Please enable JavaScript.
Coggle requires JavaScript to display documents.
More methods geared to address overlapping complexes - Coggle Diagram
More methods geared to address overlapping complexes
Decomposing PPI networks for complex discovery
Solution
divide proteins into subsets annotated by cell localization GO term (where they are in the cell)
remove hub proteins (proteins that have a lot of connections), cluster, bring them back
Differentiation
preprocessing/postprocessing methods
Needs
better prediction of overlapping complexes; capture information about interaction mutual exclusivity
Benefits
super easy concept
existing clustering algorithms can be improved
Regularizing predicted complexes by mutually exclusive protein-protein interactions
Differentiation
expresses protein interaction mutually exclusive (PIME) constraint as regularization term in previous work PPSampler2
Benefits
also improves detection of small complexes of PPSampler algo
idea can be applied in another form
Solution
using a previous clustering (non-overlapping) result, find groups of mutually exclusive interactions in each cluster, for each group, remove all but one, try to find configuration where cluster remains connected
Needs
better prediction of overlapping complexes; capture information about interaction mutual exclusivity
Protein complex prediction by date hub removal
Solution
collect gold-standard list of date hub proteins and study topological features of said proteins in PPI network
two stage method: combine clustering results before and after removing date hub proteins
Differentiation
used empirical data on date hub proteins to establish less arbitrary thresholds for choosing date hub proteins
used with CMC algorithm
Needs
better prediction of overlapping complexes
unclear/faulty characterization of date and hub proteins in literature
Benefits
idea can be applied to supplement existing clustering algorithms
RocSampler: regularizing overlapping protein complexes in protein-protein interaction networks
Solution
ensure partitions are at least of size two
cooling operation from simulated annealing
modify partitioning scheme to allow overlapping partitions and allow overlapping predictions
Differentiation
builds up on previous work on PPSampler2, Metropolis-Hasting (Markov chain Monte Carlo)
Needs
need to capture fact that complexes can overlap (share proteins), many algorithms use a partition model
Benefits
allows overlapping complex predictions
retains advantages of PPSampler2