winEnd) implied by tStart,
tEnd vectors. Coverage can be most efficiently calculated by
first obtaining all reads overlapping the range using the
getOverlappingRanges function then projecting them into the same
or smaller range
"""
assert(len(tStart) == len(tEnd))
res = np.zeros(shape=winEnd-winStart, dtype=np.uint)
# Clip to window and translate.
# Be careful to avoid underflow!
tStart_ = np.clip(tStart, winStart, winEnd) - winStart
tEnd_ = np.clip(tEnd, winStart, winEnd) - winStart
for (s, e) in zip(tStart_, tEnd_):
res[s:e] += 1
return res
def makeReadLocator(cmpH5, refSeq):
"""
Return a function which can be called iteratively to find reads
quickly.
"""
if not cmpH5.isSorted: raise Exception("CmpH5 is not sorted")
refInfo = cmpH5.referenceInfo(refSeq)
offStart, offEnd = refInfo.StartRow, refInfo.EndRow
if (offEnd - offStart > 0):
refAlignIdx = cmpH5.alignmentIndex[offStart:offEnd, ]
returnEmpty = False
else:
refAlignIdx = cmpH5.alignmentIndex[1:2, ]
returnEmpty = True
def f(rangeStart, rangeEnd, justIndices = False):
if returnEmpty:
## This looks strange, but the idea is that a rowless matrix
## still has columns and these are what I want to preserve --
## h5py objects cannot be subset by a vector of length 0,
## however, numpy allows this.
idxs = np.array([], dtype = 'uint32')
else:
idxs = getOverlappingRanges(refAlignIdx.tStart, refAlignIdx.tEnd,
refAlignIdx.nBackRead, refAlignIdx.nReadOverlap,
rangeStart, rangeEnd)
if justIndices:
return(idxs + offStart)
else:
return(refAlignIdx[idxs,])
return f
def getReadsInRange(cmpH5, coords, justIndices = False):
"""
Return an ndarray representing the portion of the reads which
overlap the range specfied by coords, where coords is a
three-tuple composed of (refSeqID, rangeStart, rangeEnd). Here,
cmpH5 is an hdf5 object representing a pointer to a sorted cmp.h5
file.
"""
if not cmpH5.isSorted: raise Exception("CmpH5 is not sorted")
return makeReadLocator(cmpH5, coords[0])(coords[1], coords[2], justIndices)
def getCoverageInRange(cmpH5, coords, rowNumbers=None):
"""
Return a vector of length: coords[2] - coords[1] where each
element represents the number of reads overlapping that position
in the cmp.h5 file.
"""
if not cmpH5.isSorted: raise Exception("CmpH5 is not sorted")
if rowNumbers==None:
rowNumbers = getReadsInRange(cmpH5, coords, justIndices=True)
if (len(rowNumbers))==0:
return np.array([0]*(coords[2] - coords[1]))
else:
return(projectIntoRange(cmpH5.tStart[rowNumbers], cmpH5.tEnd[rowNumbers], coords[1], coords[2]))
|
