Sign in or Join FriendFeed
FriendFeed is the easiest way to share online. Learn more »
Cytoscape Symposium 2010

Cytoscape Symposium 2010

An unofficial group for the Cytoscape Symposium and Retreat 2010, Ann Arbor Michigan, USA
Morgan
After the break: Trey Ideker leading the discussion on RAVIN P41 #cytoscape
Cytoscape applied for a grant and is now being elevated to the level of an NIH resource. - Morgan
Some positions are being created as a result that weren't possible before. Three new positions: - Morgan
1. Program coordinator/executive director (0.5 FTE)- responsible for reporting to NCRR, administration, anything else that needs doing, facilitate software development process, service and collaborations with resource users, track publications, track technology project milestones - Morgan
2. Outreach coordinator (>1 FTE): Documents software and developer practices, design Cytoscape tutorials, present Cytoscape tutorials at conferences, roving engineer who works with plugin coordinators, formally collecting user feedback, discussion lists, help desk, Facebook, Twitter, wiki, Google Summer of Code. Difference between developer and user outreach. - Morgan
3. Cytoscape architect: oversee the overall design of the codebase, best practices, coordinating the code that comes into the core, must have software development experience, but the position may not require coding, develop software development best practices - Morgan
These three positions are at the RAVIN level and are intended to coordinate with individual centers/project managers at the different locations. - Morgan
Perhaps an alternate role you may want to consider is "Analyst" - a person primarily reposnsible for scoping out requirements and setting priorities on it. - Ami Khandeshi
Morgan
Cytoscape 2.8 Overview and Schedule (Mike Smoot) #cytoscape
Custom graphics: overlay protein structures, glossy overlays, Google charts, etca - Morgan
Equations for attribute browser: like excel (eg. =ABS($gal4R)), can only reference attributes from the same node at the moment. Either right-click on the column or use equation editor. - Morgan
Improved CyGroups - Morgan
Search for plugins - Morgan
Bug fixes! - Morgan
Morgan
Gary Bader leading the discussion on #cytoscape UI issues
Some users can't distinguish between tabs on the attribute browser, or don't see any visible results - Morgan
When importing attributes, there is no visible feedback - Morgan
VizMapper doesn't always update with attributes - Morgan
Suggestion: select-all button for attributes - Morgan
Need to click directly on nodes and edges to interact with them and this can be difficult when things are small - Morgan
Control panels have a minimum size. Dynamic hiding might be good. - Morgan
Size of the window should be stored in the Cytoscape program instead of the session. - Morgan
Hiding attributes from the user that are used solely for vizMapping (upcoming feature?) - Morgan
Setting default file opening directory during installation or when loading files - Morgan
Setting memory levels in the UI may not be practical, but it's easier in 2.7 - Morgan
Better plugin classification - Morgan
How about a (small, lower right corner) memory usage bar. To give a sense of how bad things are. - Terry Weymouth
Nielsen's heuristic evaluation guidelines for any UI - Usability engineering guidelines http://www.useit.com/papers... - Morgan
I think the core idea of this session is that the core developers are very happy to hear any feedback or feature tweaks or suggestions and encourage all users to report on problems they have. - Morgan
Morgan
#Cytoscape Core Developer's Hackathon today: join the discussion!
Mike Smoot leading the discussion today. We're making plans today on where to be a year from now - Morgan
Didn't hear what Trey said. Gary Bader mentioned that some issues came up in the Developer's workshop relating to the UI. - Morgan
As discussed in the mini-workshop, Cytoscape 3.0 will use OSGi/Spring and Maven, and a development schedule based around modules - Morgan
Javadocs, stable API, code reviews, and unit testing for 3.0 - Morgan
Cytoscape 3.0 is already functional, the modules for round one should extend the basic functionality, and be ready around Halloween - Morgan
OSGi is a framework that allows for easy modularization of the code - Jochen Weile
Trey Ideker: what parts of the modules depend on OSGi/Spring? Many people are struggling with the large spec and W3C people have warned him not to go down that road. - Morgan
Part of the design would have to re-written if OSGi/Spring were dropped. - Morgan
In the mini-workshop, a shim to hide the plugin developer from OSGi was decided on. - Morgan
Can't see the whiteboard! - Morgan
Plugins that depend on other plugins need to have the same versions of dependencies, but in general OSGi should isolate plugins from each other using different ClassLoaders. - Morgan
Differences between 2.- and 3.-: Node ids will be classes, CyDataTables instead of CyAttributes, no root graph, no Giny. True hierarchical networks, independent networks. - Morgan
A few years ago, plugin developers agreed that there should be a clean break so that minor versions don't break plugins. - Morgan
OSGi also allows hiding of Cytoscape internal implementation details so that they are not part of the public API. - Morgan
Problem with OSGi is that it might be too difficult to maintain? - Morgan
My opinion: code re-use, while having an initial learning curve, makes maintainance easier in the long run, because any bugs can be re-directed to OSGi developers rather than Cytoscape developers. - Morgan
Cytoscape shim: allows developers to deal only with basic Java, and should be easy to maintain - Morgan
CyShim interface will have all of the factory methods necessary to deal with Cytoscape. - Morgan
Any opinions out there on Ant vs. Maven? In my opinion, I think enough IDEs can use Maven invisibly that it doesn't need to be shimmed. - Morgan
I use both ant and maven, often in the same project. Maven is unsurpassed when it comes to "standard" builds and dependency management. Ant is way, way easier for hacking together custom initialization and configuration scripts. - Terry Weymouth
Maven is in many ways more comfortable to use than Ant and provides many more features. On the other hand it has its own difficulties. It's version tracking for example is a minefield. - Jochen Weile
But with Maven you don't need anything in the POM other than the name of the project, and it can figure out how to build it without you needing to define a build target. - Morgan
I don't think that Ant is any better at that though. - Morgan
Nice example of a new 3.0 "Homer" plugin using the shim with one (!) dependency. - Morgan
Code completion in Eclipse or other IDE should be available with the compiled jar file and a javadoc jar. - Morgan
Scooter Morris likes it! - Morgan
If people want to use the full features of OSGi, they can easily peel back the abstraction layer and access the full functionality. - Jochen Weile
So if an plugin developer wants to use OSGi, would they just have to implement that CyShim interface with their own OSGi/Spring configuration? - Morgan
The shim will grow with each completed module. - Morgan
The shim can probably be re-named to something like "PluginAdapter". - Morgan
We're all curious to see what a plugin would look like without the shim. There will be website-based tutorials and Maven archetypes that will build the structure for you. - Morgan
Coffee break! - Morgan
Gary Bader
Keiichiro Ono
Cytoscape Retreat Day 3: Developer Hackathon (Broadcasting live at http://www.ustream.tv/channel...) #cytoscape
Morgan
Stephen Friend, M.D., Ph.D, "It is not what we do but how we do it." #cytoscape
Why aren't we able to identify effective therapeutic agents? - Morgan
The problem is that pathway maps are not sufficient. - Morgan
"Our maps are not fitting reality and we need to acknowledge that fact." - Jochen Weile
Bottom-up systems biology models (equations, etc.) are not what are needed. - Morgan
No proteome, genome, transcriptome, RNAome is going to give you a picture of causality. - Morgan
What is required for a model of causality? Co-expression networks are only descriptive. Bayesian networks CAN be used, but carefully. - Morgan
Sage bionetworks: non-profit association in several locations to train and examine this top-down modelling concept - Morgan
Slides are available from the Sage website: http://www.sagebase.org/trainin... - Morgan
We live in a hunter-gatherer society where scientists acquire data, and then it is "theirs". - Morgan
Data mining: my data's mine, your data's mine. This is slowing science down! - Morgan
Need a fundamental shift in research in order to support the movement towards P4 medicine. - Morgan
Great talk. - Morgan
Jochen Weile
Brian Athey, PhD, U Michigan: "NIH National biomedical computing networks to enable network biology" #cytoscape
Follows a similar line to Gil Omenn's talk this morning in terms of biology as an information science - Morgan
Cytoscape sits at the center of data analysis - Jochen Weile
"The demise ofthe expert-based practice is inevitable" Human cognitive capacity remains steady while the amount of information required to make decisions is constantly increasing. - Morgan
Many different associations have different methods of making sense of knowledge and have different names for the same thing, and the existing infrastructure in the USA for moving research from the bench to the bedside. - Morgan
55 members of the CTSA National Consortium for translational science. They are dispersed across the USA and have 6 strategic goals relating to moving from basic science to patient treatment - Morgan
Jochen Weile
Peter Zandstra, Ph.D., U Toronto: "Cell-cell interaction networks" #cytoscape
Cell interaction networks have been used for blood cells, development, and also stem cell biology - Morgan
There are low numbers of haematopoetic stem cells available. Growing stem cells is difficult, with at most 2-5 doublings. We know it's possible in vivo, but not yet in vitro. - Morgan
transplanting blood stem cells always results in less stem cells than normal, regardless of the number transplanted. Traditional argument is that there are limited niches for the stem cells, but they believe it has to do with feedback from more mature cells. - Morgan
Designing new, artificial niches for stem cells to grow in order to regulate their availability - Jochen Weile
"Model predicts competitive antagonism between primitive and mature cell output." - Morgan
Jochen Weile
Eric Neuman, PhD, Clinical Semantics Group, Cambridge, MA: "Visual analysis of the web of linked data", #cytoscape
The S*QL plugin for Cytoscape - Jochen Weile
Bolt & Newman: Built first Arpanet router, also created Genbank - Jochen Weile
Also work on visual analytics: Kinetic objects for finding relations beyond the characteristics of individual elements - Jochen Weile
"Biomedical innovation depends on data that is highly connected and application independent." - Morgan
All software should be able to read any data - we should never need to write another parser! - Morgan
Moving to a data-centric collaborative space: Use URIs as data entries. All software should be able to read any data! Data reference instead of data exchange! - Jochen Weile
RDF naturally flows into networks - Morgan
The LOD cloud: A network of data providers forming a semantic web of 14 billion triples - Jochen Weile
Linked Open Data movement - all databases and tools have a SPARQL endpoint - Morgan
S*QL plugin allows to use both SPARQL and SQL to create graphs and specify visualisation details and extend or filter the contents - Jochen Weile
The rest of the presentation consists of many examples of the power of S*QL queries and Cytoscape to integrate data from multiple sources and visualise them - Morgan
Exploring Disease networks in Cytoscape using the SPARQL plugin - Jochen Weile
Jochen Weile
Edward Rietman, PhD. Dana Farber Cancer Institute, Boston: "Observation of Spatial PPI Dynamics in Arabidopsis root" #cytoscape
The interactome of an organism means the complete set of molecular interactions between proteins, metabolites, DNA, RNA. This network is vastly incomplete for any known organism - Jochen Weile
PPI network covers 5-10% of all possible interactions - Jochen Weile
PPI networks can be generated by experiments (y2h, ap/ms), prediction or literature curation - Jochen Weile
Prediction efforts so far have not been very successful - Jochen Weile
With enough testing, Y2H high-throughput methods can be reliable enough to work with - Jochen Weile
CCSB has published several PPI networks for human, C. elegans and others - Jochen Weile
Now working on publishing the complete Arabidopsis PPI - Jochen Weile
Generated several function-specific subnetworks - Jochen Weile
Network-wide expression analysis over time in root development - Morgan
Mapping change in expression to location in PPI network: Series of hairball visualizations - Jochen Weile
Can use neural networks to learn from the time-series of networks - Morgan
Jochen Weile
Mark Newman, PhD., U Michigan: "Social networks and the spread of disease" #cytoscape
Traditional approaches: Divide population into compartments, assume random mixing - Jochen Weile
Study of social networks is older than facebook: dates back a hundred years - Jochen Weile
Visualising social networks mostly results in hairball pictures, that are visually stunning, scientifically worthless and published in Science or Nature ;) - Jochen Weile
So we need a way to understand these networks, to quantify them. One crucial property: Node degree (number of friends) - Jochen Weile
Hub nodes are very likely to get a disease (because of their many contacts), but also very likely to pass it on. - Jochen Weile
Most people only have very low number of partners, but some few people have large numbers of contacts, which matches with scale-free property of the network - Jochen Weile
Example SARS 2003: The first person to contract SARS was a social hub. You can follow the spread of the desease from hub to hub. Each one give the disease a big boost. The media called them "super-spreaders" - Jochen Weile
Study on pneumonia spreading in a hospital. Patients are confied to wards, but caretakers move between them, acting as hubs. - Jochen Weile
Fed example case into computer simulation of wards and caretakers. Probability of caretaker to get infected in a ward is relatively low, but the if he gets infected he is very likely to affect all other patients in all served wards. - Jochen Weile
This means that to control the disease, antibiotics should be given to the caretakers (even if they are not at risk) - Jochen Weile
Simulated contact network with households, schools, hospitals, shopping malls and work places - Jochen Weile
Simulation shows that a disease turning into an epidemic depends on it transmission probability: Over a certain threshold, the spreading explodes - Jochen Weile
Another important factor is how quickly a disease reaches a hub. If the first patients are hubs, the disease is likely to explode, too - Jochen Weile
Detailed simulation of a whole city, with building locations, daily plans for citizens and computer agents for every one of them: Can be used to simulate many circumstances of diseases - Jochen Weile
More difficult for sexually transmitted diseases: Networks change continuously - Jochen Weile
Survey: median length of a partnership; 92 days, 38% had 5 partnerships in their life. 15% had 20 or more partnerships, median gap between partnership: 121 days. Fraction of partnership that overlap: 25% Median days of overlap: 427 days - Jochen Weile
Strong correlation between age and length of partnership, as can be expected - Jochen Weile
Air travel is the primary vector for global disease spread - Jochen Weile
Prediction level of swine flu based on airline networks were extraordinarily accurate. - Jochen Weile
Morgan
Charlie Boone, PhD. U Toronto: "The genetic landscape of the cell" #cytoscape workshop
Genetic interactions in yeast. (Unexpected phenotypes as a result of double mutants) - Jochen Weile
5000 of the 6000 genes in yeast can be deleted and still result in a viable phenotype - Jochen Weile
Genetic interaction: two gene alleles combine to generate an unexpected phenotype - Morgan
Synthetic lethality: two gene deletions with no effect individually are non-viable together - Morgan
Genes that show the same pattern of genetic interactions often belong to the same pathway. - Jochen Weile
Mate yeast strain, drive them through meiosis, and then select from the resulting progeny for the double mutant using antibiotics - Morgan
One gene can have many gene interactions, ~ 50. Network is obviously very complex. - Morgan
Make the network less complex with hierarchical clustering - Morgan
Measure quantitiative genetic interactions rather than qualitative: Looking at the fitness of the double mutatants compared to the single mutants - Jochen Weile
Several layers of normalisation methods are necessary to get accurate quantification - Jochen Weile
Constructing a correlation-based network from genetic interactions - Jochen Weile
Create an edge between genes that correlate well in the genetic interaction space and annotate the edge with the correlation coefficient - Jochen Weile
Correlation network shows nice clusters for various biological processes - Jochen Weile
Resulted in identification of many new members for several pathways - Jochen Weile
Mating two very similar yeast strains (2% difference, similar to that between humans), looking for unconditional and conditional essential genes. - Morgan
Most genotype-phenotype relationships are defined by the interplay of more than four genes. Double-mutants can only shed light on a small subset of relationships - Jochen Weile
Jochen Weile
Lee Hood, MD, PhD. Institute for Systems Biology, Seattle: "Systems Medicine: The Transformation from Reactive to Proactive P4 Medicine" #cytoscape
(Lee Hood is considered the inventor of modern day sequencing technology) - Jochen Weile
P4 medicine is the medicine of the future (or present!), using systems approaches to revolutionalize the way medicine deals with patients. - Morgan
P4 medicine: Predictive, Preventive, Personalized, and Participatory - Jochen Weile
P4 medicine is the medicine of the future. It applies the systems approach to medicine - Jochen Weile
Two types of information: genomic and everything else. Two ways to connect these: biological networks and molecular machines - Morgan
P4: Hypothesis-driven and hypothesis-generating, Bases on global data acquisition - Jochen Weile
Need to be able to collect all of the information necessary, integrate it all together, and then bring it all together in networks and models that generate and drive hypotheses in systems biology - Morgan
The Institute for Systems biology takes has highest impact of papers in the US, and 3rd highest in the world - Jochen Weile
Systems biology approach to prion disease in mice. Had to investigate multiple strains in order to reduce the noise inherent in the system (subtractive biology). Reduced the potential genes from over 7000 to 333. - Morgan
Biology needs to be the driver in systems biology. Only deep understanding of underlying biology can help to draw the correct conclusions - Jochen Weile
Discovered many differentially expressed genes encoding known and novel prion disease phenotypes - Jochen Weile
Using systems apprach for blood analysis: Blood as the carrier of many organ-specific fingerprints - Jochen Weile
Every organ releases a plethora of specific proteins into the blood stream, which can be analysed - Jochen Weile
Fingerprint for brain is 110 proteins; liver is 80. - Morgan
Allows early detection methods and easy following of disease progression - Jochen Weile
New projects at ISB: Human Genome Project for the whole family, Human proteome project, P4 medicine for patient assays - Morgan
2nd generation human genome project: focus on families, comparative approach - Jochen Weile
Family-based sequencing allows for using Mendelian genetics to identify over 70% of sequencing errors - Jochen Weile
Family of four, children with Miller Syndrome and lung disease. Found rare 230 000 variants in family members. - Morgan
Chromosomal overlay for children showing crossover and allele inheritance from parents. Very shiny. - Morgan
Each child carries about 30 new mutations. - Jochen Weile
Were able to drill down to identify the four genes that were major contributors to the two diseases in the children. - Morgan
Will put up family genome information into cloud-based storage (sans the personal information) - Jochen Weile
Human genome project transformed biology and gave a huge boost to systems biology and bioinformatics - Jochen Weile
Human Proteome Project: parallel to Human Genome Project, but there are more dimensionalities to proteins compared to genes. "genome is digial; proteome is analog" - Morgan
Jim Heath is developing a mIcrofluidic protein chip: assay 2500 organ-specific proteins from a drop of blood. - Morgan
Single cell analysis: one cell line falls into different clusters based on their transcriptome. What are they doing differently from each other? - Morgan
Isolate iPS stemp cells from blood and differentiate to different tissues to get important phenotypic information. - Morgan
Obtain iPS cells from families with Alzheimer history, differentiate the stem cells into neurons and analyse their heterogeneity - Jochen Weile
P4: predictive - use DNA, health history, blood analysis, molecular imaging to assess probabilistic wellness - Morgan
Personalized: unique human variations means each patient must be his own control and requires the individualized approach - Morgan
Preventitive: design preventative drugs, transition to wellness assessment - Morgan
Participatory: patient is involved in medical choices - Morgan
Q: What will be the repercussions for ethics and society. A: In the future, many methods that are considered as ethically problematic today, will be accepted as logical. - Jochen Weile
Jochen Weile
Welcome speech by Gilbert S. Omenn, MD, PhD, Center for Computational Medicine and Bioinformatics, U Michigan #cytoscape
Systems biology is about discovering what makes living organisms tick - Morgan
Vision of biology as an information science: Masses of genomic information, Powerful computational methods - Jochen Weile
Multi-layer challenge moving from phenotype to genotype, and also many aspects of phenotype - Morgan
Anyone who wants it can have a tour of the building! :) - Morgan
Jochen Weile
It should be an exciting day today at the Cytoscape Symposium! Speakers of the first session will be Ora Pescovitz and Gil Omenn (U Michigan), Lee Hood (Institute for Systems Biology, Seattle), Charlie Boone (U Toronto). #cytoscape
Other ways to read this feed:Feed readerFacebook