Why cholesterol can be good for you: A case study in integrated postgenome science

1
Why cholesterol can be good for you A case study
in integrated post-genome science

• Andy Brass

2
Overview

• Trypanosomiasis
• A strategy for genotype to phenotype
• Functional genomics analysis
• Computational infrastructure
• Applications in intensive care

3
Introduction to trypanosomiasis
4
Trypanosomiasis Is a fatal disease of livestock.

The livestock equivalent of sleeping sickness in
humans
T. congolense, T. vivax
T brucei rhodesiense T gambiense
5
Origins of NDama and Boran cattle
NDama
Cattle Tsetse Cattle and tsetse
Cattle Glossina Cattle and Glossina
6
(No Transcript)
7
Mouse models of trypanotolerance.

8
Trypanosoma infection response (Tir) loci
C57/BL6 x AJ and C57/BL6 x BALB/C
Iraqi et al Mammalian Genome 2000 11645-648
Kemp et al. Nature Genetics 1997 16194-196
9
(No Transcript)
10
Questions

• Can we breed cattle resistant to Trypanosomiasis
  (Nagana)?
• What are the causes of the differences between
  resistance and susceptible strains and mice?

11
From genotype to phenotype a strategy
12
Tools

• Understanding phenotype
• Microarrays
• Comparing resistant vs susceptible strains
• Comparing infected vs naïve mice
• Understanding genotype
• Classical Genetics
• Mapping quantitative traits

13
The underlying assumption

• QTL studies
• what are the important differences?
• Microarray
• what are the effects of those differences?
• Both together should be more powerful than either
  on their own

14
Self-organising networks of proteins (and other)
molecules Physics
Genotype Genome sequences Annotation Tell me
everything you can about gene x SNPs QTL
Phenotype What are the symptoms of disease
y? Health Informatics
Microarrays?
15
A warning!

• Treacher Collins Syndrome (Prof. Mike Dixon)
• Simple phenotype
• Cranio-facial disorder
• some hearing loss
• Simple genotype

16
Issues raised

• Knowledge not in the literature
• In what tissue does the mutation make a
  difference?
• At what time is the effect important?
• Would not have known to look in neural crest
  during embryo development

17
Experimental design

• Middle out
• Use arrays to define phenotype
• From array to trait
• From array to genome

Transcriptome
Phenotype
Genotype
18
Functional genomics
19
The experiment
Liver
AJ
Spleen
Balb/c
Kidney
C57
0
9
17
3
7
Tryp challenge
20
Mouse Microarray data

• Hybridised spleen and liver RNA from 5 timepoints
  post infection to Affymetrix 430
• Five biological replicates for each condition.
• Each biological replicate a pool of five mice
• 225 arrays
• Data looks very clean

21
Data analysis

• Identify pathways that have excess numbers of
  responding genes
• Track genes through pathways that are suspected
  of being relevant
• Identify clusters of responding genes that have
  common transcription factor binding sites.

22
Macrophage wiring diagram

• Mapping data on an SBML wiring diagram of a
  macrophage (Kitano, 2004)

23
Wiring diagram of a macrophage
24
C57 day 3 vs day 0
25
C57 day 7 vs day 0
26
C57 day 9 vs day 0
27
C57 day 17 vs day 0
28
C57 vs AJ day 0
29
C57 vs AJ day 3
30
C57 vs AJ day 7
31
C57 vs AJ day 9
32
C57 vs AJ day 17
33
Hypotheses the phenotype

• Early time point
• Stress response and protection against apoptosis
  (TGF-beta, Il-6)
• Middle time point
• Th1/Th2 switch
• Late time point
• Cholesterol synthetic pathways

34
Apoptosis response
35
Th1/Th2 switch
Classically activated macrophages
Alternatively activated macrophages
Th2 signal (IL4, IL10)
C57
AJ/Balb
0
7
9
3
17
36
Cholesterol cycle
HDL
LDL
Tissues
37
Cholesterol synthesis
38
Functional test

• Feeding experiment
• Modify cholesterol via diet
• Does it affect infection

39
Total Cholesterol levels
40
Cholesterol

• C57 are resistant and maintain cholesterol levels
  well during infection
• AJ are susceptible cholesterol values fall
  rapidly
• Increasing cholesterol lowers parasite levels
  but increases anaemia
• (C57 are slow to switch on bile acid synthesis
  a mechanism for maintaining cholesterol?)

41
Cholesterol Hypothesis

• Maintaining cholesterol levels in acute infection
  is important (macrophage function, suppressing
  immune damage?)
• C57 mice atherosclerotic but trypanotolerant
• AJ very non-atherosclerotic but susceptible

42
A computational infractructure
Genotype
Transcriptome
Phenotype
Define in terms of Pathways Kegg GO Curated
Look at genes in the QTL What pathways do
QTL Intersect What pathways are enriched In the
transcriptom
Which pathways Might be involved
in Phenotype Text Literature
43

• Data sets are getting very big
• Breaking existing tools (Grid)
• Need to support induction/hypothesis
• Need information from many different resources
  (warehouses solutions a non-starter!)

44
Difficult questions that should be easy

• Tell me everything you can about my favourite
  protein
• What name 14-3-3 epsilon, ywhae
• What proteins interact with protein x
• What resources?
• What quality?

45
(No Transcript)
46
Popular tasks

• Tell me everything about my favourite x
• Triaging a list

Extra annotation
Re-annotated list
classifier
Big list
Black box filter
Smaller list
47
Application of workflows

• What genes are in the QTL?
• From genome out
• Tell me everything you can about my favourite
  affy probe
• Characterise the array data
• Look for intersections
• What could myGrid do to help?

48
Example workflow

• Identify all genes in the QTL (ensembl)
• Identify all KEGG pathways associated with the
  genes
• Identify regulated genes (maxD)
• Identify KEGG pathways of regulated genes
• Look at the intersection (common pathways

49
Self-organising networks of proteins (and other)
molecules Physics
Genotype Genome sequences Annotation Tell me
everything you can about gene x SNPs QTL
Phenotype What are the symptoms of disease
y? Health Informatics
Microarrays?
50
Genes in QTL
Hypothesis?
Re-annotated smaller list
Big list
Black box filter
Smaller list
Classifier
Re-annotated smaller list
Big list
Black box filter
Smaller list
Genes responding on the Affy chip
51
Architecture
myGrid
Workflows
Bioconductor Matlab
maxdBrowse
Pierre
Instance Store
WSDL
Web services layer
Metadata layer/ontologies
FUGE
Knowledge Base
Metadata databases
maxdLoad
PEDRo
Raw data layer
52
QTL workflow
53
Results

• Being systematic matters!
• Identified intersection in adipocytokine pathway
• RXRbeta is under the Tir1 QTL peak
• RXRbeta has polymorphisms which go with phenotype
• RXRb sits at the inflammation/lipid pathway
  intersection

54
Parallels with Intensive care
Data from Hope Hospital
55
Acknowledgements

• Trypanosomiasis
• Steve Kemp (Liv)
• Harry Noyes (Liv)
• Anthea Broadhead (Liv)
• John Gibson (Armidale)
• Morris Agabo (ILRI)
• Helen Hulme
• Leo Zeef
• Tim Hinsley
• Kathryn Else
• Jans Naesson (ILRI)
• Wellcome
• maxd
• Norman Morrison
• David Hancock
• Giles Velarde
• Michael Wilson

• E-Science
• Carole Goble
• Norman Paton
• Paul Fisher
• Katy Wolstencroft
• Hannah Tipney
• Antoon Goderis
• Tom Oinn (EBI)
• Connie Hedeler
• Peter Li
• NIBHI
• Pat Baker
• Iain Buchan
• Salford
• John New
• Paul Dark

56
Modelling transcription factors