Hans A. Kestler

1
MCGH Analyzer

• Hans A. Kestler
• André Müller

2
Data processing steps

• Scanning of the DNA chips (normal and switched)
• 2 Channels (Cy 5 and Cy 3)
• Build mean/median over the pixels
• Further processing with MCGH Software

3
MCGH software

• Background reductioncalculate intensities
  according to the background
• Quality control of the spotsreject spots not
  fitting the quality criteria
• Accumulate spots to clones
• Check testreject clones not fitting the visual
  options
• Select control clones
• Reduce control clones
• Main calculation loop

4
Overview
5
Background reduction

• Background reduction to get intensities
• No reduction
• Fixed reduction
• Local reduction
• Global reduction
• Local Fixed reduction
• Global Fixed reduction
• Compute log Ratios
• log( IntCy3 / IntCy5 )
• log( IntCy5 / IntCy3 )

6
Quality control

• Reject spots with
• flags marked by the scanning software(bad, not
  found, absent, normal ...)
• A background intensity brighter than the
  foreground (new!)
• Min/Max reduction
• Reject the n smallest ratios
• Reject the n largest ratios

7
Spots to clones

• Accumulate the non-rejected spot values
• Mean
• Standard deviation
• Median
• over
• Intensities (Cy3, Cy5)
• log Ratios
• New Feature Reject clones with less than
  SpotLowerBound valid spots.

8
Check test

• Reject clones if at least one of these conditions
  holds
• Me(di)an background intensity gt Background upper
  bound
• Me(di)an Cy3 Intensity lt Me(di)an Cy3 background
  intensity x Intensity lower bound
• Standard deviation Cy3 Ratio gt Ratio SD upper
  bound

9
Select control clones

• Only non-rejected clones will be selected as
  control clones.
• Manual selectionSelect clones with id 91 or
  k or K or ?91 as control clone
• Automatic selection
• No AutoBandCutoffPercentage clones from the
  middle band
• AutoBandSelect band around the median

10
Reduce control clones

• Some of the control clones will be rejected ...
• Cutoff PercentageReject the n smallest
  ratios
• Without Cutoff BandReject the n largest
  ratios
• Cutoff BandReject band around the median

11
Main calculation loop

• Calculate control means (the mean/median over all
  control clones/spots)
• Normalize ratios (subtract control mean from the
  ratio)
• Calculate tolerance value Ts standard
  deviation of the ratios of the observed
  clonen the number of valid spots in this
  clonet value of the t-statistic significance
  niveau
• Force T-Test Reject clones with T gt Force T
  Value
• C Check Replace tolerance values with
  possible greater values.
• Find clone with maximum tolerance and reject it
  if its tolerance value T is gt Force T Value
• Perform T Test and evaluate result value.
• Everything has to be recalculated if a control
  clone will be rejected.

12
The C Check

• The clone tolerance values are now recalculated
  according to the following scheme
• If the new tolerance value is greater than the
  old T will be replaced by the new value

13
The T Test

• If Force T is set, the value will be set
  to the Force T Value
• otherwise it is the greates tolerance value found
  in the clones.

14
The T Test (2)

• Calculation of the result value R
• T Test

• No T Test thresholding

In this routine the test T gt Force T Value
will be performed repeatedly
15
NCBI Clone Database

• Integration of the NCBI component database
• Automatically mapping of clone ids to accession
  numbers, genomic clone locations and clone status
  information according to an up-to-date database
• Direct import of the NCBI file format

16
Database-generated Information
Accession-Number
Start-Base
End-Base
Clone-State
17
Batch Processing

• One ore more file pairs can be added to a session
• All computations are performed simultaneous on
  the included datasets

18
Diagrams functions

• Ratio-profiles of multiple clone sets can be
  shown in one diagram

19
Ideogram Browser 1

• Independent portable Java application
• Automation from MCGH-Analyzer with JNI
• Generation of ideogram drawings from the NCBI map
  database
• Direct representation of gain and lost markers of
  multiple clone sets
• Scalable and scrollable graphs

20
Ideogram Browser 2
21
Software Structure 1

• Excel as convenient platform with widely known
  user interface for
• Table representation
• Diagram drawing
• User interaction
• Windows DLL written in C for high performance
  using COM automation
• Platform-independent Java-Application for
  visualizing ideograms (can be docked to the DLL
  via JNI)

22
Software Structure 2
23
Future Features

• Copy number estimation
• Global thresholds
• Adaptive (local) thresholds
• Wavelets
• Adaptive weights smoothing
• NCBI database online update
• Interface to the R platform