identification of communities from multiplex networks
TRANSCRIPT
Systems Biology
Genotypes
Phenotypes
-OmicsNetworks
Large-scale Protein-protein Interaction Networks
Interactomes
How to extract relevant biological information ?
Identification of communitiesClasses / Clusters / Modules
Protein cellular functions
Group of tightly linked genes/proteins
6
Inspired from P. Aloy, ECCB 2014
Protein - Structures Functional modules Interaction Networks
ComprehensivenessPrecision
• Topology : Density, Modularity • Integration of external data • Overlapping / Disjoint
Multiple networks to depict protein cellular functions
PPI (12110 / 60669)Spike Biocarta Kegg Nci
Co-Expression (9912 / 1107547)
Pathway-related (8839 / 166761)
CORUM complexes (2528 / 36762)
• Does the combination of networks improves the detection of communities?
• How to combine multiple networks to identify communities?
Network aggregation
Union, Sum
Intersection
Consensus partitioning
Multiplex framework
Multiplex, Multi-layer, Multislice
Multiplex framework
• Multiplex-modularity
Protocole
• Multiplex framework
• Networks aggregation
• Consensus partition
• Modularity
• Modularity Optimization (Louvain algo)
• Multiplex-modularity
• Mx-modularity Optimization (adapted Louvain algo)
Simulations (SBMs)
Considering many networks improves the detection of communities
Simulations (SBMs)
Mx framework more efficient with heterogeneous and incomplete networks
Multiplex
SumUnionIntersection
Application to real biological networks
Pathways Co-expr PPI Complexes
• Granularity parameter from 1 to 15 • Community number, sizes, similarities
=> More balanced
Union aggregationSum aggregation
Multiplex-modularity
Resolution parameter γ
Annotated
clusters(%
)
151413121110987654321
0.5
0.45
0.4
0.35
0.3
0.25
0.2
0.15
0.1
0.05
Community GOBP annotations
=> More annotated
=> Multiplex framework is well suited to identify communities from multiple biological networks
Module enriched in Coffin-Siris syndrome genes
PPIPathwaysCo-
expression Complexes
SMARCC1
MLLT1
ACTG1
SMARCD1
SMARCD2
DPF3
ARID1B
MCPH1
SMARCA2
DPF2
CDX2
ZNF14
PCSK4
SMARCD3
SMARCA4
SCYL1
BAZ1B
CARM1
AFF1
PBRM1
PRMT5
ARID2
DRAP1
GATA1
ZNF2
KLF1
ACTL6B
SMARCB1
SLC10A1
UNKL
SMARCC2
ARID1A
SMARCE1
C11orf74
SMARCC1
MLLT1
ACTG1
SMARCD1
SMARCD2
DPF3
ARID1B
MCPH1
SMARCA2
DPF2
CDX2
ZNF14
PCSK4
SMARCD3
SMARCA4
SCYL1
BAZ1B
CARM1
AFF1
PBRM1
PRMT5
ARID2
DRAP1
GATA1
ZNF2
KLF1
ACTL6B
SMARCB1
SLC10A1
UNKL
SMARCC2
ARID1A
SMARCE1
C11orf74
SMARCC1
MLLT1
ACTG1
SMARCD1
SMARCD2
DPF3
ARID1B
MCPH1
SMARCA2
DPF2
CDX2
ZNF14
PCSK4
SMARCD3
SMARCA4
SCYL1
BAZ1B
CARM1
AFF1
PBRM1
PRMT5
ARID2
DRAP1
GATA1
ZNF2
KLF1
ACTL6B
SMARCB1
SLC10A1
UNKL
SMARCC2
ARID1A
SMARCE1
C11orf74
SMARCC1
MLLT1
ACTG1
SMARCD1
SMARCD2
DPF3
ARID1B
MCPH1
SMARCA2
DPF2
CDX2
ZNF14
PCSK4
SMARCD3
SMARCA4
SCYL1
BAZ1B
CARM1
AFF1
PBRM1
PRMT5
ARID2
DRAP1
GATA1
ZNF2
KLF1
ACTL6B
SMARCB1
SLC10A1
UNKL
SMARCC2
ARID1A
SMARCE1
C11orf74
Molti
https://github.com/gilles-didier/MolTiDidier et al. Accepted in PeerJ
Acknowledgments
Elisabeth Remy Laurent Tichit
Alain Guénoche Gilles Didier
José-Luis Portero
PEPS