For the CASP15 RNA category, we aimed to assess predictions by directly comparing to expiremental data. In the respository is all scripts and code used to obtain and analyse scores. All map-to-model scores used are molecule agnostic and thus can be applied to RNA as here or other molecules, however, some of the pre-processing and other scores herein are RNA specific.
To install all necessary python packages, in this repository run:
conda env create -f environment.yml
conda activate casp_rna_em
pip install .
This will also install TEMPy and rna-tools.
To run this code always activate the appropriate conda environment before running.
conda activate casp_rna_em
Obtain an academic liscense and download phenix. Install according to instruction, for example for linux:
tar -xf phenix-installer-*.tar.gz
./install --prefix="desired_location"
You can either add source desired_location/phenix-?.??.?-????/phenix_env.sh to your ~/.bashrc or in applicable function calls in this repository specify phenix_location='desired_location/phenix-?.??.?-????/build/bin/'.
The binaries used are phenix.clashscore, phenix.rna_validate, phenix.model_vs_map, phenix.mtriage, and phenix.dock_in_map.
Versions 1.18.2 and 1.20.1 (linux) were tested.
There are models whose geometries surpass the checks of phenix. They will crash phenix and not recieve phenix scores. This can be manually overwritten if desired by changing max_reasonable_bond_distance in master_params_str to 100 in phenix_location/modules/cctbx_project/mmtbx/monomer_library/pdb_interpretation.py
Download, agreeing to non-commerical liscence, and install according to the instructions for your opperating system ChimeraX. You can either add alias ChimeraX='chimerax-location/bin/Chimerax' or in applicable function calls in this repository specify chimerax_location='chimerax-location/bin/'
Qscores were calculated but not used in the final ranking because of the low resolution of the maps.
Section under construction.
Simply install according to:
git clone https://github.com/pylelab/USalign.git
cd USalign
make
You can either add export PATH=$PATH:/location/USalign to your ~/.bashrc or in applicable function calls in this repository specify usalign_location='/location/USalign'.
Request LGA download.
tar -xf LGA_package_src.*.tar.gz
cd LGA_package_src/
Add to you ~/.bashrc:
export PATH=$PATH:lga-location
ulimit -s unlimited; lga
Examples of running the pipeline can be found in example_run_scripts/ with a full run in example_run.py and then parralizable options compatible with a slurm computer cluster run with example_run_para.py. To combine results from these runs, use the file followed by example_run_combine.py once all the calculations are done.
Code can be found in process_files.py. With examples on how these steps were run results/rna_results.ipynb, example_run_combine.py.
For CASP15 we choice to compare all submitted models from each group (5) to the target and take the best score. If there was multiple conformation we compared to all conformations and took the best score. See reduce_df.
For CASP15, we ranked based on Zscore. For each target and metric a Zscore was calculated using the best score, as described above. Outliers (Z<-2) were removed for the final mean and standard deviation calculations. See get_zscore. The weighted Zscore was then calculated for each target and group, see get_weighted_sum_z. The final ranking was decided, for each group, by summing Z-scores for every target, ignoring any Z<0, see get_group_score.
Visualizations used in the manuscript can be found in results/rna_results.ipynb. Further, chimerax sessions to create the figures can be found in this folder, and the maps and models related to these files can be found here.
For the CASP15 RNA expirementalist paper, R1149 and R1556 were analysed in detail including all models junction geometry and angle. This analysis is in the R1149_R1156_analysis folder.
CASP15 papers are in preperation