Skip to content

Added method for predicting directly from sequence with contact map #16

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
wants to merge 3 commits into
base: master
Choose a base branch
from
Open

Added method for predicting directly from sequence with contact map #16

Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
aab24d2
added predict_with_cmap method
SoliareofAstora Nov 25, 2021
e03290f
sequence prediction using sequence only
SoliareofAstora Dec 25, 2021
25fc53f
prediction data dumping hotfix
SoliareofAstora Dec 25, 2021
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
40 changes: 40 additions & 0 deletions deepfrier/Predictor.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -54,6 +54,8 @@ def __init__(self, model_prefix, gcn=True):
self.model_prefix = model_prefix
self.gcn = gcn
self._load_model()
self.prot2goterms = {}
self.goidx2chains = {}

def _load_model(self):
self.model = tf.keras.models.load_model(self.model_prefix + '.hdf5',
Expand Down Expand Up @@ -130,6 +132,44 @@ def predict(self, test_prot, cmap_thresh=10.0, chain='query_prot'):
self.goidx2chains[idx].add(chain)
self.prot2goterms[chain].append((self.goterms[idx], self.gonames[idx], float(y[idx])))

def predict_with_cmap(self, seqres, cmap, chain):
self.Y_hat = np.zeros((1, len(self.goterms)), dtype=float)
self.data = {}
self.test_prot_list = [chain]
if self.gcn:
S = seq2onehot(seqres)
S = S.reshape(1, *S.shape)
A = cmap
y = self.model([A, S], training=False).numpy()[:, :, 0].reshape(-1)
self.Y_hat[0] = y
self.prot2goterms[chain] = []
self.data[chain] = [[A, S], seqres]
go_idx = np.where((y >= self.thresh) == True)[0]
for idx in go_idx:
if idx not in self.goidx2chains:
self.goidx2chains[idx] = set()
self.goidx2chains[idx].add(chain)
self.prot2goterms[chain].append((self.goterms[idx], self.gonames[idx], float(y[idx])))

def predict_from_sequence(self, sequence, chain):
self.test_prot_list = [chain]
self.Y_hat = np.zeros((len(self.test_prot_list), len(self.goterms)), dtype=float)
self.data = {}

for i, chain in enumerate(self.test_prot_list):
S = seq2onehot(str(sequence))
S = S.reshape(1, *S.shape)
y = self.model(S, training=False).numpy()[:, :, 0].reshape(-1)
self.Y_hat[i] = y
self.prot2goterms[chain] = []
self.data[chain] = [[S], sequence]
go_idx = np.where((y >= self.thresh) == True)[0]
for idx in go_idx:
if idx not in self.goidx2chains:
self.goidx2chains[idx] = set()
self.goidx2chains[idx].add(chain)
self.prot2goterms[chain].append((self.goterms[idx], self.gonames[idx], float(y[idx])))

def predict_from_PDB_dir(self, dir_name, cmap_thresh=10.0):
print ("### Computing predictions from directory with PDB files...")
pdb_fn_list = glob.glob(dir_name + '/*.pdb*')
Expand Down