Health and CS Philip Chan DNA Genes Proteins

Health and CS Philip Chan

DNA, Genes, Proteins n What is the relationship among n DNA n Genes n Proteins ?

DNA, Genes, Proteins n What is the relationship among n DNA n Genes n Proteins ? n Some DNA regions are called genes n Which are blueprints for making proteins

Gene expression n How “active” the gene is n Measuring gene expression can help characterize diseases

Cancer Subtypes n Why do we want to find subtypes? n For each cancer patient n We measure gene expression n How can we find out cancer subtypes?

Problem Formulation n Input n Expression levels (values) of each gene n n Multiple patients Number of subtypes (clusters) n Output n Cancer subtypes (clusters)

Clustering n Ideas?

Clusters (Subtypes) n Clusters n Similar within a cluster n Different across clusters n We need to define distance (similarity) n Between two patients in terms of gene expression

Distance Function n a and b: two patients n ai and bi: expression level of gene i

Distance Function n

K-means Clustering Algorithm 1. Pick k random patients as centroids 2. Assign each patient to the cluster with the closest centroid 3. Repeat Calculate the centroid for each cluster b. Assign each patient to the cluster with the closest centroid Until no changes in cluster membership a.

Calculating Centroid n Let centroidi n the expression of gene i of the centroid n centroidi = avg. expression of gene i in the cluster

Animation n http: //shabal. in/visuals/kmeans/1. html