timkartar
commented
1 week ago Hi, for your first question, you need to provide me with the file and the full error log, if you would like me take a look.
For the second one, here's some code to help you do it from the output files for 3q05.pdb (biological assembly). Please note that this prints average and max aggregations. You need to implement the log sum aggregation, which should be easy.
from Bio.PDB import PDBParser, PDBIO
import numpy as np
import sys, os
import matplotlib.pyplot as plt
from matplotlib import rc
import seaborn as sns
sns.set_theme()
sns.set_style(style="ticks")
rc('font',**{'family':'sans-serif','sans-serif':['Helvetica']})
#rc('text', usetex=True)
parser = PDBParser()
pdb = "3q05"
model = parser.get_structure("3q05","3q05.pdb")[0]
cwd = os.getcwd()
#atoms = list(model.get_atoms())
#for item in atoms:
# item.set_bfactor(0)
atoms = list(filter(lambda a: a.element != 'H', model.get_atoms()))
v_prot_all = np.load(cwd + "/"+ pdb + "_v_prot.npy")
interface_atoms = np.load(cwd + "/{}_edge_index.npy".format(pdb))[0,:]
v_prot = v_prot_all[interface_atoms]
diffs = np.load(cwd + "/{}_diffs.npy".format(pdb))
diffs = diffs/diffs.max()
atom_dict = {}
for i in range(len(atoms)):
k1 = "{:.2f}".format(atoms[i].coord[0])
k2 = "{:.2f}".format(atoms[i].coord[1])
#print((k1, k2))
atom_dict[(k1, k2)] = atoms[i]
#print(atom_dict)
plt_dict = {}
for i in range(len(v_prot)):
k1 = "{:.2f}".format(v_prot[i,:][0])
k2 = "{:.2f}".format(v_prot[i,:][1])
try:
atom = atom_dict[(k1, k2)]
except:
print(res.get_resname(),res.get_id()[1], res.get_parent().get_id())
continue
res = atom.get_parent()
key = "{}{}{}".format(res.get_resname(),res.get_id()[1], res.get_parent().get_id())
if key not in plt_dict:
plt_dict[key] = [diffs[i]]
else:
plt_dict[key].append(diffs[i])
print("{},{},{}".format(atom.name, key, diffs[i]))
final_dict = dict()
for key in plt_dict:
k = list(key)
k[1] = k[1].lower()
k[2] = k[2].lower()
k = "".join(k)
final_dict[k] = [np.mean(plt_dict[key]), np.max(plt_dict[key])]
keys = np.array(list(final_dict.keys()))
vals = [final_dict[i][1] for i in keys]
order = keys[np.argsort(vals)[::-1]]
import pandas as pd
plt_df = pd.DataFrame.from_dict(final_dict, orient="index", columns=["Average","Max"])
plt_df = plt_df.reset_index().melt(id_vars=["index"])
fig, ax = plt.subplots(figsize=(8,3.5))
plt_df.sort_values(by=['value'])
print(plt_df.to_csv())
palette = {"Average": "black", "Max":"firebrick"}
sns.barplot(x="index", y="value", hue="variable", data=plt_df, order=order[:20], palette=palette)
#print(plt_df.to_csv())
#ax.set_ylabel("Aggregated Network importance \n(atoms within 5 $\AA$) of CG DNA")
#ax.set_xlabel("Interface residues")
ax.legend()
xticklabels = ax.get_xticklabels()
ax.set_xticklabels(xticklabels, rotation = 45, ha="right", fontsize=15)
#yticklabels = ax.get_yticklabels()
#ax.set_yticklabels(yticklabels, fontsize=15)
plt.tight_layout()
#plt.show()
#plt.savefig("./interpret_residue_wise.svg")
plt.close()
Drjackxiaoyuchen
Hi, Team,
Great work on the model, however, is it possible to directly use the docked DNA-Protein model as input for DeepPBS to run a mutagenesis scan? I have a biological background with little computational experience, so I tried but the input format is incompatible any suggestions? And I have a 2nd question regarding retrieving the RI data shown in your figure 4f, is there any easy way to retrieve it from the final pse file or an easier way to visualize the npz file?