 |
|
| FriendFeed is the easiest way to share online. Learn more » |
ISMB 2008:
Pedro Beltrao
Interaction Networks and Disease Room 701A - http://www.iscb.org/ismb200...
July 21, 2008
- Comment
- Like
- Share
Hsuan-Cheng Huang, Euan, Thomas Lemberger and 3 other people liked this
An excellent line up
- Roland Krause
Mark Gerstein on protein function (from networks)
- Roland Krause
Human variation (SNP and CNV) in networks
- Roland Krause
As a side note Mark Gerstein has a few "life streams" (http://gerstein.info/streams/)
- Pedro Beltrao
positive selection largely takes place at the network periphery (in the human interactome), Nielsen et al. PLoS Biol 2005
- Michael Kuhn
Use human variation data (SNPs and copy number variation) mapped to human interaction networked. Discussing the differences between adaptive and neutral evolution. Looked for proteins under positive selection in the interactome using sequence analysis.proteins under positive selection were found at the periphery of the networks
- Pedro Beltrao
poorly connected proteins (on the periphery) are also closer to the outside of the cell. I wonder if this is not a technical problem of finding interactions for membrane proteins for example.
- Pedro Beltrao
Some of work on networks he was mentioning was done by P. Kim (http://homes.gersteinlab.org/people...)
- Pedro Beltrao
Adaptive evolution on the periphery, example CD4 and HIV.
- Roland Krause
Well, there is also the problem of biased gene conversion which inflates the number of proteins appearing to be under positive selection. Would be interesting to explore this critically.
- Roland Krause
but the cd4 example is clearly not influenced by that
- Roland Krause
On CD4 protein, surface residues that interact with HIV gp120 are under positive selection
- Michael Kuhn
Can use structures to find mutually exclusive protein-protein interactions (Kim et al. Science 2006)
- Michael Kuhn
Hmm, Mark Gerstein mostly talked about previously published results... HIV being the only disease-related topic he mentioned
- Michael Kuhn
:) Acknowledgments: Co-author network
- Michael Kuhn
This session is packed.
- Roland Krause
But the CD4 stuff is at least relevant, not just some OMIM mapping without a cause
- Roland Krause
Next up: Haiyuan Yu, former grad student in Mark Gersteins lab, now post-doc with Marc Vidal
- Michael Kuhn
Haiyuan Yu talking about "High Quality Binary Protein Interaction Map Reveals Properties of Yeast Interactome Network"
- Pedro Beltrao
There we go again: Christian von Mering's infamous "Y2H isn't reliable" figure :)
- Michael Kuhn
yeast two hybrid evangelism :) Ito non core is very noisy should not be used for analysis. (the net is very slow today)
- Pedro Beltrao
RTFP, we do say that Ito core is superior to the whole data set, but its only ~850 interactons.
- Roland Krause
To compare literature curated interaction with Y2H they used only the literature curated that are supported by one report. He says that that is the fair comparative test. Using different methods to compare Y2H with literature curated interactions
- Pedro Beltrao
interaction re-tested by Y2H- core Y2H from previous tests retest very well. The literature curation re-test at similar or even lower quality. pull down interaction re-test worse than Y2H but he stresses that the interaction types are different and that the tests that they are using Y2H and PCA detect binary interactions.
- Pedro Beltrao
Its fractions, not absolute numbers and only the specificity, not sensitivity, so...
- Roland Krause
He suggests that MIPS or other complex interactions data can not be used directly to benchmark binary interactions. He defines a set of high quality binary interactions for benchmarking.
- Pedro Beltrao
re-mapped yeast interaction network repeated 3X. 2930 interactions among 2000 proteins (combining also previous interactions)
- Pedro Beltrao
Its a good map, just very incomplete and its quite hard to envisage how to close it.
- Roland Krause
ongoing project: look for mutants that disrupt particular interactions.
- Pedro Beltrao
Again, a talk that didn't address disease at all even though Marc Vidal is doing some work on this (Yıldırım et al. Drug-target network. Nature Biotechnology (2007) vol. 25 (10) pp. 1119-26)
- Michael Kuhn
I remember from another talk that they were mentioning a maximum coverage of Y2H of 20% but I am not sure. They were saying that basically we will need many different methods to get higher coverage.
- Pedro Beltrao
The problem is just that all interaction screens provide different aspects - binary (y2h), complexes (apms), functional interaction (genetic) - and just throwing them together won't cut it.
- Roland Krause
Frederick P. Roth - Mouse Functional Linkage Graphs and Complex Disease
- Pedro Beltrao
Fritz Roth: Functional prediction in mouse as a competition
- Roland Krause
He is talking about MouseFunc (http://hugheslab.med.utoronto.ca/supplem...) predicting function for mouse
- Pedro Beltrao
Precision of functional association (36 top novel predictions) precision between 72 to 83%. Evaluated by other lab.
- Pedro Beltrao
There are a number of real improvements in this work, the limitation to 31-100 GO term specificity in particular
- Roland Krause
I wish the speakers would go more into detail, it's a special session after all
- Roland Krause
Next one! Trey Ideker
- Michael Kuhn
plug for Cyotscape plug-ins, most popular: Bingo w/ 5000 downloads in last year
- Michael Kuhn
One last though about Roth's talk. He was another researcher mentioning the need to understand how many different mutations together might cause disease.
- Pedro Beltrao
based on pathway assembly, study disease using networks: network-based disease diagnosis; functional separation of disease gene families; moving from GWAS to network-wide "pathways" association (PAS)
- Michael Kuhn
Is it really necessary to advertise Cytoscape? Sounds like advertising Blast to me.
- Roland Krause
General description of his Trey Ideker's lab. How to combine different interaction networks and to study them to understand evolution, function and disease. First work on networks and disease, combining gene expression profiles and protein interactions.
- Pedro Beltrao
Tuesday 2:45 , 718B talk by members of his lab on dissection function in large transcription networks. All of the disease work was referenced for other talks. He is going to talk about the position effects of transcription factor binding sites in the sub-telomere
- Pedro Beltrao
position effects in the genome: order matters, even in eukaryotes; look at distance from nearest telomere in yeast. transcription factors have particular binding profile, bimodal distribution of telomere distances
- Michael Kuhn
Some TFs bind in a bi model distribution along the genome. Some of them bind the sub-telomere in a condition dependent fashion
- Pedro Beltrao
He said these distance effects were unpublished and not talked about before, but this was also shown in Barcelona in April or May
- Michael Kuhn
He suggested in a previous talk that this could be due to physical localization of transcription withing the nucleus. More intuitive could be that it is some sort of condition dependent sequestration of TFs
- Pedro Beltrao
condition-specific behavior: changes in environmental tradition sometimes leads to appearance of another telemore distance mode
- Michael Kuhn
Lat talk: Nevan Krogan http://kroganlab.ucsf.edu/ (disclaimer, I am biased :)
- Pedro Beltrao
Last talk of the session: Nevan Krogan
- Michael Kuhn
how to combine protein and genetic interaction networks. Starting by introducing quantitative genetic interactions.
- Pedro Beltrao
epistatic interactions measured by colony size - EMAPs a good methods papers on this is http://genomebiology.com/2006...
- Pedro Beltrao
using correlation of genetic interactions between gene pairs to predict complex and functional association.
- Pedro Beltrao
Genetic Map of 743 Genes Involved in Chromosome Biology; Collins et al, Nature, 2007. Pairs of genes coding for physically interacting proteins often have similar genetic ixn profiles and positive genetic ixns
- Michael Kuhn
Talking about Assem's work on S. pombe genetic interaction map on DNA machinery
- Pedro Beltrao
550 S.pombe genetic interaction map that is comparable (orthologous as well) to the one in S. cerevisiae
- Pedro Beltrao
He is talking about the RNAi cluster found within the genetic interaction network. A new component of RNAi machinery was found in the map. Follow up experiments confirm the result
- Pedro Beltrao
how conserve are genetic interaction between S. cerevisiae and S. pombe ?
- Pedro Beltrao
the conservation of genetic interactions (S scores) between pairs of proteins is very low R=0.14 but higher if they interact (defined by co-complex).
- Pedro Beltrao
conservation is found at the level of modules (complexes) but not between them
- Pedro Beltrao
Best conservation at the complex level - but complexes are often in the nucleus and the nucleus is best conserved. Not a contradiction to what is presented but softens the generality, I would think.
- Roland Krause
genetic interactions within modules not conserved
- Roland Krause
Bonus talk! Frank Alber, USC: Determining the subunit architectures of macromolecular assemblies. [Interaction networks? Disease? Hello?]
- Michael Kuhn
Frank Alber on subunit architectures of complexes
- Roland Krause
Frank Alber talk on determining big macromolecular assemblies structures by integrating different structural data.
- Pedro Beltrao
Introduction to nuclear pore complex. used developed computational approach and experimental data: copy numbers, sedimentation coefficients, imuno-electron microscopy (15 to 20nm resolution), shape from electron microscopy, affinity purification for proximity information
- Pedro Beltrao
used all the constraints to build optimized models of protein localization probability within the complex and used known structures to fill the structure. Nice pictures and movie of the whole complex
- Pedro Beltrao
Stunning display of the 450 proteins in the nuclear pore complex. Amazing work, really.
- Roland Krause