Expression analysis introduction

1
Expression analysis introduction
Brian P. Dalrymple
2
Not about gene expression anymore
Single cell analysis
EST/cDNA arrays
Oligo arrays
RNASeq
Slot blots
10,000s
100,000s
10s
1,000s
Good gene coverage
Splice variants ncRNAs etc
More and more samples
candidates
Limited sampling
DISCO analysis
Correlation networks
Reverse engineering networks
Network comparisons
?????
3
Or even more ambitiously
Environment And Management
Populations
Metabolites, fluxes and communication
Organisms
Organism subsystems
Tissues/organs
Neighbouring cells
Cells
Cellular subsystems
Genes, proteins and genome sequences
4
How is gene expression data being used?
Populations
Metabolism and communication
Organisms
Organism subsystems
SNPs
Tissues/organs
Gene expression
Neighbouring cells
Cells
Cellular subsystems
5
Deeper and deeper for all
Single cell analysis
EST/cDNA arrays
Oligo arrays
RNASeq
Slot blots
10,000s
100,000s
10s
1,000s
Good gene coverage
Splice variants ncRNAs etc
More and more samples
candidates
Limited sampling
DISCO analysis
Correlation networks
Reverse engineering networks
Network comparisons
?????

• Size of networks sets to increase 10 to 100 fold
• Current visualisation tools and our heads
  struggle to cope with 5000 genes
• Sean Grimmond will shine a light on this
• How far can we make predictions about protein
  activity based on gene expression data
• High throughput gene expression data will remain
  relatively cheap compared to proteomics and
  metabolomics data
• Nick Hudson will give an example of recent
  progress in this area
• Validation of predictions
• Already most predictions from analysis of gene
  expression data are never tested
• Shiv. Hiriyur-Nagaraj

6
Talking about predictions

• Perhaps lots of biologists are wary of
  predictions made from complex statistical
  analysis of complex datasets
• Especially as they may not completely understand
  the process by which the predictions are made
• How do we know that our predictions are good, or
  which methods are better than others?
• Mainly anecdotal
• But there is a initiative in place to try to
  achieve a measure of the accuracy DREAM2,
  recently described in the Annals NY Acad
  Sciences, focussing on networks
• Makes interesting reading, perhaps or perhaps not
  surprising that most bioinformatics methods for
  network predictions were not much better than
  random.
• Most successful approaches not necessarily those
  generally expected
• LOTS of scope for testing accuracy and improving
  quality of predictions
• Gene expression analysis is here to stay and a
  central component of molecular biology research

7
Gene expression in the future
TFs/Histones
DNA sequence
ChIP on Chip
Methylation
TF-binding sites
Other purtebations
Molecular Systems Biology
Gene expression
Knock down
Metabolites
Knock in
Knock out
Protein expression
Post trans modification
Proteins