Gene and Protein Networks II Monday, April 16 2006

1
Gene and Protein Networks IIMonday, April 16 2006

• CSCI 4830 Algorithms for Molecular Biology
• Debra Goldberg

2
Outline

1. Recap
2. Confidence assessment, edge prediction (contd)
3. Predicting protein function
4. Predicting protein complexes/functional groups
5. Network integration
6. Caveats, cautions, practical issues

3
Summary of network models
Random not grown, low clustering, short distances, Poisson degree distribution
Regular (lattice) high clustering,long distances
Small world high clustering, short distances
Scale-free power law degree distribution
Hierarchical high clustering, modular, power law degree distribution
4
There is information in a genes position in the
network

• We can use this to predict
• Relationships
• Interactions
• Regulatory relationships
• Protein function
• Process
• Complex / molecular machine

5
Confidence assessment

• Can use topology to assess confidence if true
  edges and false edges have different network
  properties
• Assess how well each edge fits topology of true
  network
• Can also predict unknown relations

6
Prediction

• A v-w edge would have a high clustering
  coefficient

v
w
7
Outline

1. Recap
2. Confidence assessment, edge prediction (contd)
3. Predicting protein function
4. Predicting protein complexes/functional groups
5. Network integration
6. Caveats, cautions, practical issues

8
Interaction generality

• Confidence measure for edge based on topology
  around neighbors.

Saito, Suzuki, and Hayashizaki 2002,2003
9
Confidence assessment

• Integrate experimental details with local
  topology
• Degree
• Clustering coefficient
• Degree of neighbors
• Etc.
• Used logistic regression

Bader, et al., Nature Biotechnology 2003
10
The synthetic lethal network has many triangles
Xiaofeng Xin, Boone Lab
11
2-hop predictors for SSL

• SSL SSL (S-S)
• Homology SSL (H-S)
• Co-expressed SSL (X-S)
• Physical interaction SSL (P-S)
• 2 physical interactions (P-P)

Wong, et al., PNAS 2004
12
Multi-color motifs
Zhang, et al., Journal of Biology 2005
13
Outline

1. Recap
2. Confidence assessment, edge prediction (contd)
3. Predicting protein function
4. Predicting protein complexes/functional groups
5. Network integration
6. Caveats, cautions, practical issues

14
Computationally predicting protein function

• Homology
• Machine Learning
• Graph-theoretic methods

15
Majority method

• Consider immediate neighbors
• Guilt by association
• Schwikowski, et al., Nature Biotechnology 2001

16
Neighborhood method

• How does frequency affect assignment?
• Consider a given radius
• Hishigaki, et al., Yeast 2001

17
Minimum Cut methods

• Minimize interactions between proteins with
  different annotations
• Vazquez, et al., Nature Biotech. 2003
• Karaoz, et al., PNAS 2004

18
Functional flow

• Use network flow algorithm to transport
  function annotation
• Nabieva, et al., Bioinformatics 2005

19
A Markov Random Field method

• Function prediction based on
• Frequency of each function
• neighbors
• of these neighbors with function in question
• Functional linkage graph
• Iterate twice
• Letovsky and Kasif, Bioinformatics 2003

20
Outline

1. Recap
2. Confidence assessment, edge prediction (contd)
3. Predicting protein function
4. Predicting protein complexes/functional groups
5. Network integration
6. Caveats, cautions, practical issues

21
Community structure

• Proteins in a community may be involved in a
  common process or function
• Communities are dense subgraphs with sparse
  interconnections

22
Hierarchical clustering (1)Using natural edge
weights

• Gene co-expression
• e.g., Eisen MB, et al., PNAS 1998

from www.medscape.com
23
Hierarchical clustering (2)Adjacency vector

• Function cluster Tong et al., Science 2004
• Find drug targets Parsons et al., Nature
  Biotechnology 2004

24
Topological overlap

• A measure of neighborhood similarity

Ravasz, et al., Science 2002
25
Spectral clustering

• Compute adjacency matrix eigenvectors
• Each eigenvector defines a cluster
• Proteins with high magnitude contributions

Bu, et al., Nucleic Acids Research 2003
positive eigenvalue negative eigenvalue
26
Dense subgraphs

• Spirin and Mirny, PNAS 2003
• Find fully connected subgraphs (cliques), OR
• Find subgraphs that maximize density 2 m / (n
  (n-1))
• Bader and Hogue, BMC Bioinformatics 2003
• Weight vertices by neighborhood density,
  connectedness
• Find connected communities with high weights

27
Betweenness centrality

• Consider the shortest path(s) between all pairs
  of nodes
• Betweenness centrality of an edge is a measure
  of how many shortest paths traverse this edge

• Edges between communities have higher centrality

Girvan , et al., PNAS 2002
28
Finding motifs
29
Finding motifs
30
Motif function and aggregation
31
Motif function and aggregation
32
Outline

1. Recap
2. Confidence assessment, edge prediction (contd)
3. Predicting protein function
4. Predicting protein complexes/functional groups
5. Network integration
6. Caveats, cautions, practical issues

33
Relationships between network data types

• Distinct data sources generally lead to better
  inferences.
• Associations not independent
• Errors independent

34
Various methods with varying goals
35
Incorporating experimental conditions
Luscombe, et al., Nature 2004
36
Party and date hubs

• Protein interaction network
• Partition hubs by expression correlation of
  neighbors

Han, et al., Nature 2004
37
Network connectivity

• Scale-free networks are
• Robust to random failures
• Vulnerable to attacks on hubs
• Removing hubs quickly disconnects a network and
  reduces the size of the largest component

Albert, et al., Nature 2000
38
Removing date hubs shatters network into
communities
Date Hubs
Party Hubs
Many sub-networks
A single main component
39
Multiple species
40
Network alignment

• Across or within species
• Interaction network and genome sequence
• e.g., Ogata, et al., Nucleic Acids Research 2000

41
Outline

1. Recap
2. Confidence assessment, edge prediction (contd)
3. Predicting protein function
4. Predicting protein complexes/functional groups
5. Network integration
6. Caveats, cautions, practical issues

42
Bias Protein abundance

• Abundant proteins are
• more likely to be represented in some types of
  experiments
• More likely to be essential
• Correlation between degree (hubs) and
  essentiality disappears or is reduced when
  corrected for protein abundance

Bloom and Adami, BMC Evolutionary Biology 2003
43
Bias Degree correlation

• Anti-correlation of degrees of interacting
  proteins disappears in un-biased data

Coulomb, et al., Proceedings of the Royal
Society B 2005
44
Data quality and sparseness
45
No gold standard

• Insufficient highly-accurate data
• Gold-standards often used to train and validate
• Insufficient standardization of procedures

46
Significance
47
Final words

• Network analysis has become an essential tool for
  analyzing complex systems
• There is still much biologists can learn from
  scientists in other disciplines
• Network analysis is itself a new and evolving
  field