Transcript odedMagger
Bioinformatics: Cool stuff you can do with Computers and Biology Oded Magger Tel Aviv University / Autodesk inc. GIP course 2010 When two fields come together • Computer science has applications in many fields (economics, physics, history, you name it.) • The field of biology is being transformed: • Tons of biological data. • “Human genome project” • “Large scale experiments” • Complex computations and algorithms. • Biologists who know how to send an Email. 2 Genetics 101 3 Genetics 101 DNA mRNA 4 Protein Challenge examples • Find a drug for a disease. – Infer its side effects. – Understand how it influences the body. • Find out what causes genetic disease. • Figure out what a gene actually does. • Decipher the secrets of evolution!!!11 – How similar are two genes? – Construct the tree of life. • Medical image analysis ( = “sir, I’m afraid you have cancer”). 5 Case study #1 PRINCE: Associating genes and protein complexes with hereditary disease Background • Objective: to correctly predict the molecular causing factors of hereditary disease. • Causal genes • How are causal genes identified today? • Association studies. • Prioritizing genes in an interval. Done computationally using sequence similarity, functional similarity, network data and more. • Important insight: proteins causing diseases with similar phenotype tend to lie close in PPI network. 7 PPI – “Protein facebook” 8 From a problem to an algorithm • Important insight: proteins causing similar diseases tend to lie close in PPI network. • If many of your friends or friends of friends in Facebook study computer science, chances are that so are you. • Similar diseases – Natural Language Processing (the computer reads the diseases encyclopedia). 9 From a problem to an algorithm • If related parts of some car break down, the ‘symptoms’ will be similar. 10 How does PRINCE work? Interval 10P p1 p2 p3 p4 d1 p6 p5 d2 p8 p7 d4 p10 p11 0.1 q 0.3 d3 p9 0.7 d5 0.1 0.9 Strengths of PRINCE • A network based method. • Fast, propagation-based method converged to accurate matrix-based solution. • Global inference: • Inference not limited to the direct vicinity of genes for which prior knowledge exists. • Smooth function over the network. • Smart normalizations: • Logistic transformation of disease similarity metric. • Edges between two high-degree nodes have 12 lowered weight. PRINCE beats the competition! 13 Case study #2 Metabolic models: Simulation of life! (Parts stolen from Tomer Shlomi and Eytan Ruppin) Metabolism • The body is a huge factory for assembling and disassembling molecules ( = stuff). • Factory workers: special proteins called enzymes. • A single transformation of one set of material to another is called “reaction”. A reaction has a rate. (“One worker can turn 2 cows to 50 steaks in one 15 hour.”) Metabolic network 16 Metabolic simulation! • Start with a set of molecules and their amounts, a set of enzymes and their amount – and press “play”… • If you want details – wait until your third year… • Relies on material from the Linear Algebra and Algorithms courses. 17 What is it good for? • Medicine: • Diagnosis. • Metabolic diseases (any diabetics in the crowd?) • New cures for cancer. • Biotechnology • Genetically engineered bacteria and yeast manufacture products of interest with high efficiency (Insulin, bio-fuels, beer). • Easier to run experiments on a computer than in 18 the lab (1 computer = 1000 lazy biochemists). Questions? 19 Thank you for listening! 20