NK MacLennan, J Dong, S Horvath, L Ornelas, L Rahib, K Dipple and ERB McCabe.

Network Analysis of Glycerol Kinase Deficient Mice Predicts Genes Essential for Survival: A

Systems Biology Approach

NK MacLennan, J Dong, S Horvath, L Ornelas, L Rahib, K Dipple and ERB McCabe.

UCLA, Los Angeles, CA, United States.

Glycerol Kinase

• Catalyzes the reaction

Glycerol glycerol 3-phosphate, a substrate for gluconeogenesis and lipid metabolism

Human Glycerol Kinase Deficiency (hGKD)

• hGKD is an X-linked inborn error of metabolism.

• Symptoms include metabolic and central nervous system deterioration.

• Treatment: low-fat diet.

• There is no satisfactory correlation between GKD genotype and phenotype.

Mouse Model of GKD

• GK knockout (KO) mice model the human GKD phenotype. Huq et al., Hum Mol Genet. 1997; Kuwada et al., Biochem Biophys Res Commun. 2005

• Unlike humans, mice die at 3-4 days of life (Dol).

Objective

• Identify genes associated with survival of WT mice using network analysis that relates a measure of differential expression to connectivity.

• Highly connected highly differentially expressed genes have been found to be predictors of survival.

Methods• Microarray analysis on liver mRNA

• Expression data was filtered for the top 10% most varying probe sets for Weighted Gene Co-Expression Network Analysis (WGCNA).

WT WTKO C

Weighted Gene Co-Expression Network Analysis

(WGCNA) Overview

http://www.genetics.ucla.edu/labs/horvath/CoexpressionNetwork/

Construct a networkRationale: make use of interaction patterns between genes

Identify modulesRationale: module (pathway) based analysis

Relate modules to external informationArray Information: Sample dataGene Information: EASERationale: find biologically interesting modules

Find the key drivers in interesting modulesTools: Module connectivity, causality testingRationale: experimental validation, therapeutics, biomarkers

Study Module Preservation across different data Rationale: • Same data: to check robustness of module definition• Different data: to find interesting modules

Construct a networkRationale: make use of interaction patterns between genes

Identify modulesRationale: module (pathway) based analysis

Relate modules to external informationArray Information: Sample dataGene Information: EASERationale: find biologically interesting modules

Find the key drivers in interesting modulesTools: Module connectivity, causality testingRationale: experimental validation, therapeutics, biomarkers

Study Module Preservation across different data Rationale: • Same data: to check robustness of module definition• Different data: to find interesting modules

Construct a Network

Microarray gene expression data

Gene expression correlation

Correlation Matrix

Power adjacency function generates a weighted network

| ( , ) |ij i ja cor x x

Construct a networkRationale: make use of interaction patterns between genes

Identify modulesRationale: module (pathway) based analysis

Relate modules to external informationArray Information: Sample dataGene Information: EASERationale: find biologically interesting modules

Find the key drivers in interesting modulesTools: Module connectivity, causality testingRationale: experimental validation, therapeutics, biomarkers

Study Module Preservation across different data Rationale: • Same data: to check robustness of module definition• Different data: to find interesting modules

Module Identification• WGCNA aim: Detect

modules. • Modules are groups

of highly correlated, highly connected genes.

• Defined with the standard distance measure: 1-correlation.

Construct a networkRationale: make use of interaction patterns between genes

Identify modulesRationale: module (pathway) based analysis

Relate modules to external informationArray Information: Sample dataGene Information: EASERationale: find biologically interesting modules

Find the key drivers in interesting modulesTools: Module connectivity, causality testingRationale: experimental validation, therapeutics, biomarkers

Study Module Preservation across different data Rationale: • Same data: to check robustness of module definition• Different data: to find interesting modules

Connectivity (k) and Gene Significance (GS)

• A measure of a gene’s connection strength to other genes in the whole network.

• Use both k and GS

Module Connectivity

Gen

e S

ign

ific

ance

(G

S)

Construct a networkRationale: make use of interaction patterns between genes

Identify modulesRationale: module (pathway) based analysis

Relate modules to external informationArray Information: Sample dataGene Information: EASERationale: find biologically interesting modules

Find the key drivers in interesting modulesTools: Module connectivity, causality testingRationale: experimental validation, therapeutics, biomarkers

Study Module Preservation across different data Rationale: • Same data: to check robustness of module definition• Different data: to find interesting modules

Construct a networkRationale: make use of interaction patterns between genes

Identify modulesRationale: module (pathway) based analysis

Relate modules to external informationArray Information: Sample dataGene Information: EASERationale: find biologically interesting modules

Find the key drivers in interesting modulesTools: Module connectivity, causality testingRationale: experimental validation, therapeutics, biomarkers

Study Module Preservation across different data Rationale: • Same data: to check robustness of module definition• Different data: to find interesting modules

Results• Unsupervised hierarchical clustering

analysis revealed that overall gene expression profiles of the dol 1 and 3 KO mice differed from WT.

Dol 1 Dol3

Identify Modules and Study Module Preservation

Dol 1 Dol 3

Dol 3 colors Dol 1 colors

Relate Modules to Gene SignificanceGlycerol Kinase Knockout Status

DOL 1 KO• Blue: Underexpressed• Turquoise: Overexpressed

DOL 3 KO • Blue: Underexpressed• Brown: No relationship• Turquoise: Overexpressed

Relate Modules to External Information

Functional Group Enrichment Dol1 Dol3Mitotic cell cycle,

transcription factor binding, response to DNA damage stimulus, protein metabolism,

apoptosis, cell death.

Organic acid/carboxylic acid, lipid, amino acid, steroid and carbohydrate metabolism.

Mitotic cell cycle, protein metabolism, epigenetic regulation of gene expression.

Carboxylic acid/organic acid, fatty acid, amino acid and glucose metabolism.

Find the Key Drivers in Interesting ModulesDol1 Dol3

Module Connectivity

Gen

e S

ign

ific

ance

Module Connectivity

Gen

e S

ign

ific

ance

Module Connectivity

Gen

e S

ign

ific

ance

Module Connectivity

Gen

e S

ign

ific

ance

GKGPDVDAC

GKTATHNF4a

TATHNF4a

GPDVDAC

ACOTACOTPSATPSAT

BCL2BIDGADD45TRP53inp1

ACOTACOTPSATPSATPLK3PLK3

Validation Studies

• Cell Culture– ACOTACOT– PSATPSAT– PLK3PLK3

• KO Mice– ACOTACOT

Summary

• Dol 1 Blue module:

– Genes underexpressed in KO

– GK gene module membership

– Enriched with Apoptosis/ cell death genes

Summary

• Dol 3 blue module:

– Genes Underexpressed in KO

– Loss of Apoptosis/ cell death gene enrichment

Summary

• Dol 1 and 3 Turquoise module:

– Genes overexpressed in KO – ACOT, PSAT, PLK3ACOT, PSAT, PLK3 connected

Summary

• Gene validation studies supported the WGCNA.– ACOTACOT– PSATPSAT– PLK3PLK3

Conclusion

• WGCNA permits the reduction of high dimensionality data to low dimensionality output that is more easily understood– Revealed novel target genes possibly

essential for survival of WT– Provided evidence of an apoptotic role for GK

that is lost in GKD

Acknowledgements

• McCabe Lab

• Dipple Lab

Cell Culture Validation

0

50

100

150

200

250

300

350

***

******

**

***

Gyk Acot Gyk Psat Gyk Plk3Clofibrate Naltrexone Paclitaxel

% o

f C

on

tro

l

GK GKGK

Choice of Power, β