Multi-target compounds, or polypharmacological agents, hold significant potential for complex diseases like cancer, where single-target therapies are often insufficient. A lack of high-quality bioactivity data limits progress in this field, especially for compounds interacting with multiple proteins simultaneously. This study introduces MT-ConBiFormer-GPT, a deep generative model designed explicitly for low-data, multi-target molecular generation, focusing on the critical PI3K-AKT-mTOR cancer signaling pathway. The framework integrates a variational autoencoder with a BiFormer encoder to capture long-range dependencies in SMILES strings, reducing the quadratic computational complexity associated with standard transformers and mitigating semantic discontinuities. It employs a SMILES-GPT decoder for progressive molecule generation and follows a three-phase training pipeline: unsupervised pre-training, supervised contrastive learning, and curriculum-based fine-tuning. The framework's efficacy was evaluated through a rigorous, multi-stage assessment. First, the framework was evaluated through benchmarking against state-of-the-art models, with a specialized head-to-head variant, MT-ConBiFormer-GPT_H2H, demonstrating superior performance, thereby validating its generalizability from oncology to neuropsychiatry. An internal ablation study further revealed that the full MT-ConBiFormer-GPT significantly outperformed its baseline, MT-BiFormer-GPT, in both dual- and triplet-target generation tasks, highlighting the advantages of the contrastive learning stage. Additionally, the foundational Base-BiFormer-GPT architecture, a model lacking both the contrastive and curriculum learning stages, highlighted its intrinsic robustness by achieving competitive outcomes in a distinct omics-driven design task. Docking simulations and mechanistic analyses show that the generated molecules, including high-fidelity and scaffold-hopping candidates, display more favorable binding modes than reference inhibitors. This study presents a flexible and computationally efficient framework for multi-target drug discovery in data-limited settings.

BACKGROUND: The Drug-Target Interaction (DTI) prediction uses a drug molecule and a protein sequence as inputs to predict the binding affinity value. In recent years, deep learning-based models have gotten more attention. These methods have two modules: the feature extraction module and the task prediction module. In most deep learning-based approaches, a simple task prediction loss (i.e., categorical cross entropy for the classification task and mean squared error for the regression task) is used to learn the model. In machine learning, contrastive-based loss functions are developed to learn more discriminative feature space. In a deep learning-based model, extracting more discriminative feature space leads to performance improvement for the task prediction module. RESULTS: In this paper, we have used multimodal knowledge as input and proposed an attention-based fusion technique to combine this knowledge. Also, we investigate how utilizing contrastive loss function along the task prediction loss could help the approach to learn a more powerful model. Four contrastive loss functions are considered: (1) max-margin contrastive loss function, (2) triplet loss function, (3) Multi-class N-pair Loss Objective, and (4) NT-Xent loss function. The proposed model is evaluated using four well-known datasets: Wang et al. dataset, Luo's dataset, Davis, and KIBA datasets. CONCLUSIONS: Accordingly, after reviewing the state-of-the-art methods, we developed a multimodal feature extraction network by combining protein sequences and drug molecules, along with protein-protein interaction networks and drug-drug interaction networks. The results show it performs significantly better than the comparable state-of-the-art approaches.

BACKGROUND: Due to the high resource consumption of introducing a new drug, drug repurposing plays an essential role in drug discovery. To do this, researchers examine the current drug-target interaction (DTI) to predict new interactions for the approved drugs. Matrix factorization methods have much attention and utilization in DTIs. However, they suffer from some drawbacks. METHODS: We explain why matrix factorization is not the best for DTI prediction. Then, we propose a deep learning model (DRaW) to predict DTIs without having input data leakage. We compare our model with several matrix factorization methods and a deep model on three COVID-19 datasets. In addition, to ensure the validation of DRaW, we evaluate it on benchmark datasets. Furthermore, as an external validation, we conduct a docking study on the COVID-19 recommended drugs. RESULTS: In all cases, the results confirm that DRaW outperforms matrix factorization and deep models. The docking results approve the top-ranked recommended drugs for COVID-19. CONCLUSIONS: In this paper, we show that it may not be the best choice to use matrix factorization in the DTI prediction. Matrix factorization methods suffer from some intrinsic issues, e.g., sparsity in the domain of bioinformatics applications and fixed-unchanged size of the matrix-related paradigm. Therefore, we propose an alternative method (DRaW) that uses feature vectors rather than matrix factorization and demonstrates better performance than other famous methods on three COVID-19 and four benchmark datasets.

BACKGROUND: Drug-target interaction (DTI) plays a vital role in drug discovery. Identifying drug-target interactions related to wet-lab experiments are costly, laborious, and time-consuming. Therefore, computational methods to predict drug-target interactions are an essential task in the drug discovery process. Meanwhile, computational methods can reduce search space by proposing potential drugs already validated on wet-lab experiments. Recently, deep learning-based methods in drug-target interaction prediction have gotten more attention. Traditionally, DTI prediction methods' performance heavily depends on additional information, such as protein sequence and molecular structure of the drug, as well as deep supervised learning. RESULTS: This paper proposes a method based on deep unsupervised learning for drug-target interaction prediction called AutoDTI++. The proposed method includes three steps. The first step is to pre-process the interaction matrix. Since the interaction matrix is sparse, we solved the sparsity of the interaction matrix with drug fingerprints. Then, in the second step, the AutoDTI approach is introduced. In the third step, we post-preprocess the output of the AutoDTI model. CONCLUSIONS: Experimental results have shown that we were able to improve the prediction performance. To this end, the proposed method has been compared to other algorithms using the same reference datasets. The proposed method indicates that the experimental results of running five repetitions of tenfold cross-validation on golden standard datasets (Nuclear Receptors, GPCRs, Ion channels, and Enzymes) achieve good performance with high accuracy.

BACKGROUND: Moonlighting proteins (MPs) are a subclass of multifunctional proteins in which more than one independent or usually distinct function occurs in a single polypeptide chain. Identification of unknown cellular processes, understanding novel protein mechanisms, improving the prediction of protein functions, and gaining information about protein evolution are the main reasons to study MPs. They also play an important role in disease pathways and drug-target discovery. Since detecting MPs experimentally is quite a challenge, most of them are detected randomly. Therefore, introducing an appropriate computational approach to predict MPs seems reasonable. RESULTS: In this study, we introduced a competent model for detecting moonlighting and non-MPs through extracted features from protein sequences. We attempted to set up a well-judged scheme for detecting outlier proteins. Consequently, 37 distinct feature vectors were utilized to study each protein's impact on detecting MPs. Furthermore, 8 different classification methods were assessed to find the best performance. To detect outliers, each one of the classifications was executed 100 times by tenfold cross-validation on feature vectors; proteins which misclassified 90 times or more were grouped. This process was applied to every single feature vector and eventually the intersection of these groups was determined as the outlier proteins. The results of tenfold cross-validation on a dataset of 351 samples (containing 215 moonlighting and 136 non-moonlighting proteins) reveal that the SVM method on all feature vectors has the highest performance among all methods in this study and other available methods. Besides, the study of outliers showed that 57 of 351 proteins in the dataset could be an appropriate candidate for the outlier. Among the outlier proteins, there were non-MPs (such as P69797) that have been misclassified in 8 different classification methods with 16 different feature vectors. Because these proteins have been obtained by computational methods, the results of this study could reduce the likelihood of hypothesizing whether these proteins are non-moonlighting at all. CONCLUSIONS: MPs are difficult to be identified through experimentation. Using distinct feature vectors, our method enabled identification of novel moonlighting proteins. The study also pinpointed that a number of non-MPs are likely to be moonlighting.