Comparing cancer models using gene expression of genetic pathways - - PowerPoint PPT Presentation

Comparing cancer models using gene expression of genetic pathways and other gene lists

Tauno Metsalu Data mining research seminar 17.12.2012

Introduction

• Before testing a drug on human, it has to

be tested and effective on a simpler model

• Most potential drugs lack the expected

efficacy on human, even when working properly on a model

• It is important to understand the exact

aspects how models could be improved

Drug development pipeline

Model systems

Cell line as a model system

• Cheaper
• Stays alive for longer
• Less variable
• Differs from real cancer!
• How much and in which sense does it

differ?

Collected expression data

Microarrays (1)

Microarrays (2)

Gene lists

• We use two different ways to define

gene lists

• Pathways – groups of genes with a

specific biological function (KEGG database)

• Interaction partners – we take all

interaction partners of a specified gene (FunCoup database)

Similarity

• Absolute similarity – Spearman’s

correlation converted into 0-1 scale

• Relative similarity – relative rank of

absolute similarity among all gene lists (used just for better intuition)

Data integration

mouse samples mouse pathway genes human samples human pathway genes InParanoid ortholog mapping pairwise comparisons green: data used

Analysis

• We used two different genes as basis:

EGFR and MTOR

• They have enough data for both gene

list definitions (pathways and interaction partners)

EGFR

MTOR

Summary

• We proposed a pipeline to compare

cancer models

• It is possible to integrate data from

different platforms and species (human and mouse)

• MTOR is better retained in cancer

models than EGFR

Acknowledgements

• Prof Jaak Vilo
• Priit Adler
• Raivo Kolde