Supporting materials for manuscript on the design of knotted proteins.
Before using this software, please obtain a license (free for academic users) for the Rosetta macromolecular modeling suite. Information on licenses can be found here: https://rosettacommons.org/software/license-and-download
First, it's important to realize that back in the dark ages of protein design (pre-2020, pre-hallucination, pre-diffusion), and with the fragment-based sampling approaches employed here, it took hundreds to thousands of CPU hours to generate good populations of de novo design models with novel topologies. This is because, in general, the approaches for backbone topology sampling were brute force random sampling of (for example) helix lengths and loop torsions. So we relied on many independent simulations, each several minutes long, to inefficiently explore the very large space of structures and sequences. All the calculations described in this paper were run on a large computing cluster.
Second, this design code is rather old (some of the original trefoil simulations date back to 2015), and was written on a branch of the Rosetta package during a time of rapid change in the framework for modeling symmetric systems (like these tandem repeat proteins). Unfortunately, the code became stranded on this branch and was never integrated into the main trunk of Rosetta. For that reason, it's not being released as a mode of the publicly released modeling package. This means that right now, it's only available as a binary executable (download link below). However, Rosetta is currently being transitioned to free open source software, which should make it possible to release the code on this branch in a form that anyone could compile for themselves.
Third, I wouldn't personally recommend that anyone actually use this design code for
new design calculations, since design methods have advanced so dramatically in the
past couple of years. The key concept in this paper is very simple: that you can
get a
-
trefoil_flags.txtandpentafoil_flags.txt: Command line flags for running the first-round design calculations, in which we explore a range of different helix lengths. See theRunningsection below for details of what the command line would look like. -
trefoil_resampling_flags.txtandpentafoil_resampling_flags.txt: Command line flags for running the second-round design calculations, in which we focus in on specific helix lengths and backbone turn conformations. -
input/Folder with input files for the calculations. -
src/Folder with C++ files that, together with the Rosetta library source code, implement the design calculations. In this folder aresymdes.ccwhich is the main executable file, and a few supporting*.hhfiles.
Some of the inputs for the design calculation (for example, the file of backbone fragment torsion angles) are too big to host on github, so we combined them all into a big ZIP file, downloadable here:
https://www.dropbox.com/s/r3kqntd6bdtzy6b/big_input.tgz?dl=0
After downloading the file, move it into this folder and uncompress it, for example with the command
tar -xzf big_input.tgz
A pre-compiled binary for linux is available here:
https://www.dropbox.com/s/zktq4yvfzvw5xtp/symdes.static.linuxgccrelease?dl=0
A large set of trefoil and pentafoil design models can be downloaded here:
A typical command line might look something like:
./symdes.static.linuxgccrelease @trefoil_resampling_flags.txt
Multiple jobs could be launched with the same command line (on a computing cluster,
for example) and they will share their work using temporary files
created using the -shared_output_tag argument as a filename prefix.
This is mainly important for score-based filtering of final design output.
For the second round of pentafoil design calculations, we used the Protein-MPNN neural network to select new sequences starting from the backbones of the original design models.
https://github.com/dauparas/ProteinMPNN
For selecting second round pentafoil designs we modeled their structures with AlphaFold and looked for sequences that were predicted to fold into the desired topology.