Of course I also have to write something about Cryo-EM in my thesis for this
I found the following video to be a nice summary:
This is especially usefull because of the papers listed in the description (as in any meaningfull science
video on youtube I would hope ^^)
As I am writing the fundamentals part of my Thesis I have to read up on a lot of stuff, like fore example the experimental methods for getting a protein structure. And although the new “hot” method for this is of Cryo-EM, I found it quite interesting to understand how to get from a scattering pattern to a protein structure. Which basically is getting from a scattering pattern to an electron density by doing a fourier transformation.
Where F is the according 3D scattering profile that one gets by measuring the scattering from many angles. But that is already too much I actually simply wanted to drop 3 good videos from The Royal Institution in here that I found:
A longer talk (around 1h)
And two shorter videos
This really helped me understand a bit more about the differences between AlphaFold1 to AlphaFold2 and RosettaTR to RoseTTAFold respectively. Very cool, if somewhat long talk (well its a seminar ^^).
Particularly interesting is the performance comparison between the different version of RoseTTAFold and AlphaFold2 as illustrated on this slide of the presentation :
I guess I will have to give RoseTTAFold a try, especially since its supposedly less resource hungry. And maybe a little bit less of hazzle to install (if you do not have the luxury of just using docker on your computational infrastructure, … so like in most HPC environments …)
This is going to be a very Rosetta specific article, if you haven’t used it before this probably is not for you. In short: I recently had to do a lot of simulations in Rosetta, ranging from simple relaxations over protein design to comparative modelling. As the problem with simulations in Rosetta is mostly to make it run I usually do not add too much after I have done this. Now Rosetta Script offers this nice functionality of <SimpleMetrics> where you can specify all kind of metrics that should be evaluated for the generated poses.
So in order to do this afterwards without adding additional overhead to an barely already running script we can simply write a new rescore.xml where we insert all the metrics our heart desires
Here a PDB consisting of 3 chains A,B and C is to be rescored. The Protein here was a Polymerase with DNA bound, which is
the reaseon for the different chains: The protein is chain A and the two strains of the DNA are labeled as chains B and C.
Therefore we need multiple entries in the <RESIDUE_SELECTORS> region to do things with these regions.
Initially I wanted to calculate the interface energy not only between the DNA and Protein but also between the Protein and the single DNA strains. However it seems you can’t reuse a simple metric. At least I always got parsing errors from the Rosetta XML script verification stage.
Only upon removing all but one of the identical Simple_Metrics, would my script run.
To rescore the previously generated poses we need to generate a pdb_list as input, which is simply a text file containing the path
to all the pdb poses. To run the rescoring it is convenient to write all the flags into a single file rosetta_rescore.options.
Using this we can then run the rescoring directly by doing
mpirun –np $N_CORES rosetta_scripts.mpi.linuxgccrelease @rosetta_rescore.options
As I did my run on the LB2 Cluster at TU-Darmstadt, I wrote a slurm script for doing this which looks as follows.
And there you have it, this is nothing special however I often enough forget how to do it and therefore thought
it could be interesting to have this documented somewhere, where it eventually might also help someone else.
I am in the process of customizing and starting this blog. As the note suggest I am planning
to write about various subjects I encountered during the work for my PhD, which in short is: How to waste a lot of computer power in order to hopefully find out something interesting about Proteins, Optimization Algorithms and hopefully the underlying physics of these. (Yes there is underlying Physics for Optimization Algorithms xD, at least for Monte Carlo Optimization the analogy is relatively obvious from a Physicists point of view. I probably should write an article about that).
When I am finished playing around with Word Press here I will probably write a real first blog article. Until then – Ben