MDDI-SCL: predicting multi-type drug-drug interactions via supervised contrastive learning

The joint use of multiple drugs may cause unintended drug-drug interactions (DDIs) and result in adverse consequence to the patients. Accurate identification of DDI types can not only provide hints to avoid these accidental events, but also elaborate the underlying mechanisms by how DDIs occur. Several computational methods have been proposed for multi-type DDI prediction, but room remains for improvement in prediction performance. In this study, we propose a supervised contrastive learning based method, MDDI-SCL, implemented by three-level loss functions, to predict multi-type DDIs. MDDI-SCL is mainly composed of three modules: drug feature encoder and mean squared error loss module, drug latent feature fusion and supervised contrastive loss module, multi-type DDI prediction and classification loss module. The drug feature encoder and mean squared error loss module uses self-attention mechanism and autoencoder to learn drug-level latent features. The drug latent feature fusion and supervised contrastive loss module uses multi-scale feature fusion to learn drug pair-level latent features. The prediction and classification loss module predicts DDI types of each drug pair. We evaluate MDDI-SCL on three different tasks of two datasets. Experimental results demonstrate that MDDI-SCL achieves better or comparable performance as the state-of-the-art methods. Furthermore, the effectiveness of supervised contrastive learning is validated by ablation experiment, and the feasibility of MDDI-SCL is supported by case studies. The source codes are available at https://github.com/ShenggengLin/MDDI-SCL. Supplementary Information The online version contains supplementary material available at 10.1186/s13321-022-00659-8.


Autoencoder
Autoencoder is an unsupervised neural network model, which includes two parts: encoder and decoder. The function of the encoder is to encode high-dimensional features into low-dimensional latent features to let the neural network to learn the most informative features. The objective of a decoder is to map back the latent features into a reconstruction of the original input. The best scenario is that the output of the decoder can perfectly or approximately approach the original input. The autoencoder with single linear layer is calculated by following formulas, , where � �푛� � is the input feature vector of the autoencoder, and � 푒푛 and � 푒푛 are the parameters of the encoder. � is the activation function. 퐹푒푎� 푒 ℎ� 푒푛 is the latent feature vector. � 푒 and � 푒 are the parameters of the decoder.
� �푛� � is the output of the decoder.

Memantine Isoniazid
Isoniazid may decrease the excretion rate of Memantine which could result in a higher serum level.

Cilostazol Clonidine
The metabolism of Clonidine can be decreased when combined with Cilostazol.

Cyclosporine Memantine
Cyclosporine may decrease the excretion rate of Memantine which could result in a higher serum level.

Gemfibrozil Lobeglitazone
The metabolism of Lobeglitazone can be decreased when combined with Gemfibrozil.

Miconazole Ifosfamide
The metabolism of Ifosfamide can be decreased when combined with Miconazole.

Dosulepin Lobeglitazone
Dosulepin may decrease the hypoglycemic activities of Lobeglitazone.

Isoniazid Celecoxib
The metabolism of Celecoxib can be decreased when combined with Isoniazid.

Mycophenolic acid Atomoxetine
Atomoxetine may decrease the excretion rate of Mycophenolic acid which could result in a higher serum level.

Ketoconazole Lobeglitazone
The metabolism of Lobeglitazone can be decreased when combined with Ketoconazole.

Nicardipine Lobeglitazone
The metabolism of Lobeglitazone can be decreased when combined with Nicardipine.

Dabigatran etexilate
Brivaracetam Brivaracetam may decrease the excretion rate of Dabigatran etexilate which could result in a higher serum level.

Miconazole Doconexent
The metabolism of Doconexent can be decreased when combined with Miconazole.

Isoniazid Dosulepin
The risk or severity of serotonin syndrome can be increased when Isoniazid is combined with Dosulepin.

Carbinoxamine Cariprazine
The risk or severity of adverse effects can be increased when Carbinoxamine is combined with Cariprazine.

Imipramine Methylergometrine
The risk or severity of hypertension can be increased when Methylergometrine is combined with Imipramine.

Aranidipine Amylnitrite
Aranidipine may increase the vasodilatory activities of Amyl Nitrite.

Bortezomib Hydroxyurea
The risk or severity of adverse effects can be increased when Bortezomib is combined with Hydroxyurea.

Arsenic trioxide Naldemedine
Arsenic trioxide may decrease the excretion rate of Naldemedine which could result in a higher serum level.

Etacrynicacid Diazoxide
The risk or severity of adverse effects can be increased when Etacrynic acid is combined with Diazoxide.

Lidocaine Artenimol
The metabolism of Lidocaine can be decreased when combined with Artenimol. Debrisoquine Artenimol The metabolism of Debrisoquine can be decreased when combined with Artenimol.

Daunorubicin Alfuzosin
The metabolism of Alfuzosin can be decreased when combined with Daunorubicin.

Epinephrine Mifepristone
The therapeutic efficacy of Mifepristone can be decreased when used in combination with Epinephrine.

Mycophenolic acid Bosutinib
The risk or severity of adverse effects can be increased when Mycophenolic acid is combined with Bosutinib.

Doconexent Isavuconazole
The metabolism of Doconexent can be decreased when combined with Isavuconazole.

Hydroxychloroquine Cilostazol
The risk or severity of QTc prolongation can be increased when Cilostazol is combined with Hydroxychloroquine.

Lorpiprazole Benzatropine
Benzatropine may decrease the excretion rate of Lorpiprazole which could result in a higher serum level.

Granisetron Brexpiprazole
The risk or severity of adverse effects can be increased when Granisetron is combined with Brexpiprazole.

Fexofenadine Eltrombopag
The excretion of Fexofenadine can be decreased when combined with Eltrombopag.

Imipramine Artenimol
The metabolism of Imipramine can be decreased when combined with Artenimol. Fluvoxamine Acetazolamide The risk or severity of adverse effects can be increased when Acetazolamide is combined with Fluvoxamine.

Agomelatine
Cyproterone acetate The metabolism of Agomelatine can be increased when combined with Cyproterone acetate.

Mycophenolic acid Apalutamide
Mycophenolic acid may decrease the excretion rate of Apalutamide which could result in a higher serum level.

Lobeglitazone Lumacaftor
The metabolism of Lobeglitazone can be increased when combined with Lumacaftor.

Danazol Bosentan
The metabolism of Bosentan can be decreased when combined with Danazol.