Create and assess protein networks through molecular characteristics of individual proteins

1
Create and assess protein networks through
molecular characteristics of individual proteins

• Yanay Ofran et al. ISMB 06
• Presenter Danhua Guo
• 12/07/2006

2
Roadmap

• Motivation
• Introduction
• Methods
• Results and Discussion
• Conclusion

3
Motivation

• Study of biological systems relies on network
  topology.
• Integrating protein information into the network
  enhance the analysis of biological systems.

4
Introduction

• Protein-Protein Interaction (PPI) Network
• Help identify process or functions
• Major problem
• Generation problem
• Experimental errors should not be in the network
• In vitro should be include in the network
• Data representation problem
• Essential connection between PPI and protein

5
Introduction

• An ideal framework
• Macro level network topology
• Micro level characteristics of each protein
• Localization
• Functional annotation

6
Introduction

• Protein interaction Network Assessment Tool
  (PiNAT)

7
Methods

• Large-scale Assessment of PPIs
• Based on localization
• Based on GO annotation (if applicable)
• Automatic generation of networks
• Get submitted list of proteins from user
• Search DIP and IntAct
• Display of networks in the cellular context
• Alzheimers disease related pathway

8
Methods

• Localization criteria
• LOCtree classify eukaryotic proteins (60)
• Threshold confidence score gt4
• PHDhtm predict transmembrane helices (7)
• Threshold average score among 20 reliable
  predictions gt8.5
• Experiment on 4800 interactions (2191 proteins)
• High-confidence prediction 2312 (1482 proteins)
• Total protein pairs 1,097,421
• Binomial approximation to the cumulative
  hypergeometric probability distribution to get a
  p-value for over and under representation

9
Methods

• GO criteria
• The functionality annotation of a protein
• Distance between 2 GO terms measure the
  similarity
• m,n respective numbers of annotations in i and j
• simGo GO similarity defined by Lord et al.
• Ck, Cp respective individual annotation in
  protein i and j
• Cjmax Cks most similar term in j
• Cimax Cps most similar term in i

10
Methods

• Display of networks in the cellular context
• Based on LOCtree and PHDhtm predictions
• Generate Graph Markup Language (GML)
• Localization overide rule
• High PHDhtm gt High LOCtree gt Low PHDhtm gt Low
  LOCtree

11
Results

• Interactions across subcellular compartments
• Intra-compartment interactions high score
• Distant compartment low score
• Nearby compartment likely

12
Results

• Likely and unlikely interactions across GO
• Likely gt3.25
• Unlikely lt1.3
• Neutral else

13
Result

• Alzheimer in the perspective of PiNAT
• Reflects the unclarity regarding Amyloid beta A4
  protein (APP) s localization
• APP interacts extensively with almost every
  compartment of the cell

14
Result

• APPs role in Alzheimer
• APP-related PPI deemed unlikely
• Conflicts between 2 scoring systems

15
Conclusion

• Molecular knowledge and network structure can
  enhance our understanding of biological
  processes.
• PiNAT is efficient and meaningful.