Bioin401 Project: AffyPipe Final Presentation

1
Bioin401 Project AffyPipeFinal Presentation

• Yifeng Liu,
• Chelsea Ju,
• Chunyan Meng,
• April 10, 2006

2
Outline

• Pipeline software structure
• CEL file upload and access control
• Clustering
• Gene function prediction

3
Pipeline - Implementation

• Apache web server interacts with perl cgi scripts
  on the server side. The cgi scripts call the
  backend R scripts to do the data processing and
  analysis
• The cgi scripts are modulized based on the
  functions, such as DE.pm for differential
  expression analysis, Cluster.pm for gene chip
  clustering, Annotation.pm for gene function
  annotation.

4
Pipeline User Access Control

• The user has to provide a valid email address.
  Javacript will check the format of the email
  string.
• Number of CEL files can be easily restricted to
  protect the server.

5
CEL File Upload

• User can upload the CEL files as control and
  experimental samples
• Confirm with the user before upload

6
Gene Chip Clustering

• Hierarchical Clustering methods can be applied on
  the gene chips based on the expression of the
  selected genes from the previous steps.
• Heat-map of the selected genes expression level
  can be generated if the user wants.

7
Agglomeration methods for distance calculation

• The following methods are supported ward --
  Ward's minimum variance method single -- Also
  called connected, single linkage or nearest
  neighbour complete -- Furthest neighbour or
  compact average-- Also called "average linkage"
  and similar to an algorithm called "centroid"
  mcquitty-- McQuitty's method median -- Median
  (as opposed to average) similarity centroid--
  Geometric centroid
• A help page is available for detailed explanation

8
Distance measures for distance matrix computation

• The following methods are supported euclidean
  -- Usual square distance between the two vectors
  (2 norm).maximum-- Maximum distance between two
  components of x and y (supremum norm)manhattan--
  Absolute distance between the two vectors (1
  norm)canberra-- sum(x_i - y_i / x_i y_i).
  Terms with zero numerator and denominator are
  omitted from the sum and treated as if the values
  were missing.binary-- (aka _asymmetric binary_)
  The vectors are regarded as binary bits, so
  non-zero elements are 'on' and zero elements are
  'off'. The distance is the _proportion_ of bits
  in which only one is on amongst those in which at
  least one is on.minkowski-- The p norm, the pth
  root of the sum of the pth powers of the
  differences of the components. A help page is
  available for detailed explanation
• A help page is available for detailed explanation

9
Gene Chip Cluster
10
Gene Expression Heat-map

• Orage-Red sequential color range

d7b
d7a
d7a-hiv
d7b-hiv
11
Gene Function Prediction

• Input
• Unknown genes
• Annotated genes with GO ids and biology terms
• Output
• Predicted gene functions with GO ids and biology
  terms indicated by the nearest annotated gene in
  the same cluster in the context of the selected
  interesting genes

12
Example Input

• X56681_s_at GO0006357(regulation of
  transcription from RNA polymerase II
  promoter) GO0006355(regulation of transcription,
  DNA-dependent) GO0045449(regulation of
  transcription) GO0000785(chromatin) GO0003702(RN
  A polymerase II transcription factor
  activity) GO0030528(transcription regulator
  activity)
• M15205_at GO0006139(nucleobase, nucleoside,
  nucleotide and nucleic acid metabolism) GO0044237
  (cellular metabolism) GO0008152(metabolism) GO00
  50875(cellular physiological process) GO0044238(p
  rimary metabolism) GO0005737(cytoplasm) GO000479
  7(thymidine kinase activity) GO0019136(deoxynucle
  oside kinase activity) GO0019206(nucleoside
  kinase activity)

HG2815-HT4023_s_at HG2815-HT2931_at
13
Example Output

• HG2724-HT2820_at GO0016021(integral to
  membrane)GO0031224(intrinsic to membrane)
• HG1869-HT1904_at GO0008429(phosphatidylethanolam
  ine binding)GO0005543(phospholipid
  binding)GO0008289(lipid binding)

14
Possible future work for gene function prediction

• Associate the gene function based on the distance
  from different clusters.
• Clustering only the selected interesting genes
  may not give strong support for the gene function
  prediction. More factors should be considered.