The simultaneous use of two or more drugs in clinical treatment may raise the risk of a drug–drug interaction (DDI). DDI prediction is very important to avoid adverse drug events in combination therapy. Recently, deep learning methods have been applied successfully to DDI prediction and improved prediction performance. However, there are still some problems with the present models, such as low accuracy due to information loss during molecular representation or incomplete drug feature mining during the training process. Aiming at these problems, this study proposes an accurate and simple framework named PTB-DDI for drug–drug interaction prediction. The PTB-DDI framework consists of four key modules: (1) ChemBerta tokenizer for molecular representation, (2) Bidirectional Long Short-Term Memory (BiLSTM) to capture the bidirectional context-aware features of drugs, (3) Multilayer Perceptron (MLP) for mining the nonlinear relationship of drug features, and (4) interaction predictor to perform an affine transformation and final prediction. In addition, we investigate the effect of dual-mode on parameter-sharing and parameter-independent within the PTB-DDI framework. Furthermore, we conducted comprehensive experiments on the two real-world datasets (i.e., BIOSNAP and DrugBank) to evaluate PTB-DDI framework performance. The results show that our proposed framework has significant improvements over the baselines based on both datasets. Based on the BIOSNAP dataset, the AUC-ROC, PR-AUC, and F1 scores are 0.997, 0.995, and 0.984, respectively. These metrics are 0.896, 0.873, and 0.826 based on the DrugBank dataset. Then, we conduct the case studies on the three newly approved drugs by the Food and Drug Administration (FDA) in 2024 using the PTB-DDI framework in dual modes. The obtained results indicate that our proposed framework has advantages for predicting drug–drug interactions and that the dual modes of the framework complement each other. Furthermore, a free website is developed to enhance accessibility and user experience.
Prostate Cancer (PCa) is the sixth most common and the second deadliest cancer among men in the world. While the diagnosis of PCa mostly depends on the decisions of pathologists based on clinical symptoms of the patients and their careful observations upon the patients' prostate biopsy, in particular, the biopsy of the patients' glands. Often there are disagreements and debates among pathologists in deciding the Gleason grade of the same biopsy samples. Majority votes take place to avoid disputes. In this research, we aim to find an automatic method to facilitate the process of determining Gleason grade of PCa. In order to achieve that, we explore both Support Vector Machine (SVM) and Convolutional Neural Networks (CNNs) in order to determine whether the local patches of the Tissue Micro-Array (TMA) images contain glands, border region of glands, or stroma, and it turns out that with limited amount of labelled data, SVM outperforms the other. With the results from classifiers, we can filter irrelative portions of images. This processed TMA images would allow pathologists to focus on the locations of glands existing potentially. Moreover, with processed images, the next level of training – predicting the Gleason grade of PCa and generating corresponding images under certain Gleason grade – using models like CNNs and Generative Adversarial Networks (GANs) is expected to be much faster.
Human African trypanosomiasis (HAT), caused by Trypanosoma brucei , remains a neglected tropical disease with a critical shortage of therapeutic options, underscoring the need for new chemotypes. Cell division cycle‐2‐related kinase 12 (CRK12) has emerged as a genetically essential and chemically validated target in kinetoplastids, yet experimentally supported CRK12‐targeting chemotypes against T. brucei remain limited, and no experimental CRK12 structure is available. Here, we developed an integrated computational‐experimental screening workflow to accelerate CRK12‐guided hit discovery, combining ligand‐based prescreening, structure‐based virtual screening, molecular dynamics (MD) refinement, and experimental validation. An 8.6‐million‐compound library was prescreened using MACCS fingerprint screening and complex‐based pharmacophore screening, followed by staged docking and MD simulations to prioritize candidates for experimental testing. Notably, four of ten purchased compounds showed strong growth inhibition against T. brucei at 10 μM, with IC 50 values of 6.09, 1.47, 0.81, and 1.33 μM, demonstrating a promising hit rate for this data‐limited target. Binding‐mode analysis revealed a conserved hinge‐anchoring interaction pattern in the ATP‐binding pocket, providing a structural rationale for follow‐up analogue design. Overall, this study identifies new chemical starting points for anti‐ T. brucei drug discovery and demonstrates the utility of an AlphaFold2‐enabled CRK12‐guided screening strategy for targets with limited structural and ligand data.