-
Notifications
You must be signed in to change notification settings - Fork 15
/
info.txt
129 lines (83 loc) · 4.12 KB
/
info.txt
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
================================
Genomic Association Tester (GAT)
================================
The genomic association tester (GAT) tests for association
between intervals in a genome analysis context.
GAT answers the question, if a set of intervals on a genome
is statistically significantly associated with one or more
other sets of genomic intervals. Assocation typically means
overlap, but also other measures like the distance between
elements can be tested.
The tests are performed by simulation and within a genomic context.
The simulation part takes care of segment length distribution.
The genomic context can optionally take into account chromosomal
location (autosomes versus X-linked features) or isochores.
This module has been inspired by the TheAnnotator tool by
Gerton Lunter and the scripts of Caleb Webber.
The differences are:
* permit more measures of association. The original Annotator used nucleotide
overlap, but other measures might be useful like number of elements overlapping
by at least # nucleotides, proximity to closest element, etc.
* permit incremental runs. Annotations can be added without recomputing the samples.
Work space
----------
Inputs:
1. A set of :term:`intervals` ``S`` with segments to test.
2. A set of :term:`intervals` ``A`` with annotations to test against.
3. A set of :term:`intervals` ``W`` describing a work space.
Measures of association:
1. Number of bases overlapping
2. Number of intervals intersecting
3. Distance between segment and annotation
1. Closest distance of segment to annotation
2. Closest distance of annotation to segment
Sampling strategys
==================
Sampling creates a new set of :term:`interval` ``P``. There
are several different strategies possible.
Annotator strategy
------------------
Samples are created in a two step procedure. First, the length
of a sample segment is chosen randomly from the empirical segment
length distribution. Then, a random coordinate is chosen. If the
sampled segment does not overlap with the workspace it is rejected.
Before adding the segment to the sampled set, it is truncated to
the workspace.
If it overlaps with a previously sampled segment, the segments
are merged. Thus, bases shared between two segments are not counted
twice.
Sampling continues, until exactly the same number of bases overlap between
the ``P`` and the ``W`` as do between ``S`` and ``W``.
Note that the length distribution of the ``P`` is different from ``S``.
This method is quick if the workspace is large. If it is small,
a large number of samples will be rejected.
This sampling method is compatible with both distance and overlap
based measures of associaton.
Workspaces and isochores
========================
Workspaces limit the genomic space available for segments and annotations.
Isochores split a workspace into smaller parts that permit to control for
confounding variables like G+C content.
The simplest workspace is the full genome. For some analyses it might be better
to limit to analysis to autosomes.
Examples for the use of isochores could be to analyze chromosomes or chromosomal arms
separately.
If isochores are used, the length distribution and nucleotide overlaps are counted per isochore
ensuring that the same number of nucleotides overlap each isochore in ``P`` and ``S`` and the
length distributions per isochore are comparable.
Empirical length distribution
-----------------------------
The empirical length distribution is created from all :term:`intervals`
in ``S``. The full segment length is chosen even if there is partial overlap.
Optionally, the segment can be truncated. From Gerton::
What is the best choice depends on the data. Not truncating can lead
to a biased length distribution if it is expected that segments that
only partially overlap the workspace have very different lengths. However,
truncating can lead to spurious short segments.
Implementation
==============
Sampler - returns a sample of randomly placed segments
Quantifiers
.. glossary::
Interval
a (genomic) segment with a chromosomal location (contig,start and end).