Michael Kuhn
Keynote: Aviv Regev on Modular Biology
Short bio on Aviv: http://www.broad.mit.edu/about.... (Much longer bio by the introductory speaker) - Shirley Wu
talking about TF network reconstruction network. outcome is still a network with not functional annotation. Important to reach a map that explains the biology - Pedro Beltrao
want to reconstruct biological function, but get something that we can comprehend (complex networks vs. "nice" pathway representations). - Michael Kuhn
unifying concept: functional models, sets of entities that perform together to perform an identifiable and distinct function - Michael Kuhn
Idea is to reconstruct how signals are processed and processes carried out so we can understand what is happening at every level of the pathway. But this often leads to nasty hairball diagrams (she shows example). Rather, we'd prefer a detailed diagram of the inputs, outputs, dynamics of processes that allow hypotheses to be made and tested (shows example) but these are painstakingly created by hand by biologists - Shirley Wu
caveat: modules we find from genetics/molecular interactions are not necessarily functional modules - Michael Kuhn
So, regulatory networks only today - Roland Krause
are modules true bio units ? are modules a useful concept ? can we identify the pathways controlling the modules ? - Pedro Beltrao
3 questions: 1) are modules true bio entities?, 2) is modularity useful for representing biology? 3) can we identify pathways controlling these modules? - Shirley Wu
Are modules true biological entities. ref: tanay, regev, shamir, PNAS 2005 - Michael Kuhn
Certain orthologous gene groups are co-expressed across different species. Not surprising. But are the regulatory mxns also conserved? Yes - in many cases, the TF binding motifs are conserved, as well as the TFs. BUT, in some cases, the proteins binding these sites are different across species. - Shirley Wu
examples of conserved and divergent transcription regulation of modules - Pedro Beltrao
Identify phylogenetic profile of cis-regulatory elements using sequence from many different species. Don't have expression data for all of them but assume that the orthologous genes are co-expressed bc they are co-exp in more distant species. - Shirley Wu
by mapping gene sets across different ascomycota between s. pombe and s. cerevisiae they could find the conservation or divergence of over-represented motifs across those species. - Pedro Beltrao
a very neat example of evolution of TF binding sites by a redundant intermediate with two binding sites. - Pedro Beltrao
Trying to determine whether changes in regulatory sites result from gradual evolution or "switching"? Use phylogenetic analysis of site sequences to examine evolution of site - Shirley Wu
Yep, this is great work even though the ribosome is probably one of the "easy" modules - Roland Krause
New regulatory sites "invading" the promoters of an ancestral site. Regulatory redundancy --> new site may push out ancestral site. - Shirley Wu
Open questions: Where do all the new binding sites come from? - Roland Krause
How do new sites are created ? is it positive selection or neutral drift ? - Pedro Beltrao
Dawn Thompson lead authors phylogeny of modules - Pedro Beltrao
13 species probed in carbon sources, data gathered: expression and LC-MS.MS metabolic data.... I would really like to have that data to use :) - Pedro Beltrao
Algorithm to infer a module phylogeny: MoPhy - Shirley Wu
Great, a keynote with novel data on parallel profiling of fermentation in 13 yeast species - Roland Krause
comparative expression in glucose depletion. divergence along evolutionary time occurs first for the dynamics of response - Pedro Beltrao
Moving on to second question: is modularity a useful perspective on system fxn? - using cancer as an example (lots and lots of data available, results already kind of modular - gene lists) - Shirley Wu
Building a module map of cancer (as opposed to just compiling gene lists). Have expression data and gene sets (Go categories, clusters, pathways) - Michael Kuhn
take gene sets and find repressed/induced gene sets - Michael Kuhn
Modules: core genes of (merged) gene sets. Can test modules vs. arrays - Michael Kuhn
Arranged gene sets with significant changes against arrays in a matrix. Clustered the gene sets. Then threw out genes that did not contribute to clustering. Merged the genes remaining in gene sets in each cluster to get modules. Picture of a Cancer Module Map - Shirley Wu
491 modules vs 94 conditions - Michael Kuhn
Example result - bone proliferation pathways showing up in primary breast cancer expression. Why? Pubmed search shows that many cancers, inc breast cancer, like to metastasize to bone. But this is primary, not metastatic tissue. Maybe, bone proliferation module turning on may foretell metastasis of the tumor? - Shirley Wu
Cancer and ES modules similarity - Roland Krause
ES signature could be used a signal for prognostic. what is the mechanism ? - Pedro Beltrao
Last part of talk: can we identify the pathways controlling modules? - Shirley Wu
From gene expression - assign genes to modules according to expression, infer regulatory pathway between modules that explains the expression patterns, throw away the gene assignments and re-assign, iterate until convergence. - Shirley Wu
It works! But why should it work? Many regulators are controlled post-transcriptionally. Turns out many of them are regulated post-transcriptionally by the targets they themselves control. - Shirley Wu
gene sets are important and useful. from my view it has to do with selection pressure. selection for function not for how the function is accomplished - Pedro Beltrao
Models don't take into account Time (also not space:) , at different time scales (from environmental time to evolution time) - Pedro Beltrao
future outlook: look for principles and tools to look at time at different scales. - Pedro Beltrao
Four nice keynotes so far but this was probably the best - Roland Krause
yes, very cool talk - Pedro Beltrao
Go women in science! So far Claire and Aviv's talks have been my favorites. - Shirley Wu