Biology Reference
In-Depth Information
the primary screen are re-tested independently by Y2H
using fresh yeast cells (verification). In our most recent
Y2H-based binary interactome mapping efforts, DB-X/
AD-Y pairs that score positive in at least three out of four
replicates during the verification step were considered
high-quality verified Y2H interactions
[55,57]
. Further
sequencing can be undertaken to confirm the identity of the
reported protein pairs
[55]
. Ultimately, this pipeline can
systematically generate large numbers of highly repro-
ducible binary protein interactions.
Reproducibility, however, does not necessarily guarantee
high dataset quality. Technical artifacts could result in protein
pairs appearing reproducibly positive despite actually being
false positives. To obtain a quantitative estimate of the quality
of a protein
be capable of doing so
[84,85]
. Individual protein interac-
tion assays seem optimized for the detection of a certain
subtype of binary interactions, although the biochemical
and structural biases of these assays remain poorly under-
stood. Intuitively, interactions involving membrane
proteins would be expected to perform better in the cell
membrane environment of the split-ubiquitin system than
in the nuclear environment of the Y2H system
[65,47]
.It
will therefore be necessary to join forces and use multiple
assays to fully map the reference binary interactome of any
organism. We are confident that
this challenge can be
overcome within the next decade.
protein interactome dataset, an integrated
empirical framework for quality control of interactome
mapping was proposed in 2009
[36]
. At the heart of this
endeavor was the recognition that high-throughput inter-
actome mapping needed to be rigorously calibrated, like any
well-controlled reliable small-scale experiment. Using this
framework, protein
e
Large-Scale Co-Complex Interactome
Mapping
To fully map the reference interactome, it is operationally
helpfultogobeyondbinaryprotein interactions and
identify protein complexes within cells
[86]
.Protein
complexes typically contain five to six different proteins,
within a wide range from two to hundreds in a variety of
stoichiometries
[87]
. The concentration and binding
affinity of the protein subunits determine complex
assembly according to the law of mass action
[88]
.Two
proteins in isolation may have only weak or no propen-
sity to form a binary interaction. Owing to cooperative
and allosteric effects, a third protein may have a high
affinity to both simultaneously, so that the resulting
protein complex is considerably more stable than the sum
of its component affinities
[89]
. Hence, even if a refer-
ence binary protein
protein interactome maps generated
with the mapping pipeline described herewere shown to have
high precision (80
e
100% of reported protein pairs are true
positives) but low sensitivity (~10
e
15% of all interrogated
true positive interactions are captured in the experiment)
[36,39,52,55]
. Because of this low sensitivity, binary inter-
actome maps generated so far represent small fractions of the
underlying true interactomes. This explains why only
a marginal number of protein interactions are found in
multiple binary interactome maps assembled independently
for the same organism
[36,39,55]
.
How far along is the journey towards a complete binary
protein
e
protein interactome had been fully
mapped, co-complex interactome network maps would
still provide novel protein interactions and bring
a fundamentally different view of interactome organiza-
tion. The characterization of entire protein complexes, as
they assemble in cells, is a necessary route to gather
information on gene function and biological systems
[90
e
protein interactome reference map? The answer
requires an estimation of the size of such a reference map for
any given species. Many statistical methods have been
designed to this end, often based on dataset overlap and
hypergeometric distributions
[36,75
e
82]
. Mapping of the
binary interactome of the model organism S. cerevisiae is
estimated to be the closest to completion, with ~6
e
30%
coverage already obtained (~2900 binary protein interactions
of demonstrated high technical quality detected, out of an
estimated total of 10 000
e
94]
.
The most common methodologies currently used for the
mapping of co-complex interactomes rely on protein
complex purification followed by identification of constit-
uent proteins by mass spectrometry. These experiments
necessitate a trade-off between throughput, reproducibility
and physiological setting. Cellular proteins range in their
abundance up to seven orders of magnitude in humans
[95]
and five in yeast
[96]
. Protein complexes therefore need to
be purified from the soup of cellular extracts without losing
too many components, while at the same time avoiding
those proteins that are extremely abundant and co-purify
artificially
[97
e
45 000, assuming one splicing
isoform per gene)
[39,46,49]
. However, most of the task
remains to be accomplished.
How can this daunting challenge be overcome? Inspi-
ration is drawn from the history of genome sequencing,
which underwent a disruptive shift in the late 1990s. Like
sequencing at that time, Y2H-based mapping is currently
seeing more efficient automation, stricter quality control
and innovative technology development which together are
increasing productivity while reducing cost
[36,62,83]
.Itis
unlikely that Y2H-based mapping alone will be sufficient to
complete a comprehensive reference binary protein
e
100]
. A fraction of protein complexes may
consist of dedicated elements, but most complexes also
include abundant proteins that participate in several other
complexes,
e
pro-
tein interactome map. No single interaction assay may ever
e
such as chaperone proteins. Purification
