SLIDE 1
Human Interactome
2
RAD50 MSH4 MSH5 PCNA BRCA2 FEN1 RAD51 DMC1 MED6 RFC1
Marinka Zitnik - Stanford University - http://snap.stanford.edu/pathways
Large-Scale Analysis of Disease Pathways in the Human Interactome - - PowerPoint PPT Presentation
Large-Scale Analysis of Disease Pathways in the Human Interactome - - PowerPoint PPT Presentation
Large-Scale Analysis of Disease Pathways in the Human Interactome Marinka Zitnik Joint work with Monica Agrawal and Jure Leskovec Human Interactome RAD50 RFC1 BRCA2 MSH4 PCNA FEN1 MED6 MSH5 DMC1 RAD51 Marinka Zitnik - Stanford
SLIDE 2
SLIDE 3
Human Interactome
3
RAD50 MSH4 MSH5 PCNA BRCA2 FEN1 RAD51 DMC1 MED6 RFC1
Network biology: Interacting proteins tend to lead to similar phenotypes
[Menche et al., Science 2015, Costanzo et al., Science 2016]
Marinka Zitnik - Stanford University - http://snap.stanford.edu/pathways
SLIDE 4
Disease Pathways
§ Pathway: Subnetwork of interacting proteins associated with a disease
4
RAD50 MSH4 MSH5 PCNA BRCA2 FEN1 RAD51 DMC1 MED6 RFC1
Lung carcinoma pathway
Marinka Zitnik - Stanford University - http://snap.stanford.edu/pathways
SLIDE 5
This Work: Research Question
What is the protein interaction network structure of disease pathways?
5 Marinka Zitnik - Stanford University - http://snap.stanford.edu/pathways
SLIDE 6
Disease Pathway Dataset
§ Protein-protein interaction (PPI) network culled from 15 knowledge databases:
§ 350k physical interactions, e.g., metabolic enzyme-coupled interactions, signaling interactions, protein complexes § All protein-coding human genes (21k)
§ Protein-disease associations:
§ 21k associations split among 519 Mendelian and complex diseases
§ Disease categories, e.g., cancers (68), nervous system diseases (44), cardiovascular diseases (33), immune system diseases (21) § Pros: Experimentally validated data, comprehensive analysis
6 Marinka Zitnik - Stanford University - http://snap.stanford.edu/pathways
SLIDE 7
Prediction Task
7 Marinka Zitnik - Stanford University - http://snap.stanford.edu/pathways
SLIDE 8
Methods and Setup
§ 5 methods: neural embeddings, matrix completion, neighbor scoring, diffusion, connectivity significance
§ Get a score for each node: probability that protein is associated with a disease
§ For each disease:
§ Train the method using training proteins § Predict disease proteins in test test
8 Marinka Zitnik - Stanford University - http://snap.stanford.edu/pathways
SLIDE 9
Prediction Results
9
§ Best performers:
§ Random walks hits@100 = 0.36 § Neural embeddings hits@100 = 0.30
§ Worst performer:
§ Neighbor scoring hits@100 = 0.24
hits@100 hits@100 hits@100 Full results for all methods in the paper.
Marinka Zitnik - Stanford University - http://snap.stanford.edu/pathways
SLIDE 10
Prediction Results
10
§ Best performers:
§ Random walks hits@100 = 0.36 § Neural embeddings hits@100 = 0.30
§ Worst performer:
§ Neighbor scoring hits@100 = 0.24
hits@100 hits@100 hits@100 Full results for all methods in the paper.
Limited success of current methods Failure cases not well understood
Marinka Zitnik - Stanford University - http://snap.stanford.edu/pathways
SLIDE 11
What is the network structure of disease pathways?
11 Marinka Zitnik - Stanford University - http://snap.stanford.edu/pathways
How can we explain failure cases of disease pathway prediction?
SLIDE 12
Competing Views
1. Current: Traditional network clusters
§ Well connected internally § Localized in the PPI net § Few edges pointing outside
2. Our work: Multi-regional objects
§ Loosely interlinked § Distributed in the PPI net § Many edges pointing outside § Higher-order connectivity
12 Marinka Zitnik - Stanford University - http://snap.stanford.edu/pathways
SLIDE 13
Are Pathways Well Interlinked?
13 Marinka Zitnik - Stanford University - http://snap.stanford.edu/pathways
vs.
Modularity ≈ 0 Modularity ≈ 1
SLIDE 14
Are Pathways Well Interlinked?
§ No! - Pathways are embedded within PPI net § Modularity: Interactions within the pathway minus the expected interactions
14 Marinka Zitnik - Stanford University - http://snap.stanford.edu/pathways
vs.
Modularity ≈ 0 Modularity ≈ 1
SLIDE 15
Are Pathways Connected?
15 Marinka Zitnik - Stanford University - http://snap.stanford.edu/pathways
vs.
Pathway components = 1 Pathway components = 4
SLIDE 16
Are Pathways Connected?
No! - Pathways have fragmented PPI structure: § 16 pathway components § 10% of pathways have 60+% proteins in the largest component
16 Marinka Zitnik - Stanford University - http://snap.stanford.edu/pathways
vs.
Pathway components = 1 Pathway components = 4
SLIDE 17
Do Pathways Localize in Net?
17
vs.
Localized pathway Dispersed pathway
Marinka Zitnik - Stanford University - http://snap.stanford.edu/pathways
SLIDE 18
Do Pathways Localize in Net?
18
vs.
Localized pathway Dispersed pathway
Marinka Zitnik - Stanford University - http://snap.stanford.edu/pathways
SLIDE 19
Do Pathways Localize in Net?
19 Marinka Zitnik - Stanford University - http://snap.stanford.edu/pathways
Disease pathways are weakly embedded in the PPI network, e.g.:
SLIDE 20
Pathways are Multi-Regional!
20 Marinka Zitnik - Stanford University - http://snap.stanford.edu/pathways
SLIDE 21
How To Proceed?
§ Network motifs: Higher-order network structures
21 Marinka Zitnik - Stanford University - http://snap.stanford.edu/pathways
SLIDE 22
How To Proceed?
§ Network motifs: Higher-order network structures
22 Marinka Zitnik - Stanford University - http://snap.stanford.edu/pathways
Do disease pathways utilize higher-order network structure?
SLIDE 23
Counting Network Structures
§ 73 possible structures of size 2 to 5 nodes (edge à size-5 clique)
23 Marinka Zitnik - Stanford University - http://snap.stanford.edu/pathways
SLIDE 24
Are Network Motifs Abundant?
24 Marinka Zitnik - Stanford University - http://snap.stanford.edu/pathways
SLIDE 25
Are Network Motifs Abundant?
25
Cardiovascular diseases, e.g., Cardiomyopathy, Tachycardia Cancers, e.g., Tumor of salivary gland, Thyroid carcinoma
Marinka Zitnik - Stanford University - http://snap.stanford.edu/pathways
SLIDE 26
Are Network Motifs Abundant?
26
Cardiovascular diseases, e.g., Cardiomyopathy, Tachycardia Cancers, e.g., Tumor of salivary gland, Thyroid carcinoma
§ Higher-order structures provide additional signal past edge connectivity § Lead to better performance (11%, avg.) § Example: Hearing loss: hits@100 = 0.03 à à hits@100 = 0.77
Marinka Zitnik - Stanford University - http://snap.stanford.edu/pathways
SLIDE 27
Summary & Conclusions
§ Current method assumptions not valid § Propose new prediction paradigm:
§ Disease pathways are loosely interlinked § Multi-regional objects with regions distributed throughout the PPI network § Higher-order connectivity is important snap.stanford.edu/pathways
27 Marinka Zitnik - Stanford University - http://snap.stanford.edu/pathways