Biological Networks Gavin Conant 163B ASRC [email protected] 882-2931Types of Network • Regulatory • Protein-interaction • Metabolic • Signaling • Co-expressingGeneral principle Gene/protein/enzyme Relationship between genesExample • We can create a network where actors and actresses are the nodes • Two actors are joined if they co-stared in a film • Kevin Bacon gameExample • We can create a network where actors and actresses are the nodes • Two actors are joined if they co-stared in a film • Kevin Bacon game – On average, two actors can be linked by 3.65 films – Christopher Lee is actually more highly connected than Kevin Bacon Watts, D. J., and S. H. Strogatz. 1998. Collective dynamics of 'small-world' networks. Nature 393:440-442.Example—Protein interaction networks • Many or most proteins bind to other proteins – Generally not covalent – But can still be • Long term (multi-peptide enzyme complexes) • Short term (signal transduction) – Generically treat this as protein interactionsMethods to identify protein interactions • Lots of small-scale experiments • We will discuss the ―genome-scale‖ methods • Earliest: yeast two-hybrid – Uses yeast as a tool—can be done for proteins from any organism – But has been done at genome scale in yeast Lac promoter LacZ reporter gene Introduce a plasmid with a reporter gene: turns colonies blue if LacZ is on and a galactose-analog (―X-gal‖) is present TF TF Transcription • Originally, the transcription factor induced transcription of the gene and turned the colonies blueLac promoter LacZ reporter gene The different domains of the TF can function independently TF1 Transcription • TF now functions if the two pieces are ―close‖ • Note however that the two pieces do not ―stick together‖ on their own TF2 TF1 TF2Lac promoter LacZ reporter gene Tether each piece of the TF to another protein Transcription • We are testing if the bait and prey interact • If they do, they will bring the two halves of the TF close enough to turn on LacZ • If the colonies turn blue, the two proteins interaction TF1 Bait TF2 Prey TF1 Bait TF2 PreyYeast two-hybrid results • > 60% of the proteins in the yeast genome have been tested for some interactions • Question, if there are 5000 unique proteins in yeast, what is the possible number of interactions? • Problems with two-hybrid method – We’ve altered the proteins—false negatives and false positivesYeast two-hybrid results • Problems with two-hybrid method (cont.) – Probably won’t be able to sample all possible interactions – Doesn’t account for time of protein expression or location • We could infer that two proteins interact when they are in fact never at the same place at the same timeMass-spec. methods • Start by adding a tag to the ―bait‖ protein of interest: Bait gene TAP tag • Now grow cells with this construct: copies of the bait protein will be taggedInteractions vs. complexes • Two-hybrid methods find pairwise protein interactions • Here, we are looking at larger groups of proteins, aka complexes • Complex is a somewhat vague term – Length of residue? – Functional?Mass spec. continued • Extract the cell proteins without disturbing the protein complexes • From those proteins, extract any complexes with a member having the tag – Several steps – Uses an antibody to the tag for identification • Result is complexes that have the bait protein as a memberActual mass spec • We now have a group of complexes which the bait protein is a member of • What other proteins are present? Unfolded protein Proteolytic cleavage Many short peptide fragmentsSeparation of fragments • Magnetic field separates fragments based on: – Charge – Mass • Result is a list of ions with mass and change that are present MagnetIdentifying the ions 1. The mass spectrometer gives us a list of peptide fragment masses and charges 2. You search the genome for all possible peptide fragments and calculate their mass and charge 3. Match 1 to 2! – Obviously this is a computational challengeMass-spec protein complexes • We put the bait-centered protein complexes into the mass-spec and identify the peptides present • Look those up in the genome to identify the proteins present • This gives us a list of the proteins in the complex for that bait proteinGavin, A. C. et al., 2002. Functional organization of the yeast proteome by systematic analysis of protein complexes. Nature 415:141-147.In yeast • ~590 proteins used as bait • ~ 232 complexes found among these – ~130 proteins used as baits did not return a complex – But complexes can also interact with each otherBuilding networks • Protein interaction data has lots of uses • One is to study biological networks • Networks are an abstraction of the relationship among entities!Jeong, H., S. P. Mason, A.-L. Barabási, and Z. N. Oltvai. 2001. Lethality and centrality in protein networks. Nature 411:41-42.Network Measurements • Simple: Degree distribution – How many edges? – Expressed as a distributionNetwork Measurements • Simple: Degree distribution – How many edges? – Expressed as a distribution • Common degree distributions: – Normal: Mean number of edges is m – Exponential • P(x)≈ex • Power-law – P(x) ≈ xaComparing exponential and ―scale-free‖ networks • Exponential – P(x)≈ex – Most nodes have similar numbers of edges – Sensitive to node loss • Scale-free – P(x) ≈ xa – Few nodes with many edges (aka hubs) – Less sensitive to random node loss Albert, R., H. Jeong, and A. L. Barabasi. 2000. Nature 406:378-382.Other statistics • Mean path length • Longest path (aka diameter) • Clustering coefficient – Number of connections between your ―neighbors‖ over the possible number of connections • Number of componentsFeatures of protein interaction networks • ―Small world‖ • There are a few proteins with very many interactions (hubs) • Proteins are ―cliquish‖: – If you and I interact, and I interact with Susan – It is more likely you and Susan will interact • Most proteins ―talk‖ to every other protein in very few stepsUses of interaction networks • Predict disease-related genes – The density of protein interactions that a protein shares with other proteins can be used to predict whether it is likely to influence a particular disease • Follow
View Full Document