results: davis-shannon CV #46

results/model_media/davis/train_log/DGM_davisD_nomsaF_binaryE_64B_0.0001LR_0.4D_2000E_chemGPTLF.json

@@ -0,0 +1,221 @@
+{
+    "train_loss": [
+        1.9954961412380228,
+        1.1123259179788616,
+        1.0628154035987316,
+        1.0195641736376337,
+        0.9921986040354303,
+        0.9692184450681758,
+        0.9488197967360636,
+        0.940734565818468,
+        0.9137962934833053,
+        0.9058670597727847,
+        0.8862639353074858,
+        0.8738508635093417,
+        0.8678539732256821,
+        0.8494533173146501,
+        0.8396369166078821,
+        0.8195936667477399,
+        0.8189072838399263,
+        0.7990547935354763,
+        0.7925020887030173,
+        0.7814916852136365,
+        0.7716743027445225,
+        0.7743883602716753,
+        0.7596010671087862,
+        0.7445095050750317,
+        0.7552647592120729,
+        0.7467957078685276,
+        0.7425954988209811,
+        0.7552545020373258,
+        0.7356984313346643,
+        0.7358773813080207,
+        0.7404806569746111,
+        0.742597665621248,
+        0.7288618015735286,
+        0.7157279677128634,
+        0.7089876080574715,
+        0.7249864521807274,
+        0.7202833260994701,
+        0.7165739839995461,
+        0.7198128473854303,
+        0.7385284647065322,
+        0.7309793036150853,
+        0.7196460504958455,
+        0.7323818235880638,
+        0.7173825457807531,
+        0.7192918718256781,
+        0.722492459147707,
+        0.7188923951584549,
+        0.7100095051563288,
+        0.7068662808268471,
+        0.7108152348680042,
+        0.714867272776906,
+        0.7058873030200469,
+        0.7120912533698488,
+        0.7118631477294638,
+        0.6987585334469919,
+        0.7243739486037366,
+        0.7150540050959825,
+        0.7067155465895755,
+        0.7103592198325601,
+        0.7352949746562976,
+        0.7051169296924388,
+        0.7122998241236252,
+        0.7036621218497774,
+        0.7038248123732948,
+        0.7041913827106488,
+        0.7179739314700122,
+        0.7103856882179337,
+        0.7186218756951589,
+        0.7088906502565452,
+        0.7301404662347869,
+        0.7173615305198241,
+        0.7239452102839683,
+        0.7309273347851977,
+        0.7180232846268774,
+        0.7328828894079918,
+        0.727330649090288,
+        0.723351289419276,
+        0.7066450789405972,
+        0.7188684371242349,
+        0.7147803407459132,
+        0.7227345445086207,
+        0.7226420441658887,
+        0.7060261584933748,
+        0.7267584679977952,
+        0.7306896013628065,
+        0.7264216862767275,
+        0.7150264033652117,
+        0.7042778410116391,
+        0.7083051127290964,
+        0.7082227646595741,
+        0.712323370999533,
+        0.6989217922833245,
+        0.7070858577694908,
+        0.6963187569620467,
+        0.6941914890390054,
+        0.6874943163526137,
+        0.6820477824410902,
+        0.6814433606423899,
+        0.6704526256473191,
+        0.668551255891149,
+        0.6652297119453418,
+        0.661144870951508,
+        0.6565740831624881,
+        0.6341712325152042,
+        0.6606567606280467,
+        0.643499315848195,
+        0.6269802997976673
+    ],
+    "val_loss": [
+        1.1018557784788425,
+        1.2535118042961158,
+        1.2187856405981676,
+        1.2760470970204116,
+        1.41563035914081,
+        1.3298013821858254,
+        1.2990144251502456,
+        1.2875569346781321,
+        1.3214407840344964,
+        1.2794410279208683,
+        1.319744275546987,
+        1.3295082387811117,
+        1.2344647842681908,
+        1.2835332232072076,
+        1.0444040316799903,
+        1.0201239069589967,
+        0.8929435918643346,
+        0.7577486736679334,
+        0.7760434191739929,
+        0.7643616874784692,
+        0.7721006384292469,
+        0.7641122302501998,
+        0.7538701039368072,
+        0.7168129901525325,
+        0.7462786953706723,
+        0.7326808813011554,
+        0.7197824846747108,
+        0.7124754123768844,
+        0.6933985606487016,
+        0.6861601175938165,
+        0.6809609269401866,
+        0.6760130381557794,
+        0.6701235365950857,
+        0.6707020423568752,
+        0.6693344269013775,
+        0.6909623357809388,
+        0.682249966654432,
+        0.7903768657953931,
+        0.7202901590730896,
+        0.7126389422280862,
+        0.7491712721756998,
+        0.7206178106454126,
+        0.6651267873543619,
+        0.6644183549333561,
+        0.7078817178355118,
+        0.6839179164416824,
+        0.7086175708553498,
+        0.7268013580026418,
+        0.7167100477489783,
+        0.8461501004681222,
+        0.7416352115425207,
+        0.8179783209400103,
+        0.8254071828286963,
+        0.8371337609761188,
+        0.9208616896025605,
+        0.97088622076816,
+        0.8756023296731961,
+        1.069556474998857,
+        0.9624758456108383,
+        0.859606430698928,
+        1.0121137776074156,
+        1.0009158287358066,
+        1.0606968613440353,
+        1.2167002147937243,
+        1.5448616045502435,
+        1.4364397041036014,
+        1.4733317930053558,
+        1.7337123687354337,
+        1.7061408191262044,
+        1.7082935327713469,
+        1.3157687119529706,
+        1.5440313264941878,
+        1.3741871246334527,
+        1.2794151946750831,
+        1.316956363407178,
+        1.0836214681243033,
+        1.4886349551956606,
+        1.446724139446541,
+        1.2230758408601685,
+        1.2782749346832423,
+        0.9854454206203861,
+        1.434194633492013,
+        0.9626338091592379,
+        1.1336133684715735,
+        1.1052076671369357,
+        1.161011285033176,
+        1.1970799049420233,
+        1.0871652196598265,
+        1.3620113483547702,
+        1.1277231313020353,
+        1.308906744110637,
+        1.5659388677027506,
+        0.8891798744897103,
+        1.1205095070768643,
+        1.0892479312224972,
+        0.7972673116668036,
+        0.8973261003230031,
+        1.087875069614574,
+        1.1212975865093975,
+        1.0956825447326477,
+        0.9849263438656184,
+        0.8330175951020216,
+        0.9281886269841179,
+        1.0319483467708424,
+        0.9451814371622703,
+        1.0264084387830237,
+        1.1390326986794022
+    ],
+    "best_epoch": 44
+}

results/model_media/model_stats.csv

@@ -25,4 +25,10 @@ DGM_davis4D_nomsaF_binaryE_64B_0.0001LR_0.4D_2000E,0.8372000580825498,0.70972544
 DGM_davis1D_nomsaF_binaryE_64B_0.0001LR_0.4D_2000E,0.8259045903365158,0.6727098716351115,0.5969862332191551,0.4777166877928197,0.3703959785636244,0.6911705200547978
 DGM_davis3D_nomsaF_binaryE_64B_0.0001LR_0.4D_2000E,0.842545467746034,0.7168413490003224,0.623085263864242,0.4159492493391408,0.3558861870119901,0.6449412758842008
 DGM_davis2D_nomsaF_binaryE_64B_0.0001LR_0.4D_2000E,0.8219248012487749,0.6776357016678246,0.5900163003204868,0.4347317921133404,0.368497045519577,0.6593419386883717
-DGM_davis0D_nomsaF_binaryE_64B_0.0001LR_0.4D_2000E,0.8118490670490435,0.681294197609124,0.5718484991508404,0.42935783222492513,0.35005875178466256,0.6552540211436517
+DGM_davis0D_nomsaF_binaryE_64B_0.0001LR_0.4D_2000E,0.8118490670490435,0.681294197609124,0.5718484991508404,0.4293578322249251,0.3500587517846625,0.6552540211436517
+DGM_davisD_nomsaF_binaryE_64B_0.0001LR_0.4D_2000E_chemGPTLF,0.7370126629894659,0.4531619146470319,0.4447325194344097,0.6708082424596539,0.5559361077674945,0.8190288410426423
+DGM_davis4D_shannonF_binaryE_64B_0.0001LR_0.4D_2000E,0.8262970114209492,0.6902457009938884,0.6196677619083105,0.4657317363498777,0.3800800188713128,0.6824454090620566
+DGM_davis3D_shannonF_binaryE_64B_0.0001LR_0.4D_2000E,0.8387058574888743,0.7253796342930805,0.6397184730308061,0.418395340180838,0.3501044126973091,0.646834863145794
+DGM_davis2D_shannonF_binaryE_64B_0.0001LR_0.4D_2000E,0.8496612461728558,0.7437110771416391,0.6565426331150037,0.391104889452298,0.3245244547831619,0.6253837937237405
+DGM_davis1D_shannonF_binaryE_64B_0.0001LR_0.4D_2000E,0.8312754273934078,0.6999447666182411,0.6286994684318588,0.448984799382394,0.354222957680865,0.6700632801328499
+DGM_davis0D_shannonF_binaryE_64B_0.0001LR_0.4D_2000E,0.8164859423418647,0.693536266777011,0.6013830893679215,0.46093123143684783,0.36639131860035223,0.6789191641402147

results/model_media/model_stats_val.csv

@@ -4,3 +4,9 @@ DGM_davis0D_nomsaF_binaryE_64B_0.0001LR_0.4D_2000E,0.8124934102025569,0.69098278
 DGM_davis3D_nomsaF_binaryE_64B_0.0001LR_0.4D_2000E,0.8195586154273969,0.717774769130841,0.5720718176514452,0.3657891513377326,0.3382081135692242,0.6048050523414406
 DGM_davis2D_nomsaF_binaryE_64B_0.0001LR_0.4D_2000E,0.8233405467261639,0.7040322378359025,0.5687852322102779,0.3543421502907676,0.3433588875976263,0.5952664531877868
 DGM_davis1D_nomsaF_binaryE_64B_0.0001LR_0.4D_2000E,0.82715800662027,0.7083018179748569,0.6132060396459778,0.4895027135441312,0.3843769813986386,0.699644705221251
+DGM_davisD_nomsaF_binaryE_64B_0.0001LR_0.4D_2000E_chemGPTLF,0.7380363162506938,0.4574955465284291,0.4466001977109979,0.6649520759195586,0.551816601340746,0.8154459368465568
+DGM_davis4D_shannonF_binaryE_64B_0.0001LR_0.4D_2000E,0.823990429488312,0.7117146794067782,0.600087030044339,0.4287954343848595,0.3553209463269589,0.654824735623861
+DGM_davis3D_shannonF_binaryE_64B_0.0001LR_0.4D_2000E,0.827734580656503,0.702739642357159,0.5768866122321109,0.3738840686242872,0.3239032915296037,0.6114606026755013
+DGM_davis2D_shannonF_binaryE_64B_0.0001LR_0.4D_2000E,0.8292712666904243,0.7228505402132716,0.5833504792222683,0.3717008538318646,0.3222885614393665,0.6096727432253016
+DGM_davis1D_shannonF_binaryE_64B_0.0001LR_0.4D_2000E,0.8248946179115073,0.7385253536342127,0.598720558202774,0.404841846456497,0.3498974830004453,0.6362718337758612
+DGM_davis0D_shannonF_binaryE_64B_0.0001LR_0.4D_2000E,0.8299184455370965,0.7248253045920677,0.6116541529841135,0.4180130896587404,0.3448278840382894,0.6465393179526984

src/models/lig_mod.py

@@ -6,6 +6,8 @@
 from selfies import encoder
 
 from src.models.prior_work import DGraphDTA
+from src.models.pro_mod import EsmDTA
+
 
 class ChemDTA(DGraphDTA):
     def __init__(self, mol_output_dim=128, dropout=0.2, *args, **kwargs):
@@ -33,6 +35,54 @@ def forward_mol(self, data_mol):
         mol_x = data_mol.lig_seq
 
 
+        # get selifes from smile
+        selfies = [encoder(s) for s in mol_x]
+
+        # get tokens
+        res = self.tokenizer(selfies, return_tensors="pt", padding=True)
+
+        res['input_ids'] = res['input_ids'].to(data_mol.x.device)
+        res['attention_mask'] = res['attention_mask'].to(data_mol.x.device)
+        res['token_type_ids'] = res['token_type_ids'].to(data_mol.x.device)
+
+        # model
+        model_output = self.model(**res).last_hidden_state
+
+        # flatten to [L, 128]
+        x = torch.mean(model_output, dim=1)
+
+        x = self.relu(self.mol_fc_g1(x))
+        x = self.dropout(x)
+        x = self.mol_fc_g2(x)
+        x = self.dropout(x)
+        return x
+
+class ChemEsmDTA(EsmDTA):
+    def __init__(self, mol_output_dim=128, dropout=0.2, *args, **kwargs):
+        super(ChemEsmDTA, self).__init__(dropout=dropout, edge_weight_opt='binary', *args, **kwargs)
+
+        print('DGraphDTA Loaded')
+        num_features_mol = 128
+
+        #### ChemGPT ####
+
+        # get tokenizer and model
+        self.tokenizer = AutoTokenizer.from_pretrained("../hf_models/models--ncfrey--ChemGPT-4.7M/snapshots/7438a282460b3038e17a27e25b85b1376e9a23e2/", local_files_only=True)
+        self.model = AutoModel.from_pretrained("../hf_models/models--ncfrey--ChemGPT-4.7M/snapshots/7438a282460b3038e17a27e25b85b1376e9a23e2/", local_files_only=True)
+
+        self.model.requires_grad_(False) # freeze weights
+
+        # adding a new token '[PAD]' to the tokenizer, and then using it as the padding token
+        self.tokenizer.add_special_tokens({'pad_token': '[PAD]'})       
+
+        self.mol_fc_g1 = nn.Linear(num_features_mol, 1024)
+        self.mol_fc_g2 = nn.Linear(1024, mol_output_dim)
+
+    def forward_mol(self, data_mol):
+        # get smiles list input
+        mol_x = data_mol.lig_seq
+
+
         # get selifes from smile
         selfies = [encoder(s) for s in mol_x]
 

src/utils/config.py

@@ -6,7 +6,7 @@
 confProDy(verbosity='none') # stop printouts from prody
 
 
-MODEL_OPT = ['DG', 'DGI', 'ED', 'EDA', 'EDI', 'EDAI', 'EAT', 'CD']
+MODEL_OPT = ['DG', 'DGI', 'ED', 'EDA', 'EDI', 'EDAI', 'EAT', 'CD', 'CED']
 
 STRUCT_EDGE_OPT = ['anm', 'af2', 'af2-anm'] # edge options that require structural info (pdbs)
 EDGE_OPT = ['simple', 'binary'] + STRUCT_EDGE_OPT

src/utils/loader.py

@@ -4,7 +4,7 @@
 from torch.utils.data.distributed import DistributedSampler
 from torch_geometric.loader import DataLoader
 
-from src.models.lig_mod import ChemDTA
+from src.models.lig_mod import ChemDTA, ChemEsmDTA
 from src.models.pro_mod import EsmDTA, EsmAttentionDTA
 from src.models.prior_work import DGraphDTA, DGraphDTAImproved
 from src.data_processing.datasets import PDBbindDataset, DavisKibaDataset
@@ -99,7 +99,13 @@ def init_model(model:str, pro_feature:str, pro_edge:str, dropout:float,
         elif model == 'CD':
             # this model only needs sequence, no additional features.
             model = ChemDTA(dropout=dropout)
-
+        elif model == 'CED':
+            model = ChemEsmDTA(esm_head='facebook/esm2_t6_8M_UR50D',
+                num_features_pro=320,
+                pro_emb_dim=512, # increase embedding size
+                dropout=dropout,
+                pro_feat='esm_only',
+                edge_weight_opt=pro_edge)
         return model
 
     @staticmethod