Transcript Algorithms and Data Analysis in Microarray Technology

Course Work Project
Project title
“Data Analysis Methods for Microarray Based Gene
Expression Analysis”
Sushil Kumar Singh (batch 2002-03)
IBAB, Bangalore
Done at
Siri Technologies Pvt. Ltd.
Bangalore
Outline







Introduction
Overview of Data Analysis
Normalization
Clustering Algorithms
Future work
Acknowledgements
Questions ???
Introduction
Overview of Data Analysis
Normalization


An attempt to remove systematic variation from
data.
Sources of systematic variation –

Biological source


Technical source



Influenced by genetic or environmental factors, Age, sex etc.
Induced during extraction, labelling, and hybridization of
samples
Printing tip problems
Measurement source


Different DNA conc.
Scanner problem
Why Normalize Data



To recognize the biological information in
data.
To compare data from one array to another.
In practice we do not understand the data –
inevitably some biology will be removed too.
Normalization methods

Methods of elements selections




Housekeeping genes
All elements
Using Spiked control
Methods to calculate normalization factor



Log ratio
Lowess
Ratio statistics
Clustering

For a sample of size “n” described by a ddimensional feature space, clustering is a
procedure that

Divides the d-dimensional features in K-disjoint
groups in such a way that the data points within
each group are more similar to each other than to
any other data point in other group.
Clustering algorithms

Unsupervised – without a priory
biological information



Agglomerative – Hierarchical
Divisive – K-means, SOM
Supervised – a priory biological
knowledge
 Support vector machine (SVM)
Hierarchical clustering (HC)

Agglomerative technique

steps





The pair-wise distance is calculated between all genes.
The two genes with shortest distance are grouped together
to form a cluster.
Then two closest cluster are merged together, to form a new
cluster.
The distances are calculated between this new cluster and
all other clusters
Steps 2 to 4 are repeated until all the objects are in one
cluster.
HC contd.

Data table
HC contd.
• Calculation of
distance matrix
using data table.
Experiment » Axis
Log ratio of genes »
Coordinates
• For n-experiments n
dimensional space
HC contd.

Distance between genes

Euclidean distance

Pearson correlation

Semi-metric distance – Vector angle

Metric distance – Manhattan or City block
HC contd.

Distance between clusters
 Single linkage clustering

Complete linkage clustering

Average linkage clustering

UPGMA
Weighted pair-group average

Within-groups clustering

Ward’s method

HC contd.


The result of HC displayed as branching tree
diagram called “Dendrogram”.
Pros and cons of HC


Easy to implement, quick visualization of data set.
Ignores negative associations between genes,
falls in category of greedy algorithms.
K-means Clustering


Divisive approach
Steps





Specify K-initial clusters and find their centroid.
For each data point the distance to each centroid
is calculated.
Each data point is assigned to its nearest centroid.
Centroids are shifted to the center of data points
assigned to it.
Steps 2-4 is iterated until centroid are not shifted
anymore.
K-means clustering contd.
x2
x1

Pros and Cons


No dendrogram
It is a powerful method if one has prior idea about
the no. of cluster, so it works well with PCA.
Future Work

It includes similar analysis on






Self Organizing Map (SOM)
Support Vector Machine (SVM)
Relevance Network
Gene Shaving
Self Organizing Tree Analysis (SOTA)
Cluster Affinity Search Technique (CAST)
Acknowledgements





Institute of Bioinformatics and Applied Biotechnology
(IBAB), Bangalore
Dr. Ashwini K Heerekar (Siri Technologies Pvt. Ltd,
Bangalore)
Dr. Jonnlagada Srinivas (Siri Technologies Pvt. Ltd,
Bangalore)
Mr. Kiran Kumar (Siri Technologies Pvt. Ltd,
Bangalore)
Mr. Mahantha Swamy MV. (Siri Technologies Pvt. Ltd,
Bangalore)
Selected references:



A Biologist Guide to Analysis of DNA Microarray
DATA, by Steen Knudsen
DNA Microarrays And Gene Expression from
experiment to data analysis and modeling, by P. Baldi
and G. Wesely
Papers:



Computational Analysis of Microarray Data by John Quackenbush, Nature
Genetics Review, June 2001, vol2.
The use and analysis of Microarray Data by Atul Butte, Nature Review drug
discovery, Dec 2002, vol1.
Microarray Data Normaliation and Transformation by John Quackenbush,
Nature Genetics.
Questions
???
Thank You