Computing Dynamical Signatures

To compute the ENCoM Dynamical Signatures from the list of mutated PDB files, we need to build a DynaSigDF (“Dynamical Signature Data Frame”) object. The DynaSigDF constructor needs the list of files, and a matching list of experimental measures (what the machine learning models will try to predict later on). For the miR-125a dataset, the maturatino efficiency measure along with the MC-Fold enthalpy of folding are stored in the data_mir125.df file.

from dynasigml.dynasig_df import DynaSigDF
import glob
import json

def load_data(filename):
with open(filename) as f:
    lines = f.readlines()
data_dict = dict()
for line in lines[1:]:
    ll = line.split()
    data_dict[ll[2]] = [float(ll[0]), float(ll[1])]
return data_dict


files_list = glob.glob('mir125a_variants/*.pdb')
beta_values = [np.e ** (x / 2) for x in range(-6, 7)]
data_dict = load_data('data_mir125.df')
exp_data = []
for fn in files_list:
    mutid = fn.split('.')[0].split('mir125a_')[-1]
    exp_data.append(data_dict[mutid])
# eff is for maturation efficiency
DynaSigDF(files_list, exp_data, ["eff", "mcfold_energy"], "combined_dsdf".format(index), beta_values=beta_values)

Note

The above code will take between 24 and 48 hours to run on a modern laptop (without parallelization). However, you do not need to run it as the saved DynaSigDF objects that contain every Dynamical Signature are part of the dynasigml_mir125a_example repository. For more details on how to run the computations yourself, including parallelization, see the Running in parallel section.