How many S. cerevisiae physical interactions in BioGRID are self-self interactions?

This blog post shows an example of an easy quire analysis to extract self-self physical interactions identified in the budding yeast Saccharomyces cerevisiae from (an old) BioGRID database. The point of these blog posts is to show some of the motivation for the development of quire, not to explain the details of the language. For details on the quire language, see the documentation (>100 pages) in the provided .tgz file.

Doing the counting

First, we need to read in Biogrid and extract out the physical interactions. We'll do this with a separate file, biogrid-physical-nosymm.q. We can then run this script with the run( @"biogrid-physical-nosymm.q" ) command in the actual script. This is one of the few code reuse strategies that quire currently supports.

##
## biogrid-physical-nosymm.q
##
#
# Read biogrid
#

scgenes     = registry( @ "../../quire_git/data/databases/gene_registry.txt" )
biogridfile = "/home/cdputnam/databases/biogrid/BIOGRID-ORGANISM-Saccharomyces_cerevisiae_S288c-3.4.152.tab2.txt"
biogrid     = network( @ biogridfile, format="biogrid_tab2", generegistry=scgenes, symmetry="none", sourcename="biogrid" )

#
# Extract physical interactions
#

physical_types = """
Affinity Capture-MS
Affinity Capture-Western
Reconstituted Complex
Two-hybrid
Co-crystal Structure
Far Western
FRET
Protein-peptide
PCA
Co-purification
Co-fractionation
Affinity Capture-Luminescence
"""

physical       = filter( biogrid, type=physical_types, silent="false" )

Next, we need to identify which of the baits and prey genes are present in a self-self interaction. Because we read in the interaction data without symmetry, we can simplify the task by recognizing that the only possible candidates must be in both the bait and prey genelists. For the Saccharomyces cerevisiae Biogrid database 3.4.152, there are 5,626 genes in the bait and prey list, but only 4,287 that are in both, so we can ignore 1,339 of them.

run( @"biogrid-physical-nosymm.q" )

baits  = element( physical, type="baits" )
preys  = element( physical, type="preys" )

common = filter( baits + preys, partition="11" ) -> flatten()

Finally, we can look for any interactions between a gene with itself by applying the gene partition of a genelist containing only one gene at a time and extracting those interactions with the gene partition "1:1".

Instead of just counting, we will collect all of the genes and observations belonging to self-self interactions for potential future use.

selfinteractors  = genelist( generegistry=scgenes )
selfinteractions = network( generegistry=scgenes )

for i in range( sizeof( candidates ) )
        gene = element( candidates, i )
        label( physical, genepartition=gene )
        genegene = filter( physical, genepartition="1:1" )
        if ( sizeof( genegene ) > 0 ) then
                selfinteractors  = flatten( selfinteractors + gene )
                selfinteractions = flatten( selfinteractions + genegene )
        fi
        orf = element( gene, 1, type="genes" )
        "gene: " + locus( orf, scgenes ) + " (" + orf + ")" + " self: " + sizeof( selfinteractions )
rof

selfinteractors

Thus we end up with 1,583 genes identified as being part of self-self interactions (observed 2,912 times for an average of 1.8 observations for each interaction).

network summary
----------------
        nsources        1
                source 1 flattened
        nbaits  1583
        npreys  1583
        ninteraction_types      11
                Affinity Capture-MS             587
                PCA                             383
                Two-hybrid                      697
                Affinity Capture-Western        335
                Reconstituted Complex           458
                Co-crystal Structure            268
                Co-purification                  66
                FRET                             41
                Co-fractionation                 34
                Protein-peptide                  40
                Far Western                       3
        ninteractions       1583
        nobservations       2912

Are reports of self-self interactions missing in some physical techniques?

We can now take the self-self interaction network and compare it to the starting network.

Technique                      Total Obs  Self Obs   Percent
-----------------------------  ---------  ---------  -------
All Techniques                    119539     2912     2.44%
-----------------------------  ---------  ---------  -------
Affinity Capture-MS                63529      587     0.92%
Affinity Capture-Western           17536      335     1.91%
Reconstituted Complex               8080      458     5.67%
PCA                                 6671      383     5.74%
Two-hybrid                         16183      697     4.31%
Co-purification                     4371       66     1.51%
Co-fractionation                    1011       34     3.36%
Protein-peptide                      849       40     4.71%
FRET                                 226       41    18.14%
Far Western                          103        3     2.91%
Co-crystal Structure                 930      268    28.82%
Affinity Capture-Luminescence         50        0     0.00%

Unsurprisingly, co-crystallization has a high percentage of self-self physical interactions. Also unsurprisingly, affinity capture mass spectrometry has very few reports of self-self physical interactions, given the fact that peptides from multiple subunits of the same protein generate the same peptide, so that identifying self-self interactions requires very specific experimental designs.