Alexander Hoffmann, University of California, San Diego, Parsimonious Gene Regulatory Network modeling reveals mRNA half-life control of the pathogen-responsive transcriptome
- innate immune response info processing
- decompose into modules.
- combinatorial code encoded by signal pathway, decoded by TFs combos on promoters.
- signaling dynamics are relevant to behavior.
- previous work on encoding the temporal code. How is that read out?
- GRNs from omits TF profiling, histones, TF binding.
- Problems: expression of TF neq activity, binding which gene?
- Activate indiv pathways, id coregulated genes.
- Gene clusters contain distinct groups of GO terms
- 1kb upstream search for TF binding sites.
- some genes nascent tx not match total mRNA. LPS required to repress mRNA degredation.
Bernie Daigle, Jr., University of California, Santa Barbara, Accelerated Maximum Likelihood Estimation for Stochastic Biochemical Systems
- estimate kinetic parameters in stochastic system.
- bistable genetic switch cross repression noisy switching.
- With exact parameters can sample with Gillespie.
- what if you have a trajectory but not kinetics. Max likelihood inference.
- Effect of discrete sampling
- method require simulating data from current guess set f pars. Bad guesses are intractable to get data. Compare instead to set of sim data that are closest to data.
- try to hit all data ‘harder’ the more data you have.
My talk. Discuss with Ido Golding’s students. (Baylor). Others
Vittorio Cristini, University of New Mexico, Tumor Modeling
- Cancer from a theoretical engineer / applied mathematician
- war on cancer has been very unsuccessful from a statistical point of view.
- Molecular model, cell model (agent based approach) tissue model (how big is the tumor?) Looking for emergent behavior.
- Intro to mathematical pathology: map patient measurements into estimations relevant to clinical planning
- Nutrient diffusion is the primary limiting factor to the growth of in-situ tumors.
- Tumors reach stationary growth states very quickly — propose as reason why mammory screening on 3 month intervals show low detection can grow quickly — isn’t that better identified by clinical study with denser sampling?
- proliferated index and apoptotic index can be directly measured from histology data (no fit).
- Model projects tumor size/region that should be removed by dissection: how far to tumorigenic cells expend past calcified region (if left can lead to second tumor formation).
- Patient specific approach to predicting tumor growth.
- mammogram prediction and actual tumor size over-predicts or under-predicts by factors of 3.
- histological grade (PI) is also a poor predictor — proliferative index and apoptotic index both increase with grade.
- what are your other parameters — growth and death rate can get from histology, 3d simulation of cell growth/spreading and migration.
- L/A is a much better predictor (growth / apoptosis).
Michael Deem, Rice University, Heterogeneous Diversity of Spacers within CRISPR
- capture of viral DNA into crisper locus rare event in lysis. Expression of spacers crRNAs silence phage RNA to provide immunity.
- why more diversity of spacer elements near beginning than end of sequence? (recent spacers more variable than historic spacers).
- model: limit system to 2 spacers total. bacteria and viruses grow. infinite population.
- more positive fitness gain from dominant viral population. First spacer is random, second spacer gets driven much more strongly to be equal to the dominant phage genotype.
- what about selection back on phage? and phage mutation? Add these, use literature claim: single basepair mutation allows phage to escape detection.
- random deletion after 30 are in, probability scales linearly with distance.
- simulation, diversity still increases monotonically with distance from leader.
Qiong Yang, Stanford University, Switch-like negative feedback promotes clock-like mitotic oscillation
- Ferrell Lab
- Early cell divisions in Xenapous: no check-point — free running.
- Divisions still synchronous / coordinated
- biochemical clock Cdk1 cyclin B1 form bistable switch.
- Cdk1 activates APC which down regulates cycB1 (complex) — makes oscillator out of bistable system.
- Reduced model cdk1-cycB call CC. CC activates self, activates A, A represses C. simple clock architecture.
- CDK1 nullcline is cubic/like curve (experimentally measured) in response to APC.
- sigmoidal response for nullcline of A as fxn of CDk1 required for limit cycle.
- more ultrasenstive larger oscillatory region.
- How about period and amplitude stability? Period more stable at high hill coefficient.
- How are you getting time in minutes in your model? — best guess. qualitative behaviors hold.
- probe in cell extract adding protein components back in while inhibiting different parts of the pathway. estimate hill coefficient of ~37 from data.
- role of noise in circuit? Effects of feedback delay
Alexandra Jilkine, University of Arizona, A Density-Dependent Switch Drives Stochastic Clustering and Polarization of Signaling Molecules
- Understanding spontaneous polarization in budding yeast
- existing model: stocahstic activation of Cdc42p + positive feedback builds cluster.
- what stops cluster when the un-clustered state is desired?
- what limits the spread of a cluster (not spread whole membrane). why only 1?
- minimal model: membrane bound diffuse slowly, free diffuse fast.
- loss of clustering for large N. Repression of clustering for small N. Emergence of individual cluster for intermideate N. Substrate sequestration Turing instability.
- small n case just like viral spread low density arrival of new components not greater than leaving rate of existing ones, even if you artificially build up a small cluster.
- Intermediate N: clans either drift together and become fused. all clans go extinct as t–> infinity (but new ones seed. Meanwhile non-dominant ones much more likely to die).
Wan-Chen Lin, University of California, Berkeley, Spatial organization in cell membranes
- Groves lab. immunological synapse
- T cell receptor and cSMAC in center. ICAM-1 larger receptors, outer ring of synapse.
- measure rate of radial movement of clusters.
- these supported lipid membrane talks all look the same to me: receptors bind and cluster. They don’t move past barriers. clustering effects signalling.
- myosin is important for signalling. Where does this fit in? –required for proper clustering / bulls-eye synapse appearance.
- dissect probability of activation as a function of number of receptors in cluster. Need at least 4 peptides to activate T-cell. (yeah that’s a nice measurement).
- use stochastic fluctuation analysis to estimate number of components.
- PALM fixed pSMAC junction (resolution 45 nm)
- larger clusters experience larger force more quickly moved to center.
Closing Banquet Talk, Peter Sorger, Harvard Medical School, Measuring and Modeling Cell Decision Processes