fMRI guided Microarray analysis

1
fMRI guided Microarray analysis

• Imaging-Guided Microarray Isolating Molecular
  Profiles That Dissociate Alzheimers Disease from
  Normal Aging
• A.C. Pereira, W. Wu S.A. Small
• Ann NY Acad. Sci. 1097, Feb 2007
• Combining Brain Imaging with Microarray
  Isolating Molecules Underlying the Physiologic
  Disorders of the Brain
• A. Pierce S.A. Small
• Neurochemical Research, Vol. 29, No. 6, June 2004

2
Crash course The CELL and microarrays in 3 slides

• Cells internal processes and inter-cell
  communication based on proteins
• Goal Figure out which proteins exist in a cell
  under some condition
• Condition e.g. disease
• Many times detect proteins differentially
  expressed e.g. disease vs. control
• Basic staining a specific protein and follow it
  under a microscope
• Next The CELL

3
From DNA to Protein

• (Final) product Protein
• Intermediate product mRNA
• Idea measure mRNA to get protein measurements
• Simultaneous measurements by hybridization

4
DNA Microarrays

• mRNA concatenation of nucleotides
• 4 types ATGC pegs/holes
• Process
• Crush cell
• Wash all but mRNA
• Glue lamps
• Spill on chip
• Shake well!

5
Sorry, 4 slides...

• Chip design probes for genes
• Light on --gt Protein exists
• Light off --gt No protein at the moment

6
Problem setting

• Given two sets of DNA microarrays
• Disease
• Control
• Extract a set of differentially expressed genes
• Feature selection for classification
• Biological significant features for downstream
  research

7
Problem setting revisited

• Given two sets of DNA microarrays
• Disease
• Control
• fMRI measurements of the two populations
• Extract a small set of differentially expressed
  pathogenic-behaving genes
• Feature selection for classification
• Biological significant features for downstream
  research

8
Nervous System Diseases

• Multiple categorizations
• Organic vs. Functional
• Anatomic vs. Physiologic
• Structural vs. Metabolic
• Physiologic molecular pathway
• Invisible to (non functional) imaging
• Not evident under microscope, no histological
  markers
• Anatomic loss/gain of tissue

9
A Needle in a Haystack

• Target Find the one(?) molecule that
  malfunctions
• Multiple molecular pathways within a neuron
• Neuronal interconnection
• Cascade/ripple throughout the system
• Molecule -gt Neuron (population)
• Neuron -gt Other neuron
• Other neuron -gt Other molecules
• Molecules might be in the same neuron population
  (feedback)
• infeasible for standard statistical analysis

10
Aging and AD

• Cognitive decrease (AD and aging)
• Differential vulnerable vs. resistant
• Memory Encoding
• Hippocampus
• Entorhinal Cortex
• Dentate Gyrus
• CA subfields
• Subiculum
• Common process
• Synaptic Failure leads to
• Cell loss / tangles / plaques
• Function, not structure!

11
Hippocampus

• AD
• Aging
• Known from postmortem,in-vitro, and fMRI
• Interconn.
• Asses all regions together

12
Microarray analysis

• Differential expression analysis
• Blind analysis
• Thousands of parameters simultaneously
• High false positives rate (multiple comparisons,
  recall FDR)
• Poor signal-to-noise ratio
• Usually produce a list of differentially
  expressed genes
• list can be very long (up to hundreds)

13
Statistical Modeling

• Temporal model
• 2nd stage for fMRI
• Double subtraction
• With sickness - basal metabolic rate changes as
  well

14
Multiple Studies

• Why fMRI and not postmortem?
• p.m. biased against earliest (and most
  discriminatory) stages
• Only fMRI can image the cell-sickness stage
• EC found to be the primary source of dysfunction
  in AD
• What about normal aging?
• Age-related changes in the EC matched
  pathological decline
• Age-related changes in the dentate gyrus (DG),
  and subiculum (SUB), matched normal aging

15
Spatio-Temporal Model

• How a pathogenic molecule should behave?
• Differentially expressed in the EC (vs. no
  differentially expression in the DG)
• Differences between AD and controls should be age
  independent
• once EC dysfunction begins it does not worsen
  across age groups or over time

16
Results

• 5 Molecules matched the pattern
• Much less than 100s!
• Best molecule VPS35
• Part of a complex that connects-to and transports
  substances within a cell
• A-beta a known smoking gun for AD
• Experiments validated
• Low VPS35 --gt High A-beta
• Required neuronal molecules in end-to-end
  transportation are not transported --gt brain
  dysfunction

17
Conclusion

• Microarrays noisy, unfocused results
• fMRI imaging in-vivo, not post-mortem
• Create statistical model (criteria) using fMRI,
  for microarray differentiation
• Lack of specific methods
• Not a parametric model, like a thumb rule
• Nice example for research advance
• My personal research is on PD
• Lots of imaging data
• Any suggestions?
• Thanks!