use precomputed embeddings

app.py

@@ -61,10 +61,16 @@ def display_dti():
                 'Select encoder for drug compound',('None', 'CDDD', 'MolBERT')
             )
             if selected_encoder == 'CDDD':
-                from cddd.inference import InferenceModel
-                CDDD_MODEL_DIR = 'src/encoders/cddd'
-                cddd_model = InferenceModel(CDDD_MODEL_DIR)
-                embedding = cddd_model.seq_to_emb([smiles])
+                #from cddd.inference import InferenceModel
+                #CDDD_MODEL_DIR = 'src/encoders/cddd'
+                #cddd_model = InferenceModel(CDDD_MODEL_DIR)
+                #embedding = cddd_model.seq_to_emb([smiles])
+                from huggingface_hub import hf_hub_download
+                precomputed_embs = f'{selected_encoder}_encoding.csv'
+                REPO_ID = "emmas96/Lenselink"
+                embs_path = hf_hub_download(REPO_ID, precomputed_embs)
+                embs = pd.read_csv(embs_path)
+                embedding = embs[smiles]
             elif selected_encoder == 'MolBERT':
                 from molbert.utils.featurizer.molbert_featurizer import MolBertFeaturizer
                 from huggingface_hub import hf_hub_download
@@ -91,10 +97,16 @@ def display_dti():
                 'Select encoder for protein target',('None', 'SeqVec', 'UniRep', 'ESM-1b', 'ProtT5')
             )
             if selected_encoder == 'SeqVec':
-                from bio_embeddings.embed import SeqVecEmbedder
-                encoder = SeqVecEmbedder()
-                embedding = encoder([sequence])
-                embedding = encoder.reduce_per_protein(embedding)
+                #from bio_embeddings.embed import SeqVecEmbedder
+                #encoder = SeqVecEmbedder()
+                #embedding = encoder([sequence])
+                #embedding = encoder.reduce_per_protein(embedding)
+                from huggingface_hub import hf_hub_download
+                precomputed_embs = f'{selected_encoder}_encoding.csv'
+                REPO_ID = "emmas96/Lenselink"
+                embs_path = hf_hub_download(REPO_ID, precomputed_embs)
+                embs = pd.read_csv(embs_path)
+                embedding = embs[sequence]
             elif selected_encoder == 'UniRep':
                 from jax_unirep.utils import load_params
                 params = load_params()