Debiasing a Protein Annotation Database
Debiaser removes bias from GAF files based on annotation information content, GO evidence, annotation source, number of proteins annotated from a given source, an date. Debiaser accepts one or more GAF files as input. The motivation for GAF lies in the observation that many organism annotations are biased due to high throughpout experimental studies (1). Removing such annotation biases can help present a more balanaced picture of protein annotations for a given organism or set of proteins.
Prerequisites
Required modules.
Modules are available in most GNU/Linux distributions, or from their respective websites.
Required files
You would need an obo formatted version of the Gene Ontology. Depending on your needs, this would usually be one of go-basic.obo or go.obo. For more details and to download either the most recent daily version or the latest version go to the Gene Ontology website.
Installation
Installing from source
git clone https://github.com/Rinoahu/debias
cd debias
python setup.py install
Installing with pip
pip install debias
OR
pip install git+git://github.com/Rinoahu/debias
Initial files
These files will be created upon running debias_prep.
debias_prep -i data/GOFILE.obo
GOFILE will usually be one of go.obo or go-basic.obo
This will generate seven files in total. Three files corresponds to the three ontologies. Three files corresponds to the mapping between each GO_term and its ancestors in its own respective ontology. The last file contains mapping from alternate GO_ID to actual GO_ID. Please use this command when a new go.obo file is released.
1. ./data/alt_to_id.graph : Needed to obtain mapping from alternate GO_ID to actual GO_ID
2. ./data/mf.graph : The MFO Ontology graph
3. ./data/bp.graph : The BPO Ontology graph
4. ./data/cc.graph : The CCO Ontology graph
5. ./data/mf_ancestors.map : The MFO Ancestors map
6. ./data/bp_ancestors.map : The BPO Ancestors map
7. ./data/cc_ancestors.map : The CCO Ancestors map
Quick setup steps
-
Download the latest go.obo file from http://www.geneontology.org/ontology/
-
Run the program
debias_prepprogram and provide the downloaded .obo file. See the usage details below. This program needs to be run only when a new .obo file needs to be used. -
Run the program
debias
usage: debias [-h] [--prefix PREFIX] [--cutoff_prot CUTOFF_PROT]
[--cutoff_attn CUTOFF_ATTN] [--output OUTPUT]
[--evidence EVIDENCE [EVIDENCE ...] | --evidence_inverse
EVIDENCE_INVERSE [EVIDENCE_INVERSE ...]] --input INPUT
[INPUT ...] [--aspect ASPECT [ASPECT ...]]
[--assigned_by ASSIGNED_BY [ASSIGNED_BY ...] |
--assigned_by_inverse ASSIGNED_BY_INVERSE
[ASSIGNED_BY_INVERSE ...]] [--recalculate RECALCULATE]
[--info_threshold_Wyatt_Clark_percentile INFO_THRESHOLD_WYATT_CLARK_PERCENTILE | --info_threshold_Wyatt_Clark INFO_THRESHOLD_WYATT_CLARK]
[--info_threshold_Phillip_Lord_percentile INFO_THRESHOLD_PHILLIP_LORD_PERCENTILE | --info_threshold_Phillip_Lord INFO_THRESHOLD_PHILLIP_LORD]
[--verbose VERBOSE] [--date_before DATE_BEFORE]
[--date_after DATE_AFTER] [--single_file SINGLE_FILE]
[--select_references SELECT_REFERENCES [SELECT_REFERENCES ...]
| --select_references_inverse SELECT_REFERENCES_INVERSE
[SELECT_REFERENCES_INVERSE ...]] [--report REPORT]
[-histogram HISTOGRAM]
optional arguments:
-h, --help show this help message and exit
--prefix PREFIX, -pref PREFIX
Add a prefix to the name of your output files.
--cutoff_prot CUTOFF_PROT, -cprot CUTOFF_PROT
The threshold level for deciding to eliminate
annotations which come from references that annotate
more than the given 'threshold' number of PROTEINS
--cutoff_attn CUTOFF_ATTN, -cattn CUTOFF_ATTN
The threshold level for deciding to eliminate
annotations which come from references that annotate
more than the given 'threshold' number of ANNOTATIONS
--output OUTPUT, -odir OUTPUT
Writes the final outputs to the directory in this
path.
--evidence EVIDENCE [EVIDENCE ...], -e EVIDENCE [EVIDENCE ...]
Accepts Standard Evidence Codes outlined in
http://geneontology.org/page/guide-go-evidence-codes.
All 3 letter code for each standard evidence is
acceptable. In addition to that EXPEC is accepted
which will pull out all annotations which are made
experimentally. COMPEC will extract all annotations
which have been done computationally. Similarly,
AUTHEC and CUREC are also accepted. Cannot be provided
if -einv is provided
--evidence_inverse EVIDENCE_INVERSE [EVIDENCE_INVERSE ...], -einv EVIDENCE_INVERSE [EVIDENCE_INVERSE ...]
Leaves out the provided Evidence Codes. Cannot be
provided if -e is provided
--aspect ASPECT [ASPECT ...], -a ASPECT [ASPECT ...]
Enter P, C or F for Biological Process, Cellular
Component or Molecular Function respectively
--assigned_by ASSIGNED_BY [ASSIGNED_BY ...], -assgn ASSIGNED_BY [ASSIGNED_BY ...]
Choose only those annotations which have been
annotated by the provided list of databases. Cannot be
provided if -assgninv is provided
--assigned_by_inverse ASSIGNED_BY_INVERSE [ASSIGNED_BY_INVERSE ...], -assgninv ASSIGNED_BY_INVERSE [ASSIGNED_BY_INVERSE ...]
Choose only those annotations which have NOT been
annotated by the provided list of databases. Cannot be
provided if -assgn is provided
--recalculate RECALCULATE, -recal RECALCULATE
Set this to 1 if you wish to enforce the recalculation
of the Information Accretion for every GO term.
Calculation of the information accretion is time
consuming. Therefore keep it to zero if you are
performing rerun on old data. The program will then
read the information accretion values from a file
which it wrote to in the previous run of the program
--info_threshold_Wyatt_Clark_percentile INFO_THRESHOLD_WYATT_CLARK_PERCENTILE, -WCTHRESHp INFO_THRESHOLD_WYATT_CLARK_PERCENTILE
Provide the percentile p. All annotations having
information content below p will be discarded
--info_threshold_Wyatt_Clark INFO_THRESHOLD_WYATT_CLARK, -WCTHRESH INFO_THRESHOLD_WYATT_CLARK
Provide a threshold value t. All annotations having
information content below t will be discarded
--info_threshold_Phillip_Lord_percentile INFO_THRESHOLD_PHILLIP_LORD_PERCENTILE, -PLTHRESHp INFO_THRESHOLD_PHILLIP_LORD_PERCENTILE
Provide the percentile p. All annotations having
information content below p will be discarded. So if 5 is provided, proteins annotated by
terms whose score is in the top 5% will be left in, the rest will be discarded.
--info_threshold_Phillip_Lord INFO_THRESHOLD_PHILLIP_LORD, -PLTHRESH INFO_THRESHOLD_PHILLIP_LORD
Provide a value t. All annotations having
information content below t will be discarded
--verbose VERBOSE, -v VERBOSE
Set this argument to 1 if you wish to view the outcome
of each operation on the console
--date_before DATE_BEFORE, -dbfr DATE_BEFORE
The date entered here will be parsed by the parser
from dateutil package. For more information on
acceptable date formats please visit
https://github.com/dateutil/dateutil/. All annotations
made prior to this date will be picked up
--date_after DATE_AFTER, -daftr DATE_AFTER
The date entered here will be parsed by the parser
from dateutil package. For more information on
acceptable date formats please visit
https://github.com/dateutil/dateutil/. All annotations
made after this date will be picked up
--single_file SINGLE_FILE, -single SINGLE_FILE
Set to 1 in order to output the results of each
individual species in a single file.
--select_references SELECT_REFERENCES [SELECT_REFERENCES ...], -selref SELECT_REFERENCES [SELECT_REFERENCES ...]
Provide the paths to files which contain references
you wish to select. It is possible to include
references in case you wish to select annotations made
by a few references. This will prompt the program to
interpret string which have the keywords
'GO_REF','PMID' and 'Reactome' as a GO reference.
Strings which do not contain that keyword will be
interpreted as a file path which the program will
except to contain a list of GO references. The program
will accept a mixture of GO_REF and file names. It is
also possible to choose all references of a particular
category and a handful of references from another. For
example if you wish to choose all PMID references,
just put PMID. The program will then select all PMID
references. Currently the program can accept PMID,
GO_REF and Reactome
--select_references_inverse SELECT_REFERENCES_INVERSE [SELECT_REFERENCES_INVERSE ...], -selrefinv SELECT_REFERENCES_INVERSE [SELECT_REFERENCES_INVERSE ...]
Works like -selref but does not select the references
which have been provided as input
--report REPORT, -r REPORT
Provide the path where the report file will be stored.
If you are providing a path please make sure your path
ends with a '/'. Otherwise the program will assume the
last string after the final '/' as the name of the
report file. A single report file will be generated.
Information for each species will be put into
individual worksheets.
--histogram HISTOGRAM, -hist HISTOGRAM
Set this option to 1 if you wish to view the histogram
of GO_TERM frequency before and after debiasing is
performed with respect to cutoffs based on number of
proteins or annotations. If you wish to save the file
then please enter a filepath. If you are providing a
path please make sure your path ends with a '/'.
Otherwise the program will assume the last string
after the final '/' as the name of the image file.
Separate histograms will be generated for each
species.
Required arguments:
--input INPUT [INPUT ...], -i INPUT [INPUT ...]
The input file path. Please remember the name of the
file must start with goa in front of it, with the name
of the species following separated by an underscore
NOTE: The files inside the folder "temp" are the one which have been generated by executing the command below
Examples
debias_prep -i data/go.obo
This command will generate seven files in total. Three files corresponds to the three ontologies. Three files corresponds to the mapping between each GO_term and its ancestors in its own respective ontology. The last file contains mapping from alternate GO_ID to actual GO_ID. Please use this command every time you update GOFILE.
debias -cprot 100 -i data/goa_yeast.gaf data/goa_dicty.gaf -a C -WCTHRESHp 2 -recal 1
This command reads from two input files one for yeast and the other for dicty. The -a C only selects the annotations which are CCO. The -WCTHRESHp argument specifies that the Wyatt Clark Threshold is a 2 percentile, which means all annotations having a Wyatt Clark Information content below 2% will be removed. Instead of providing a percentage value one can also provide a threshold value using the argument -WCTHRESH. In addition to that, those annotations will be removed which have been annotated by references that have in turn annotated more than 100 proteins. The output will be put in the current directory. It is necessary to have -recal 1 in this command since the GO_term to IC has to be created. Subsequent runs with different threshold and all other parameters fised is possible WITHOUT providing the argument -recal. This command will lead to 3 output files. One each for the two organisms and the third one is where both the organisms are combined.
debias -i data/goa_yeast.gaf data/goa_dicty.gaf -a C P -PLTHRESHp 30 -e EXPEC IBA -odir data/output -single 1
This command will read from two input files, select CCO and BPO annotations. Further, it will choose only those annotations which have been made experimentally or have been annotated computationally as "IBA" (Inferred from Biological aspect of Ancestor). In addition to that it will discard all annotations which have a Phillip Lord information content less than 30%. Instead of providing a percentage value one can also provide a threshold value using the argument -PLTHRESH. The final output will be put inside the data/output directory. You can include non existent paths. The program will attempt to create the folders if required permissions are present. This will lead to only one file, since the -single argument has been provided, which will contain all the selected annotations from both the organisms.
debias -cattn 1000 -i data/goa_yeast.gaf data/goa_dicty.gaf -a C P -einv COMPEC -pref testing -selrefinv Reactome
This command will read from two input files, select CCO and BPO annotations. Further, it will discard those annotations which have been made computationally. The program further filters out all annotations made by "Reactome". All files will be prefixed with the string "testing". Since the program creates a meaningful name for each file, the user has been given the opportunity to give a prefix.
Running test data
To test all the commands mentioned above, you can run the shell script named test.sh in the tests directory.
git clone https://github.com/Rinoahu/debias
cd ./debias/tests
bash test.sh
