Improving PPI Networks with Correlated Gene Expression Data

advertisement
Improving PPI Networks with
Correlated Gene Expression Data
Jesse Walsh
Background
• PPI networks are currently derived either
computationally or experimentally
• It is well known that there are a great number
of false positives and false negatives in
computationally derived networks
• High-throughput
– Two major published yeast PPI experiments
showed only ≈150 similar interactions out of
thousands [1]
Goal: Improve the Data Quality
• The goal of this study is to improve the quality
of computationally predicted protein-protein
interactions
• Hypothesis:
– Proteins that interact may also have similar
expression patterns
– Gene coexpression is correlated to PPIs
Previous Work
• Deane et al. (2001) [2]
– Proposed EPR metric: use
gene expression profiles to
assess the quality of
computationally predicted
protein-protein interactions
Figure adapted from Deane et al. [2]
Glimpse at the Data
Interaction Data:
DIP dataset statistics
Genome Size
Interactions
Proteins
E. Coli (2008)
4.6 million bp
7447
1863
Yeast (2001)
--
8063
4150
Yeast (2008)
12.5 million bp
18440
4943
DIP (Database of Interacting Proteins) http://dip.doe-mbi.ucla.edu/dip/Main.cgi [3]
Affymetrix Expression Data:
M3D (Many Microbe Microarrays Database)
Number of Experiments: 466
Number of Chips: 907
Genes: 4298
M3D (Many Microbe Microarrays Database) http://m3d.bu.edu/cgi-bin/web/array/index.pl?section=home [4]
Expression Data Selection
• Concerned about complications from adding
to many expression conditions
– Knockouts, over-expression, foreign genes
• Selected a group of 20 conditions that were
published as a part of the same experiment
– Hope for more homogenous data
– Allen et al. [5]
Method
Gene Expression Distance
• Expression Distance given by:
• Summed over 20 conditions
– Treated the first condition (wild-type anaerobic) as
the reference condition for lack of a true control
Expression Distance equation from Deane et al. [2]
Method
DIP Data
• DIP labels as ‘core’ or ‘non-core’
– Corresponds roughly to small scale experiments and
high-throughput experiments
• 3 Interaction Datasets
– Core
• Core interactions
– Non
• Non-core interactions
– Rand
• 100,000 random interactions were created
Method
Mapping
• DIP interaction set used uniprot protein
identifiers, while M3D used gene ids
• Ran a blast of protein sequences against
translated E. coli genome to map the datasets
together
• Lost most of my data on this step
Number of
Interactions
Mapping
Available
CORE
991
220
NON
5999
903
RAND
100,000
100,000
Results
Density distribution of squared distances
Bin size = .05
Results from Deane et al. [2]
Bin size = 1.25
Figure adapted from Deane et al. [2]
Results from Deane et al. [2]
Least Squares Factorization
Figure adapted from Deane et al. [2]
Discussion
• Mapping/Demerging problem
– Kept 1044 of my 6991 interactions (15%)
• Case study P22885
– Obsolete since 2005
– Demerged to P0A8P6 and P0A8P7
• Tyrosine recombinase xerC
• All three have a perfect ClustalW match
Discussion
• Shape of curve
– Multimers and proteins that link to themselves in
the PPI network (the zeros problem)
– 66.6% of Core, 43.6% of Non, <0.1% of Rand
Conclusion
• Cannot predict novel interactions
• Cannot assign confidence values to individual
interactions
• Can provide some measure of the overall
quality of a PPI dataset
Thank You
•
•
•
•
•
•
References:
[1] Deeds EJ, Ashenberg O, Shakhnovich EI. “A simple physical model for scaling in protein-protein
interaction networks.” Proc. Natl Acad. Sci. USA (2006) 103:311–316
[2] Charlotte M. Deane et al. “Protein Interactions: Two Methods for Assessment of the Reliability
of High Throughput Observations.” Molecular & Cellular Proteomics 1.5 349-356
[3] Salwinski L, Miller CS, Smith AJ, Pettit FK, Bowie JU, Eisenberg D (2004) The Database of
Interacting Proteins: 2004 update. NAR 32 Database issue:D449-51
[4] Faith JJ, Driscoll ME, Fusaro VA, Cosgrove EJ, Hayete B, Juhn FS, Schneider SJ, and Gardner TS.
Many Microbe Microarrays Database: uniformly normalized Affymetrix compendia with structured
experimental metadata. Nucleic Acids Research
[5] Timothy e. Allen et al. “Genome-Scale Analysis of the Uses of the Escherichia coli Genome:
Model-Driven Analysis of Heterogeneous Data Sets.” J Bacteriol. 2003 November, 185(21): 63926399
Download