9:30 am – 5:00 pm, 9:00 pm – 12:15 am

PH project

• replotting all histograms — no more collecting all the big stuff into the last bin, this was confusing.
• made new figure-producing figs (actually just histograms, the images are separate code still).
• sent updated histograms (8 total) to Ajaz, along with some comments on the fig 2 layout and captions (can we please include more labels on the figures!)

project 2, team meeting

• My tasks for next week:
  • cluster analysis for formamide titration data
  • cluster analysis of E4 data

Chromatin simulations

• test polymer simulations with greater spacing (10 nm between molecules)
• it looks like these all relax to 1, nI thought the nodes were supposed to be 10nm.
• This is probably another good question for the authors if they ever write back.

Hacking an attractive force.

• lets just build a quadratic or quartic well
• outside the well we have the energy as 1 kT here. I have since changed it to zero.
• This finally appears to be working with the code that Bogdan started ( see image below).
  • there’s this weird dip at the beginning, which could be due to the added stiffness of my polymer, it takes a while before the nodes can start interacting, but I’m not sure that’s it. Should probably investigate this more.
  • But for molecules of length 100+ we are closer to the expected 1/3rd scaling exponent.
• My hacked version of a free polymer using openmmlib seems to ignore my force though. Maybe its a distance issue. Still has a 0.6 free polymer type scaling.

10:00 am – 12:45 am

More simulation troubleshooting

Experimenting with weaker attractive forces

    energy = ("index*epsilon * .000000001 * (step(r-rm)*(r-rm)^4 * (1- step(r-2*rm))+ (rm)^4*step(r-2*rm));"
              "index=index1*index2")

reasonable scaling

10 fold stronger bonding: “sticky8” simulation

    energy = ("index*epsilon * .00000001 * (step(r-rm)*(r-rm)^4 * (1- step(r-2*rm))+ (rm)^4*step(r-2*rm));"
              "index=index1*index2")

not very substantial change in structure

further 10 fold reduction

• sticky 9
  energy = (“indexepsilon * .0000001 * (step(r-rm)(r-rm)^4 * (1- step(r-2rm))+ (rm)^4step(r-2rm));”
  “index=index1index2″)
  Despite the fact the scaling is only slightly below the SAW polymer, the system is not fully self avoiding.
  The closest two monomers at any point throughout the simulation gets closer and closer as N increases, rapidly falling below the 10 nm exclusion.
  

redoing simulations with no attractive forces

• let’s check the self avoiding condition again.
• hmmm… This actually looks very similar.
• and I thought this scaling came out to .60 last night but doesn’t seem to.
  

More issues

Now I can’t repeat Sticky6 (~flat scaling), final simulation to complete last night, in which I added a weakened attractive force back in after believing I had gotten the random walk polymer to scale properly at 0.6.

Mirny lab code

• Bashed away a bit at scaffoldExample to get it to run:
• This is very slow. But globule 128 is 128,000 monomers, so that’s probably why.

Trying to shorten simulation time by reducing data size. Wrote script to load and truncate

Finally got this working

• wrote a matlab wrapper to initialize the simulations using my RandomWalkPolymer function and to change key parameters like number of nodes and simulation time (see GitScripts)
• simulations seem to be behaving well:

Questions on polymerutils

• what are the inputs exampleOpenmm(sys.argv[1], sys.argv[2], int(sys.argv[3])) the example has 200 1 1 but doesn’t explain what these numbers mean.
• Can someone explain the shape of the Grossberg and LG ‘repulsive’ potentials? Why is there a minimum near the contact radius that goes up and saturates afterwards.

10:00 am – 5:00 pm, 10:00 pm – 12:30 am

Lab stuff

• Finally found and installed drivers for Cajal’s AMD Radeon 6900 graphics card (this card is still driving the monitors).
• installed new NVIDIA Quadro K4000 for GPU computing (still need to install the CUDA SDK etc)
• moved RAM from Tuck to Cajal. Tuck failed to boot (maybe just extremely slow, should give it an hour or so to attempt to sort itself out after windows updates, it doesn’t like these and they process very slowly).

Chromatin

Simulations

• clear exclusion issues revealed by sticky polymer sim. All polymers fit into the same volume. The expected volume of 500 monomers is 10E6, whereas the volume of the polymer (at all stages basically) is 10E4. Clearly an issue. (As is the negative scaling coefficient, since more monomers squeeze towards the center inside the volume. This is only expected in packing chains in tightly confined spaces).
• attractive and repulsive energies built off of chain_a, which is the centroid-to-centroid distance between monomers, not the hard-sphere radius distance. This is probably a problem.
• now running short sticky polymer simulations with simple change of radius for chain spacing in the energy functions.
• Also exploring the alternative Grossberg energy functions used in the Minry polymer tools (in principle I should be able to import and call these rather than copy them since they are there own classes. Though figuring out the variable space is tricky since there’s essentially no documentation of it).
• at first glance from the simulations this does not seem to have solved the problem.

More issues

• ‘GrossbergRepulsion’ energy function from Mirny lab openMM polymer doesn’t make sense to me as a repulsive energy. Why is this non-zero beyond the radius of the monomer?
  

Discussions with Bogdan

• Troubleshooting non-self avoiding behavior and wierd potentionals
• believe repulsive force was targeted at all but the attractive nodes. Removed 1-index1*index2 part of repulsive force in polymer simulations (see GitScripts for code log)
• Reducing potential energy magnitudes to minimize explosive behavior of polymer chains.

STORM

• imaging sample D07. Not looking very strong staining, especially for a 150 kb region. Something has probably gone wrong.

10:20 am – 11:15 pm

New simulations

• sticky polymer simulations. same as blue simulations but reduce binding energy 10 fold. index*epsilon * .0001 * (step(r-rm)*(r-rm)^4 * (1- step(r-2*rm))+ (rm)^4*step(r-2*rm));
• now running with 10 polymers, log scaled from 10 to 500 nodes, 3 rpts. Hopefully this will finish before tonight so I can look at it and set up a new batch of simulations.

STORM observations.

• Black domains labeling pretty sparse I think — for an 80 kb region this are pretty dim.
• background is pretty low, probably worth tripling the concentration and seeing if that helps, maybe I’ll try this tonight.
• imaging D06. These regions are WAY larger than D05. Maybe something happened in the genome here. (or the D05 stain failed more spectacularly than I thought, will have to look at the real counts).

PH project

• Analyzing S2 PH data
• sent draft of S2 PH data ANTC data to Ajaz

9:30 am – 5:45 pm

Simulations

To Analyze

• OpenMM free Polymer sims
• Odyssey Matlab periodic bonds

OpenMM free Polymers

Only a single run, but not quite the right scaling — expect 0.60. And a y-intercept at 0. See previous simulations results using my monte carlo simulator in Matlab

This is still growing substantially faster than closest packing.

Running new simulations with more repeats (4) and smaller polymers (only up to 500 — the large ones really take forever).

Matlab Periodically bound polymers

Not enough simulation time for binding nodes to find each other. Some succeed, some do not.

Future development

• wrote to RC computing about installing OpenMM and openMM-polymer on Odyssey.

STORM

• imaging D04
• sample not too bright. Hopefully still enough localizations to map out the basic structure.