You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Copy file name to clipboardExpand all lines: docs/beluga.rst
+102-1Lines changed: 102 additions & 1 deletion
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -1,4 +1,3 @@
1
-
.. _Beluga:
2
1
=======
3
2
Beluga
4
3
=======
@@ -13,3 +12,105 @@ The 2019 version of DeepSEA, nicknamed '**Beluga**', can predict **2002** chroma
13
12
Jian Zhou, Chandra L. Theesfeld, Kevin Yao, Kathleen M. Chen, Aaron K. Wong, and Olga G. Troyanskaya, **Deep learning sequence-based ab initio prediction of variant effects on expression and disease risk**. Nature Genetics (2018).
14
13
15
14
To determine if certain features (ie. transcription factors, marks, or cell types) are present/accounted for in the model, refer to the `supplemental feature table <https://s3-us-west-2.amazonaws.com/humanbase-dev/deepsea/examples/41588_2019_420_MOESM9_ESM.csv>`_ which has all the profiles used to train DeepSEA.
15
+
16
+
DeepSEA is originally described in the following manuscript:
17
+
18
+
Jian Zhou, Olga G. Troyanskaya. **Predicting the Effects of Noncoding Variants with Deep learning-based Sequence Model.** Nature Methods (2015).
19
+
20
+
To determine if certain features (ie. transcription factors, marks, or cell types) are present/accounted for in the model, refer to the `supplemental feature table <https://s3-us-west-2.amazonaws.com/humanbase-dev/deepsea/examples/41588_2019_420_MOESM9_ESM.csv>`_ which has all the profiles used to train DeepSEA.
21
+
22
+
23
+
Input
24
+
-----
25
+
26
+
DeepSEA predicts genomic variant effects on a wide range of chromatin features at the variant position (Transcription factors binding, DNase I hypersensitive sites, and histone marks in multiple human cell types). DeepSEA can also be utilized for predicting chromatin features for any DNA sequence.
27
+
28
+
File formats
29
+
~~~~~~~~~~~~
30
+
We support three types of input: vcf, fasta, bed. If you want to predict effects of noncoding variants, use vcf format input. If you want to predict chromatin feature probabilities for DNA sequences, use fasta format. If you want to specify sequences from the human reference genome (GRCh37/hg19), you can use bed format. See below for a quick introduction:
31
+
32
+
**VCF format** is used for specifying a genomic variant. A minimal example is ``chr1 109817590 - G T`` (if you want to copy cover this text as input, you will need to change spaces to tabs). The five columns are chromosome, position, name, reference allele, and alternative allele.
33
+
34
+
**Fasta format** input should include sequences of 2000bp length each. If a sequence is longer than 2000bp, only the center 2000bp will be used. A minimal example is ::
**Bed format** provides another way to specify sequences in human reference genome (hg19). The bed input should specify 2000bp-length regions. A minimal example is ``chr1 109817091 109819090``. The three columns are chromosome, start position, and end position.
79
+
80
+
Genome coordinates
81
+
~~~~~~~~~~~~~~~~~~
82
+
We support only ``GRCh37/hg19`` genome coordinates. You can use LiftOver to convert your coordinates to the correct version.
83
+
84
+
Large submissions
85
+
~~~~~~~~~~~~~~~~~
86
+
We recommend using the web server if you have <10,000 variants or sequences. You will experience degraded performance when submitting a larger set of sequences. In those instances, we suggest that you split the set into multiple <10,000 submissions, or run the standalone version on your local machine, or contact our group directly.
87
+
88
+
Output
89
+
------
90
+
91
+
Regulatory feature scores
92
+
~~~~~~~~~~~~~~~~~~~~~~~~~
93
+
* **diffs**: The difference between the the predicted probability of the reference allele and the alternative allele for a regulatory feature (:math:`p_{alt} -p_{ref}`).
94
+
* **e-value**: E-value is defined as the expected proportion of SNPs with a larger predicted effect. We calculate an 'e-value' based on the empirical distribution of that feature's effect (:math:`abs(p_{alt} -p_{ref})`) among gnomAD variants. For example, a feature e-value of 0.01 indicates that only 1% of gnomAD variants have a larger predicted effect.
95
+
* **z-score**: A scaled score where the feature diff score (:math:`p_{alt} -p_{ref}`) is divided by the root mean square of the feature diff score across gnomAD variants. Note that this is "sign-preserving", i.e. a negative z-score indicates that a mutation **decreases** the probability of a regulatory feature.
96
+
97
+
Variant scores
98
+
~~~~~~~~~~~~~~
99
+
100
+
* **Disease Impact Score (DIS)**: DIS is calculated by training a logistic regression model that prioritizes likely disease-associated mutations on the basis of the predicted transcriptional or post-transcriptional regulatory effects of these mutations (See Zhou et. al, 2019). The predicted DIS probabilities are then converted into 'DIS e-values', computed based on the empirical distributions of predicted effects for gnomAD variants. The final DIS score is:
101
+
102
+
.. math::
103
+
-log10(DIS evalue_{feature})
104
+
105
+
* **Mean -log e-value (MLE)**: For each predicted regulatory feature effect (:math:`abs(p_{alt}-p_{ref}`)) of a variant, we calculate a 'feature e-value' based on the empirical distribution of that feature's effects among gnomAD variants (see above Regulatory feature scores: e-value). The MLE score of a variant is
106
+
107
+
.. math::
108
+
\sum -log10(evalue_{feature}) / N
109
+
110
+
In-silico mutagenesis
111
+
---------------------
112
+
Perform "In silico saturated mutagenesis" (ISM) analysis to discover informative sequence features within any sequence. Specifically, it performs computational mutation scanning to assess the effect of mutating every base of the input sequence on chromatin feature predictions. This method for context-specific sequence feature extraction takes advantage of DeepSEA’s ability to utilize flanking context sequences information.
113
+
114
+
Note that ISM only accepts a sequence (FASTA file) as input.
115
+
116
+
ISM outputs effects for each of three possible substitutions of all 2000 bases, across all chromatin features.
Copy file name to clipboardExpand all lines: docs/deepsea.rst
+52-45Lines changed: 52 additions & 45 deletions
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -13,37 +13,20 @@ For older DeepSEA models see:
13
13
:doc:`beluga` (2019)
14
14
15
15
16
-
17
16
Input
18
17
-----
19
18
20
-
DeepSEA predicts genomic variant effects on a wide range of chromatin features at the variant position (Transcription factors binding, DNase I hypersensitive sites, and histone marks in multiple human cell types). DeepSEA can also be ultilized for predicting chromatin features for any DNA sequence.
19
+
DeepSEA/Sei predicts genomic variant effects on a wide range of chromatin features at the variant position (Transcription factors binding, DNase I hypersensitive sites, and histone marks in multiple human cell types). DeepSEA can also be utilized for predicting chromatin features for any DNA sequence.
21
20
22
21
File formats
23
22
~~~~~~~~~~~~
24
23
We support three types of input: vcf, fasta, bed. If you want to predict effects of noncoding variants, use vcf format input. If you want to predict chromatin feature probabilities for DNA sequences, use fasta format. If you want to specify sequences from the human reference genome (GRCh37/hg19), you can use bed format. See below for a quick introduction:
25
24
26
-
**VCF format** is used for specifying a genomic variant. A minimal example is ``chr1 109817590 - G T`` (if you want to copy cover this text as input, you will need to change spaces to tabs). The five columns are chromosome, position, name, reference allele, and alternative allele.
27
-
28
-
**Fasta format** input should include sequences of 2000bp length each. If a sequence is longer than 2000bp, only the center 2000bp will be used. A minimal example is ::
**VCF format** is used for specifying a genomic variant. A minimal example is ``chr1 109817590 - G T`` (if you want to copy cover this text as input, you will need to change spaces to tabs). The five columns are chromosome, position, name, reference allele, and alternative allele. Currently, the genome position needs to be in GRCh37/hg19
44
26
27
+
**Fasta format** input should include sequences of 4096bp length each. If a sequence is longer than 4096bp, only the center 4096bp will be used.
45
28
46
-
**Bed format** provides another way to specify sequences in human reference genome (hg19). The bed input should specify 2000bp-length regions. A minimal example is ``chr1 109817091 109819090``. The three columns are chromosome, start position, and end position.
29
+
**Bed format** provides another way to specify sequences in human reference genome (hg19). The bed input should specify 4096bp-length regions. A minimal example is ``chr1 109817091 109821186``. The three columns are chromosome, start position, and end position.
47
30
48
31
Genome coordinates
49
32
~~~~~~~~~~~~~~~~~~
@@ -57,29 +40,53 @@ We recommend using the web server if you have <10,000 variants or sequences. You
57
40
Output
58
41
------
59
42
60
-
Regulatory feature scores
43
+
Sequence classes
61
44
~~~~~~~~~~~~~~~~~~~~~~~~~
62
-
* **diffs**: The difference between the the predicted probability of the reference allele and the alternative allele for a regulatory feature (:math:`p_{alt} -p_{ref}`).
63
-
* **e-value**: E-value is defined as the expected proportion of SNPs with a larger predicted effect. We calculate an 'e-value' based on the empirical distribution of that feature's effect (:math:`abs(p_{alt} -p_{ref})`) among gnomAD variants. For example, a feature e-value of 0.01 indicates that only 1% of gnomAD variants have a larger predicted effect.
64
-
* **z-score**: A scaled score where the feature diff score (:math:`p_{alt} -p_{ref}`) is divided by the root mean square of the feature diff score across gnomAD variants. Note that this is "sign-preserving", i.e. a negative z-score indicates that a mutation **decreases** the probability of a regulatory feature.
65
-
66
-
Variant scores
67
-
~~~~~~~~~~~~~~
68
-
69
-
* **Disease Impact Score (DIS)**: DIS is calculated by training a logistic regression model that prioritizes likely disease-associated mutations on the basis of the predicted transcriptional or post-transcriptional regulatory effects of these mutations (See Zhou et. al, 2019). The predicted DIS probabilities are then converted into 'DIS e-values', computed based on the empirical distributions of predicted effects for gnomAD variants. The final DIS score is:
70
-
71
-
.. math::
72
-
-log10(DIS evalue_{feature})
73
-
74
-
* **Mean -log e-value (MLE)**: For each predicted regulatory feature effect (:math:`abs(p_{alt}-p_{ref}`)) of a variant, we calculate a 'feature e-value' based on the empirical distribution of that feature's effects among gnomAD variants (see above Regulatory feature scores: e-value). The MLE score of a variant is
75
-
76
-
.. math::
77
-
\sum -log10(evalue_{feature}) / N
78
-
79
-
In-silico mutagenesis
80
-
---------------------
81
-
Perform "In silico saturated mutagenesis" (ISM) analysis to discover informative sequence features within any sequence. Specifically, it performs computational mutation scanning to assess the effect of mutating every base of the input sequence on chromatin feature predictions. This method for context-specific sequence feature extraction takes advantage of DeepSEA’s ability to utilize flanking context sequences information.
82
-
83
-
Note that ISM only accepts a sequence (FASTA file) as input.
84
45
85
-
ISM outputs effects for each of three possible substitutions of all 2000 bases, across all chromatin features.
46
+
The Sei framework predicts 40 sequence class scores, covering a wide range of regulatory activities such as cell-type-specific enhancers and promoters, as well as 21,907 chromatin profiles for any DNA sequence.
47
+
48
+
To help interpretation, we grouped sequence classes into groups including P (Promoter), E (Enhancer), CTCF (CTCF-cohesin binding), TF (TF binding), PC (Polycomb-repressed), HET (Heterochromatin), TN (Transcription), and L (Low Signal) sequence classes. Please refer to our manuscript for a more detailed description of the sequence classes.
49
+
50
+
51
+
|Sequence class label | Sequence class name | Rank by size | Group |
0 commit comments