Functional module identification with tomato gene and metabolite expression profiles Cass Peluso Boyce Thompson Institute, Cornell University, Ithaca, NY 14853, USA

1
Functional module identification with tomato
gene and metabolite expression profilesCass
PelusoBoyce Thompson Institute, Cornell
University, Ithaca, NY 14853, USA
Project Leaders Zhangjun Fei, Ph.D, Je-Gun
Joung, Ph.D
Introduction
Results
Abstract
B
D
A
Cells carry out a multitude of complex functions
through the coordinated effort of a set of genes.
Such activity is often carried out through the
organization of the genome into regulatory
modules. Modules are sets of co-regulated genes
that share a common function. The identification
of modules, their regulators, and the conditions
under which regulation occurs is thus very
important since a good deal of a cells activity
is organized into this network of interacting
modules. It is essential that these modules be
identified and their functions be determined in
order to understand cellular responses to
internal and external signals (Segal et al.
2003). Here we report the identification of
functional modules in the tomato using gene
expression and metabolite profile datasets
generated from a set of Solanum pennellii
introgression lines.
TMV response-related gene Product (WRKY)
heat shock protein salicylic acid-binding
protein gibberellin 2-oxidase
Figure 3. Representative functional modules (A)
The inferred regulatory modules. (B) Module 35
contains a pathogenesis-related TF as a
regulator. It also has a number of genes that are
potentially involved in plant responses to biotic
and abiotic stresses. This module is thus likely
related to pathogen response, which could have
important implications for the creation of
disease-resistant tomato varieties. (C) Module 6
shows two regulators acting on gene products that
relate to the cell wall. The likely function of
this module is related to cell wall organization
and biogenesis. (D) Module 20 contains
phytofluene, a metabolite in the carotenoid
biosynthesis pathway.
C
Methods
Figure 2. Computational pipeline for module
identification

• Tomato TOM2 array transcriptional factor
  identification
• Step 1 Blast tomato TOM2 probe sequences against
  SwissProt and TrEMBL protein databases. Parse
  results using BioPerl to extract probe IDs and
  hit accessions.
• Step 2 Map TOM2 array probe IDs to GO term IDs
  using the Gene Ontology Annotation Database (GOA)
  based on their homologues in SwissProt and
  TrEMBL.
• Step 3 Associate GO IDs and GO names using the
  Gene Ontology definition file (OBO v1.2)
  downloaded from http//geneontology.org.
• Step 4 Add each GO name to each GO ID in the
  result file from Step 2.
• Step 5 Identify TOM2 array probes with GO names
  of the desired regulators.
• Tomato functional module identification
• Step 6 Impute gene expression dataset.
• Step 7 Make input expression dataset Convert
  absolute value to log value (for gene and
  metabolite profiles), choose expressed gene in
  introgression lines, and merge expression
  profiles.
• Step 8 Make Genomica input file
• 8.1 Insert associated genes (SGNs) with symbols
  (LEs) and sort.
• 8.2 Get symbols for the regulators.
• 8.3 Extract and add the expression data for the
  regulators, add the associated symbols, and
  merge them into the output file from step 8.1.

MADSbox cell wall organization and
biogenesis cell wall protein
First, a computational pipeline was implemented
to identify transcription factors on tomato TOM2
oligo-nucleotide arrays (See Fig. 2 for details).
Then, the gene expression profiles generated
using the TOM2 arrays and the targeted metabolite
profiles from twenty-three S. pennellii
introgression lines were processed and
normalized. The processed and normalized gene
expression and metabolite profiles and the set of
candidate regulatory genes on the TOM2 arrays
were then loaded into Genomica, a program that
uses an algorithm to simultaneously search for a
partition of genes into modules and for each
module's regulatory program. A module's
regulation program specifies the set of
regulators that control the module and the
expression of the genes in the module. The
program outputs a list of modules and associated
regulation programs. Fig. 3 shows several
interesting modules that were identified. Each of
the identified modules was then analyzed for GO
term enrichment using a tool in the Tomato
Functional Genomics Database. Significantly
over-represented GO terms were identified in each
module with an adjusted p-value (False Discovery
Rate, FDR) lt 0.05. A heatmap of the significance
of GO term enrichment was generated using the
web-based application Matrix2PNG, with an orange
color signifying that a module has a certain
function (Fig. 4). A list of modules and their
regulators was then processed using the program
Cytoscape, which created a module-regulator
network map, with modules in light blue and
regulators in orange (Fig. 5).

WRKY
CBF1
WRKY
NAC2
WRKY
NAC
WRKY
TAG1
WRKY4
ERF
ERF
Figure 5. The regulator-module network represents
key regulators that are linked to several
different modules. Module 35 shares the
pathogenesis-related transcriptional factor with
modules 4, 31, and 43. These modules need to be
investigated to see if they have the functional
interactions. Modules 6 and 20 also share the TMV
response-related gene product with numerous other
modules.
Figure 4. A heatmap representing the significant
biological functions of modules
References Segal E. et al. (2003) Module
networks identifying regulatory modules and
their condition-specific regulators from gene
expression data. Nat Genet 34 166-176.
Acknowledgements Thank you to BTI and Dr. Je Min
Lee for the IL datasets used and helpful comments
given.