ICSB 2009 Sunday August 31st, 2009

Keynote Lectures

Stan Leibler, Rockefeller University, Protein: from structure to function

He compared protein alignment problems with finding trends from stock market. The rationale is that, in an alignment, for every position, there is one amino acid appearing most frequent. If we use that amino acid as the consensus for that position, and calculate the correlation between different positions for the consensus sequence, we could find conserved patterns in the consensus sequence. Those conserved patterns are not required to be continuous in either primary structure or secondary structure for that protein. Based on this concept, he found highly conserved patterns in one protein family and he confirmed the amino acid inside one pattern is likely to be continuous in the tertiary structure. In the Q&A session, he confirmed that he was working to extend this method to the pathway level.

Tobias Meyer, Stanford University, Signaling systems biology of cell proliferation

For mammalian systems, there are 200 cell types and 1000 signaling proteins in each cell. The signal generally follows: input -> modules -> output. Biosensors were very helpful but hard to make them work. Thus, he used SiRNAs for 3000 genes.

Part 1 is focused proliferation RNAi set. For cells, to divide or not to divide is a crucial binary decision. Considering there are 100,000 ORC in the genome, it is not a trivial problem. Previous findings suggest pRb -> positive feedback loop -> DNA replication. He found positive feedback loop (pRb is inside the loop) -> DNA replication. His model predicated 10 million fold increase in DNA replication by 6 fold increase in CDK4 activity. He found that the degradation of p21 was a time limiting step in the process. When there was 1% p21 left, the CD4 loop was initiated. The degradation process was slow and there might be signal integrated over time.

Part 2 was focused on cell migration RNAi set. The reference is Vitorino & Meyer, Genes & Dev 2008. He created an assay and a model. It was found that EGF was not required for refilling small holes ( an area of 5 cells) but required for refilling large holes. In the model, it was necessary to include drag force between cells in order to explain the data.

Bernhard Palsson, UCSD

Genome -> Metabolic

-> Expression

-> Operational ( regulatory)

Metabolic + Expression = ME

ME + Operational = OME

Jonathan Weissman

Amyloids are ordered protein aggregates. He found it was necessary for the prion fibers to be unstable, so that new “seeds” could be created. It was crucial to achieve a balance between the propagation of prions and the proliferation of cells. He used a modified RNA protection assay to map the ribosome binding sites on the genome, utilizing deep sequencing.

ICSB 2009 Stanford Monday August 31st, 2009

Morning session Speaker one: Lucy Shapiro, Stanford CtrA controls 90 genes, and GcrA controls 50 genes. They inserted fluorescent transposons to label the position of chromosome loci inside the cell. Before replication, the chromosome was highly methylated. During the process of replicating DNA, the chromosome became half methylated, thus turns on different genes. Therefore, genes were turned on sequentially from the ORC in the DNA replication process. At the same time, protolysis is tightly connected with transcriptional regulation.

Speaker two: Christopher A. Voigt, UCSF

Part 1: Do programs behave as the sum of the parts?

Part 2: Biophysical models

Part 3: How to scale up fundamental circuts

They built a photographic bacteria by using a light sensor and a light receptor. Then they build a simple edge detection algorithm by bacteria. It is important to unit test different parts, in this case an AND gate, and then put them together.

Impedance matching. The idea came from electrical engineering, in which the dynamic ranges have to match between constitutive modules.

RBS design. They built a thermodynamic model to compute synthetic RBS. The synthetic RBS is context dependent, meaning it has to be re computed every time the trailing sequence was modified.

NOR gate. NOR gate is boolean complete. It means you can use only NOR gate to achieve all the boolean relationships. Based on the different strength of RBS, they were able to get all types of NOR gate, achieving virtually all combinations.

Speaker 3: Sunney Xie, Harvard

nm spatial resolution, ms temporal resolution, high specificity, low concentration

Single molecule detection by localization. Target search on DNA is a mixture of 1D and 3D search.

Stroboscopic illumination. Instead of modulating shutter of the camera, modulating the duration of light. A repressor spent 90% of time on the DNA.

New projects. Persisters are antibiotic resistant bacteria cells. They are rare, 1 in 1 million cells. N. Balban, S Leibler, Science 2004.

Tracking persiters by 2D microfluidic device. One of the challenges are the sample size was very small.

They tagged 1000 genes with YFP and study correlation between mRNA and protein levels in single cells. The noise was different for low and high abundant proteins.

Speaker 4: Julie A. Theriot

If you plot Mean Square Displacement (MSD) v.s. time in a log-lot plot, the slope alpha is a indication of the normality of the diffusion. Alpha = 1 for normal diffusion and < 1 for anomalous diffusion.

Proteins move sub diffusively in vitro if the medium was crowded. (David Goodsed; Banks and Fradin 2005). The lower limit was 0.75 for alpha in this case.

laplacian filter -> threshold -> centroid position -> gaussion fit

Chromosome loci moves at D = 0002 um^2/s and alpha = 0.4.

Classic polymer dynamic models, Rouse model has an alpha = 0.5 and Reptation model has an alpha = 0.25. Rouse model with memory could explain the observation. The fractional Brownian Motion was preferred over CTRW.

The elastic environment will cause a negative correlation in the auto correlation curve.

Afternoon session

Speaker 1: Grant Jensen, Caltech

Speaker 2: Stuart Kim

In collaboration with Gene Myer (Janelia).

Automatic Cell Lineage Analyzer. Xiao Lin (Stanford), Fahui Lona, HanChuan Pena, Gene Myer (Janelia)

Spatial expression map. 2D clustering to distinguish between different types of cells.

Speaker 3: Ernst H. K. Stelzer

They used pulsed UV laser cutter.

Multi view microscopy. Specimen is not flat and embedded in agar.

Structure based LSFM. 3 Terabytes data for one run.

Speaker 4: Robert Waterson, University of Washington

Bao et al. PNAS 2006. StarryNite Automated analysis. AceTree Editor.

ICSB 2009 Stanford, Tuesday September 1st, 2009

Morning session

Markus Covert, Stanford, the model of an entire cell

Flux based analysis for metabolic networks

Question 1: Can we build the whole cell today? Modules -> models -> simulation of individual modules -> integration of modules. One of the key principles is to use separation of time scales. They built visualization tools. The current status is 60% of the genes

Speaker 2: Daphner Koller, Stanford, Understand gene regulation networks

Question: How to infer regulating mechanisms from gene expression data. MRNA levles of regulators are indicative of activity. Co-regulated genes have similar regulation program. Linear regulation network models worked surprisingly well, Lee et. al. PLOS Genetics, 2009

The Puf3 modules. MRNAs are stored in P bodies.

SNPs. Not all SNPs are equal. Framework for modeling gene regulation. They further improved model by making it lineage aware.

Speaker 3: Gene Myers

He started with saying the bio image informatics is an underdeveloped field and only 10 labs in the entire world are working on this problem. “Genomes permits us to do recombinant genetics on an entire genome”. He called it system genetics, and pioneers in this field are O’Shea, Hyman, Rubin(20,000 promoter fused with GFP).

There are three problems, intracelluar, cellular collections, organism behavior.

The problem characteristics are: Poor S2N, but strong prior knowledge; Coarse sampling, limited by diffraction; Only high through put assays are worth the effort.

DAPI, mCherry are output, GFP was used to label a subset of the cells. They found stereotype in worms and flys (Chris Doe, U of Oregion).

Single cell, medium through put. Automated cell targeting by laser ablation or light activation (Svoboda lab, mouse whisky)

Structure of a fly’s brain (20,000 lines). The position of neurons in the brain are stereotypical. In each hemoshpere, there are exactly 104 neuroblast ( Tzumin Lee Lab, Boston, JFRC). Promoter segments are fused with GFP (Rubin, Tzuman) -> image register -> tracing (hardest) -> comparing

Microscopy are high through put, and the goal is to map cells at ‘particle’ level. It is surprising that the computation is the bottom-neck at each project. It is not likely that a general framework could be built. One of the new projects is to use ultrasound for kidney.

Noon session, Chemical tools

Speaker 1: Morpholinos with photocleavable hairpins. Caged morholinos could increase the specificity for doing micro array.

Speaker 2: MeroCBD. BioSensor

Speaker 3: UCSF. 500 kinases in human

Speaker 4: Tom Wandless, Stanford

Destablizing domain controlled by small molecules. Banaszynski et. al. Cell 2006

ICSB 2009 Wednesday September 8th, 2009

Signaling Networks

Galit Lahav, Harvard

input -> p53 input -> slow enzyme -> p53

The purpose of the coherent feed forward loop is to separate different time scales.

//P53 bio sensors. General framework for bio sensors, especially for bio sensors that are able to distinguish between regular form and phosphorylated form

P53 pulsing v.s. NFkB pulsing. Encourage people to look at basal dynamics. Modeling help understanding how to use coherent feed forward loop to generate pulsing

Speaker 2: Kong-Joo Lee, Ewha Womans University, Seoul, Korea

//Different time scales for different biological processes: RNAi fast, transcriptional regulation medium, translation slow, post translation slow, protein localization fast

Speaker 3: Wendell Lim, UCSF, the design principles of cell signaling

Q: How do cells process information? – Modularity; – Hierarchical organization

Classical approach is learning by destruction, while the newer approach is learning by synthesis and evolution

Key: What are the general rules of adaption ? ( Ma et. al. 2009, Cell; Kati et. al. 2008) To simplify the problem, they consider a three nodes network, with one input node, one regulatory node and one output node. In this network, there are ~16 k toplogies and ~ 10 k parameters. Only a small fraction of all configurations behaves adaptively, and all those belongs to either of the two types: first is negative feedback with buffering node; second is incoherent feed forward loop with proportioner loop

Speaker 4: Peter Sorger, MIT

//Hijack apoptosis to kill cancer cells. There are much more mitochondria in cancer cells than normal cells.

After drug treatment, there are 22% of surviving cancer cells. Compared with after antibiotic treatment, there are 10^-6 persisters (Sunny Xie). In Peter Soger’s case, there is transient resistance ( 1 day ) to multi inducers of Trail. They used DLE based phase diagrams to represent their data.

//Could it due to unequal concentration of protein reactant? We could do single cell studies with optical tweezers , droplet micro fluidic device and single point FCS.

Cell fate is determined by a race between the dead signal and the live signal, whoever arrives first determining the cell fate.

Morning session 2, Emerging Concepts in Cell Signaling:

Speaker 1: Philippe Bastiaens, Max Planck Institute of Molecular Physiology

The spatial organization of Ras signaling. How is the spatial distribution of Ras maintained? The palmitolyation is efficient at counter acting entropy driven redistribution of palmitolyated proteins. TIRF can be used to approximate SRP, because the loss of fluorescence is a combination of lateral diffusion and unbinding from the surface.

Speaker 2: Charles Boone, University of Toronto, HHMI

Yeast synthetic lethal project, trying to create double mutant for all 6000 genes, resulting 6000 by 6000 experiments.

They used correlation to cluster genes, and calculate the distance between nodes.

// could we use similar method to calculate SNP in the genome?

Speaker 3: Peter Pryciak, University of Massachussetts, subcellular colocalization

Input output behavior. What is the functional unit of scaffold mediated signaling? How is membrane localization regulated and why is it important? Two roles for scaffold protein: initiation and propagation. (Lamson 2006, Current Biology)

// could scaffold protein change the probability of collision, therefore, modifying the Gillespie algorithm?How can you increase signal strength by targeting to membrane? Because you decreased diffusion coefficient or you limited the dimension of diffusion from 3d to 1d.

FCS related. C.I. Maeder 2007, Nat Cell Biology; B.D. Slaugher 2007 PNAS;

Yablonski, 1996, PNAS ; Inouye 1997 Science.

The conclusion is membrane localization is necessary and sufficient to promote trans signaling.

Speaker 4: Sharad Ramanathan, Harvard

Q: Are there soft nodes that allow for easy changes in the network? They tried to rewire the pathway. The MPAK proteins are the most conserved parts in the MPAK pathway among species.

// PPI network between species or pathway between species or the energy for conserved proteins between species.

Afternoon session, Dynamics:

Speaker 1: Michael B. Elowitz, Caltech, How do cells encode signals?

System 1: Crz1 is a cytosolic protein, and Crz1-P-GFP is localized to nucleus. Localization to the nucleus occurs in burst. The duration of the burst is not modulated, but the frequency is modulated. This arises a general question: how do we control gene expression levels to be proportional to each other. By using frequency modulation, natural target promoters are naturally coordinated.

System 3: Drosophila. Irn Berstein (Delta attached to plate). Levine, Hua et. al. PLOS Biology, 2007. De Celis et. al., Development 1997

Speaker 2: Daniel S. Fisher, Evolutionary dynamics and biological disorgniazation

//Black swan in evolution or ecology

//Since only protein surface matters, we could use GFP as a nice scaffold and add kinase to different parts of the scaffold.

Rapid genetic changes in Citrate medium

Kerwyn C. Huang, BME, Stanford

Protein localization is important for bacteria. The system he used is MinC, MinD and MinE. Turing instability

Alexander van Oudenaarden, MIT

Degradation at burst.