evaluating machine learning approaches for aiding probe.pptVIP

Machine Learning and Genetic Microarrays Jude Shavlik David Page University of Wisconsin-Madison Goals Learn about microarray technology See some ML problem formulationsin the computational-biology literature A little on experimental pitfalls to avoid Overviews of newest “high-throughput” molecular-level data being gathered Very little on the details of machine learning algorithms Outline Molecular Biology and Microtechnology Machine Learning Applications Technological: Designing Microarrays Medical: Predicting Disease (Diagnosis, Prognosis, Treatment) Biological: Constructing Pathway Models Looking Ahead: Related Technologies Outline Molecular Biology and Nanotechnology Machine Learning Applications Technological: Designing Microarrays Medical: Predicting Disease (Diagnosis,  Prognosis, Treatment) Biological: Constructing Pathway Models Looking Ahead: Related Technologies Combining Three “Hot” Technologies Information Technology Biotechnology “Nanotechnology” The “Central Dogma” of Mol Bio The Big Picture We’d like to know which proteins are present in a given type of cell, under some conditions, etc eg, cancerous vs. non-cancerous cells However, currently can only measure RNA well Microarrays (“Gene Chips”) Specific probes synthesized atknown spot on chip’s surface Probes complementary to RNA of genes to be measured Typical gene (1kb+) MUCH longer than typical probe (24 bases) Microarray Technologies Alternate technologies exist(eg, “spotted arrays”) We won’t cover them all We’ll focus on the “market leader”(Affymetrix) How Microarrays Work Example: NimbleGen, Inc Micro Mirrors from TI From Probes Back to Genes Need algorithm for converting measured probe intensities into gene-expression levels Could simply use average probe value More (too?) complicated approaches exist (eg, Affymetrix’) Cleaning Up the Data – “Controlling Variance” Often look at relative expression levels Measurement(cancerCell) / Measurement(normalCell)