SARS-CoV-2 RBD deep mutational AlphaFold2 structures carry phenotypic information.

Scientific data.

This repository is provided alongside AlphaFold2 data found at:

https://figshare.com/projects/SARS-CoV-2_RBD_single_mutant_AlphaFold2_structures/150089

The scripts are presented to reproduce all results in the paper titled: "SARS-CoV-2 RBD deep mutational AlphaFold2 structures carry phenotypic information"

Getting started.

Download and unzip the data running (for linux/mac).

./data_prepare.sh

This data can also be downloaded manually.

Data structure created:

FASTA

• wuahn
• alpha
• beta
• delta
• eta
• omicron_ba1
• omicron_ba2

Structures

• wuahn
• alpha
• beta
• delta
• eta
• omicron_ba1
• omicron_ba2

Generated WT main varinats:

These are used for local distortion and alignment analysis.

./wt_pdbs

Data Records

Data Records: Simple statistics notebook.

Exploring simple amino acid variation statistics.

./data_usage_scripts/aa_changes_vs_RMSD.ipynb

Exploring the importance of the interface with respect to ACE2 binding values.

./interface_exploration/interface_importance.ipynb

Umap results figure:

Visualize embeddings (representations) of the generated dataset.

./projections/UMAP_all_vars_structs.ipynb

Technical Validation

Validation of local structural distortion: Visualizing structural distortion on mutation (mutation-distortion maps).

We investigate the mean distortion in distance each amino acid has for each each structure created. We observe that every mutation at a given position corresponds to a structural distortion at that position as well as some others in the structure.

./data_usage_scripts/mutation_distortion_maps.ipynb

To create all of the maps for all of the mutant clusters:

./data_usage_scripts/mutation_distortion.py

To view all of these maps.

./data_usage_scripts/all_mut_dist_maps.ipynb

Validation by predicting phenotypic quantities: Wuhan generalization experiment:

cd wuhan_generalization_experiments

To create all embeddings and run random forest classifier experiment.

./wuhan_generalization.sh

Then to visualize results with the notebook:

./RF_generalization_plot.ipynb

Validation of protein disorder: Protein Disorder analysis notebook.

Here we investigate how pLDDT (b-factor) output of the AF2 files correlate with state-of-the-art disorder estimations.

./disorder_analysis/iupred_notebook-Analyses_mod.ipynb

Validation of 3D alignment: Validation that structures are correctly aligned.

We see a direct correlation between RMSD values in the 3D space and rotation invariant adjacency matrix space.

./rotation_analysis/MSE_vs_adjacency_dist.ipynb

Usage Notes

Scripts to create PQR files

Create PQR files for further downstream tasks.

./APBS_and_PQR/PQR-APBS.ipynb

FoldX RBD protein expression prediction experiment.

Run FoldX stability function for all structures with foldX software installed.

./foldX_stability/foldx_expression_prediction.py

Antibody escape plots notebook.

Using the Bloom et al antibody escape calculator, we observe structural distortion correlates with antibody escape.

./antibody_escape/antibody_escape_vs_structuraldistortion.ipynb