Email: noah.herrington@mssm.edu

This README file explains how to use the useful scripts for AF2-based modeling of kinases

in alternative conformations.

This repository includes five scripts:

1) AlphaFold2_advanced_modified.py

- Used to run AlphaFold2 predictions of a given kinase

- Outputs 5 models

- Initial stages of program running installs extra dependencies and required packages

- "is_training" set to "True" to enable model dropout

- Takes two arguments: 1. input fasta file and 2. desired output directory

- Dependencies: biopython 1.79, jax 0.4.1, tensorflow 2.11.0

- Usage: python3 AlphaFold2_advanced_modified.py input_fasta output_dir

2) colabfold_alphafold.py

- Modified version of script from ColabFold repository

- Allows for output directory argument

- Dependencies installed by or before running AlphaFold2_advanced_modified.py

- Replace default script with this one

3) AF2_kinase_families_stackedbarplot.py

- Used to classify models downloaded from the AlphaFold2 Protein Structure Database

into their respective kinase families

- Must be used in conjunction with kinfam.csv and output from

Kincore classifier (https://github.com/vivekmodi/Kincore-standalone) as a csv,

titled "kinases_classified.csv"

- Dependencies: matplotlib 3.7.0, numpy 1.23.5, pandas 1.4.4, plotly 5.9.0

- Usage: python3 AF2_kinase_families_stackedbarplot.py

4) Kincore_ConformationDistribution_Doughnutplot.py

- Used to generate a doughnut-shaped plot of distribution of AF2-predicted models

by their conformation

- Outputs fractions of each conformation to the screen and plotly doughnut plot in browser

- Dependencies: matplotlib 3.7.0, numpy 1.23.5, pandas 1.4.4

- Usage: python3 Kincore_ConformationDistribution_Doughnutplot.py

5) MSA_models_Mobitz_plot.py

- Used to generate plot of models by pseudo-dihedral angles (proposed by Mobitz (2015)),

which group kinase models by movement of their DFG motif

- Outputs a saved hi-res image of the plot

- Allows for input of a chosen MSA depth and projection (3D/2D), where 2D represents the Mobitz

plot and 3D adds an additional dimension for RMSD with respect to a prototyical DFG-in structure

- Necessitates having downloaded the PDB structure 1ATP and renamed it

"1ATP_cAMP-dep_prot_kinase_ATP_DFGin_Reference.pdb"

- Usage: python3 MSA_models_Mobitz_plot.py

6) enrichment.py

- Used to generate enrichment plot of a series of docked models

- Allows input of colored curves by MSA Depth, pLDDT Score, or Conformation

- If coloring by pLDDT Score is desired, requires presence of CSVs containing

classification of all models at those depths (created with Kincore - Dunbrack Lab)

- Requires two positional arguments:

a) Name of the kinase, for which plots are created

b) Quality by which curves are colored (i.e., msa, plddt, conf)

- Usage: python3 enrichment.py {a} {b}