network-based target ranking for polypharmacological therapies
DESCRIPTION
2013 Summit on Translational BioinformaticsTRANSCRIPT
Francesca Vitali, Francesca Mulas, Pietro Marini and Riccardo Bellazzi
NETWORK-BASED TARGET RANKING FOR POLYPHARMACOLOGICAL THERAPIES
University of Pavia, Italy
Angelo Nuzzo IIT@SEMM, Milan, 2011
Drug Discovery
Angelo Nuzzo IIT@SEMM, Milan, 2011
Drug Discovery Today
Decline in:
• Clinical development
• Productivity
Standard approach MAGIC BULLET
SINGLE TARGET DRUGS
Emerging approach POLYPHARMACOLOGY
Angelo Nuzzo IIT@SEMM, Milan, 2011
Polypharmacology
MULTI TARGET DRUGS
• Collective weak interactions vs single strong interaction
• Improve therapeutic efficacy
- Multi-factorial diseases (e.g. cancer)
• Avoid
- Adaptive drug resistance
Drug
Protein
Angelo Nuzzo IIT@SEMM, Milan, 2011
Network-based Polypharmacology
Prior knowledge on complex diseases
DATA SOURCES +
PROTEIN INTERACTION NETWORKS
Protein
Interaction
Angelo Nuzzo IIT@SEMM, Milan, 2011
Method
Disease of interest
Network design
Network targeting
Scoring of target combinations
Ad hoc analysis of drug agents related to the selected targets
Angelo Nuzzo IIT@SEMM, Milan, 2011
Method
Disease of interest
Network design
Network targeting
Scoring of target combinations
Ad hoc analysis of drug agents related to the selected targets
Type 2 Diabetes Mellitus
(T2DM)
Angelo Nuzzo IIT@SEMM, Milan, 2011
Method
Disease of interest
Network design
Network targeting
Scoring of target combinations
Ad hoc analysis of drug agents related to the selected targets
Angelo Nuzzo IIT@SEMM, Milan, 2011
Network design
NODES Proteins
EDGES Interactions
Human proteins involved T2DM
pathways
Network
Protein – Protein Interaction Repository
Angelo Nuzzo IIT@SEMM, Milan, 2011
Biological Network Features
• Interactions
• Experimental evidence
- Not considering:
Literature mining, co-occurence, ecc.
• Confidence Score › 0.7 (high confidence)
NODES 587
EDGES 3683
Angelo Nuzzo IIT@SEMM, Milan, 2011
Disease Network
Network genes
87 disease proteins
Up- and down-regulated genes
Diabetes vs control
Angelo Nuzzo IIT@SEMM, Milan, 2011
Method
Disease of interest
Network design
Network targeting
Scoring of target combinations
Ad hoc analysis of drug agents related to the selected targets
Angelo Nuzzo IIT@SEMM, Milan, 2011
Network Targeting
• Node constraints:
• Druggable
• Not Hub
• High Bridging Centrality
Angelo Nuzzo IIT@SEMM, Milan, 2011
Druggability
LOCALIZATION
DRUG INTERACTIONS
• Cell membrane
• Extracellular space
DRUGGABLE PROTEINS
206 + 108
Chemical – Protein Interaction Repository
Angelo Nuzzo IIT@SEMM, Milan, 2011
Hub
• Hub
1. High # neighbours
2. Located between highly connected regions
Traditional pharmacology
DISCARDED AS POTENTIAL POLYPHARMACOLOGICAL TARGET
• Hub discovery method:
DISCARDED HUB 14
Vallabhajosyula et al. Identifying hubs in protein interaction networks.PLoS ONE, 4(4):e5344, 2009.
Angelo Nuzzo IIT@SEMM, Milan, 2011
Betweenness centrality
• Global bridging properties
• Influence of a node in the spread of information through the network
Bridging Coefficient
• Local bridging properties
• Penalizes hub
Bridging Centrality
Bridging centrality
BC = Bridging Coefficient
RWB = Random-walk Betweenness
Angelo Nuzzo IIT@SEMM, Milan, 2011
Bridging Centrality
Bridging centrality
BC = Bridging Coefficient
RWB = Random-walk Betweenness
Bridge node first 25% of the nodes ordered by bridging centrality
BRIDGING PROTEINS 147
Hwang et al. Bridging centrality: Identifying bridging nodes in scale-free networks.Knowledge Discovery and Data mining, 2006
Angelo Nuzzo IIT@SEMM, Milan, 2011
Target selection
47 target proteins
“Sources ” of pharmacological actions
Druggable
Non hub
Bridge nodes
Angelo Nuzzo IIT@SEMM, Milan, 2011
Method
Disease of interest
Network design
Network targeting
Scoring of target combinations
Ad hoc analysis of drug agents related to the selected targets
Angelo Nuzzo IIT@SEMM, Milan, 2011
Drug Synergy
• Reachability of a single Disease Protein
Wi,ji j
• Reachability of all DPs
Sh = shortest path
w i,j = edge weight between i and j
(i,j) = node pair
Np = # disease proteins
T
Angelo Nuzzo IIT@SEMM, Milan, 2011
Drug Synergy
TRIPLETS
• Low computational cost
• Upper limit to the number of drugs to be jointly delivered
Angelo Nuzzo IIT@SEMM, Milan, 2011
TSDS
• TSDS (Topological Score of Drug Synergy) of a triplet
TSDS
• Null distribution:
TSDS for 50 000 triplets ofproteins randomly selectedfrom the network
Combinationsof 18 targets
88 triplets with
P-Value < 0.01
Angelo Nuzzo IIT@SEMM, Milan, 2011
Method
Disease of interest
Network design
Network targeting
Scoring of target combinations
Ad hoc analysis of drug agents related to the selected targets
Angelo Nuzzo IIT@SEMM, Milan, 2011
Target AnalysisFrequency
Angelo Nuzzo IIT@SEMM, Milan, 2011
Target Analysis
Insulin-like
growth factor
family
Frequency
Angelo Nuzzo IIT@SEMM, Milan, 2011
Target AnalysisFrequency
• EXPERIMENTAL EVIDENCE
• Negative regulator of insulin signalling TO BE INHIBITED
• Extract of Larrea Tridente
1. Decrease
• plasma glucose
• triglyceride concentrations
2. Inhibits IGF1R activity
Angelo Nuzzo IIT@SEMM, Milan, 2011
Target Analysis
Alzheimer
Frequency
Angelo Nuzzo IIT@SEMM, Milan, 2011
• Strongly altered expression of
IGF1R/IGF
Target Analysis
Alzheimer
• Recent Studies
T2DM Alzheimer
• Strongly altered expression of IGF1R/IGF
Frequency
• Patients with T2DM have a 2- to 3-fold increased risk for AD.
Angelo Nuzzo IIT@SEMM, Milan, 2011
Target AnalysisFrequency
Only one standard target
Angelo Nuzzo IIT@SEMM, Milan, 2011
Conclusion
• Network-based method for feasible identification of multi-component synergy
• Data integration
• Topological feature analysis
• Scoring system
Results:
• Disease-causative pathways
• Potential multi-target nodes
• To discover new therapies
• To support the standard therapies
Future steps:
• Testing on other complex diseases
• Application to directed network
Thanks.
NETWORK-BASED TARGET RANKING FOR POLYPHARMACOLOGICAL THERAPIES