The rationale for using ADMET prediction tools in the early drug discovery paradigm is to guide the design of new compounds with favorable ADMET properties and ultimately minimize the attrition rates of drug failures. Artificial intelligence (AI) in in silico ADMET modeling has gained momentum due to its high-throughput and low-cost attributes. In this study, we developed a machine learning model capable of predicting 11 ADMET properties of chemical compounds. Each model was constructed by combining one of 40 classification algorithms including random forest (RF), extreme gradient boosting (XGB), support vector machine (SVM), and gradient boosting (GB) with one of three predefined hyperparameter configurations. This process can be efficiently performed using automated machine learning (AutoML) methods, which automatically search for the best combination of model algorithms and optimized hyperparameters. We developed optimal predictive models for 11 different ADMET properties using the Hyperopt-sklearn AutoML method. All of the developed models depicted an area under the ROC curve (AUC) >0.8. Furthermore, our developed models outperformed most of the ADMET properties and showed comparable performance in other properties when evaluated on external data sets and compared with published predictive models. Our results support the applicability of AutoML in ADMET prediction and will be helpful for ADMET prediction in early-stage drug discovery.

Mitochondrial damage induced by chemotherapeutic agents through disruption of the mitochondrial membrane potential (Δ Ψ _m ) remains a central challenge in drug development and evaluation. However, the assessment of Δ Ψ _m ‐targeting drugs using commercially available fluorescent probes is often unreliable, as these dyes can interfere with, mask, or artificially amplify drug‐induced mitochondrial dysfunction, frequently resulting in misleading conclusions and translational failure. Herein, we report a class of cationic chemotherapeutic small molecules ( DPPs ) possessing intrinsic fluorescence migration‐based self‐reporting capability, which enables direct and non‐invasive monitoring of drug action without the need for external probes. Among them, DPP‐1 and DPP‐2 disrupt mitochondrial function, trigger excessive reactive oxygen species generation, and induce highly selective apoptosis. Remarkably, both compounds exhibit concentration‐dependent mitochondrial‐to‐nuclear translocation, enabling the real‐time visualization of therapeutic progression at the subcellular level. In vivo studies further confirm their potent tumor growth inhibition and negligible systemic toxicity effects. This self‐reporting mitochondria‐targeted chemotherapeutic platform provides a highly promising strategy for integrated cancer diagnosis and precision therapy.

Membrane proteins, including small four-transmembrane proteins, play roles in membrane organization and signal transduction. Generally, they share capacities for protein-protein association and subcellular translocation to regulate cellular functions. Despite known commonalities and differences in structure and function, the relationships among the tetraspanin, CD20-like, golgi 4-transmembrane spanning transporter (Mtp), and transmembrane 4 L six (L6) families remain unclear. Here, we mostly compared TM4SF5, a transmembrane 4 L6 family member, to the genuine tetraspanins and the CD20-like family, exploring unique aspects that evolved from the small four-transmembrane proteins. TM4SF5 and other L6 membrane proteins have two conserved cysteines, affecting N-glycosylation in the large extracellular loop (LEL) and favoring a dimeric state, whereas the tetraspanins maintained their cysteine residues in addition to the CCG motif. However, TM4SF5 and other L6 membrane proteins have more sequence similarity to proteins in the CD20-like family, which lack these features in the LEL. Uniquely, TM4SF5 and other L6 membrane proteins have a β-sheet configuration in the LEL, unlike other small four-transmembrane proteins with α-helices. Therefore, the unique features in the LEL and the N-glycosylation of TM4SF5 may contribute to dimer formation and cholesterol binding, potentially influencing regulatory roles in liver malignancy development.

Mitochondrial damage induced by chemotherapeutic agents through disruption of the mitochondrial membrane potential (Δ Ψ _m ) remains a central challenge in drug development and evaluation. However, the assessment of Δ Ψ _m ‐targeting drugs using commercially available fluorescent probes is often unreliable, as these dyes can interfere with, mask, or artificially amplify drug‐induced mitochondrial dysfunction, frequently resulting in misleading conclusions and translational failure. Herein, we report a class of cationic chemotherapeutic small molecules ( DPPs ) possessing intrinsic fluorescence migration‐based self‐reporting capability, which enables direct and non‐invasive monitoring of drug action without the need for external probes. Among them, DPP‐1 and DPP‐2 disrupt mitochondrial function, trigger excessive reactive oxygen species generation, and induce highly selective apoptosis. Remarkably, both compounds exhibit concentration‐dependent mitochondrial‐to‐nuclear translocation, enabling the real‐time visualization of therapeutic progression at the subcellular level. In vivo studies further confirm their potent tumor growth inhibition and negligible systemic toxicity effects. This self‐reporting mitochondria‐targeted chemotherapeutic platform provides a highly promising strategy for integrated cancer diagnosis and precision therapy.

Nannocystin A is a macrocyclic natural product with potent cytotoxic activity and a reported mechanism involving eukaryotic elongation factor 1A (eEF1A). Here, we investigated the structure–activity relationships of the nannocystin-derived macrocyclic framework by integrating chemical synthesis, cytotoxicity assays, molecular docking, and in silico prediction of cellular distribution. Evaluation of analogues 1–6, bearing modifications at distinct regions of the macrocycle, revealed differential effects on cellular activity, predicted cellular distribution, and compatibility with the proposed eEF1A-binding site, identifying the C13 side chain as a tractable position for further modification. A focused series of C13-modified analogues 7–12 was therefore synthesized and evaluated. Among these, the homoleucine analogue 10 displayed the most favorable activity profile, retaining sub-nanomolar cytotoxicity toward SMMC7721 and Hep3B cells (IC₅₀ = 0.52 and 0.53 nM, respectively). Computational analysis suggested that the favorable profile of 10 may reflect an appropriate balance between predicted cellular access and compatibility with the proposed eEF1A-binding site, whereas bulkier or more hydrophobic C13 substituents were associated with increased predicted membrane retention and/or altered binding orientations. These findings identify C13 as a tunable region of the nannocystin-derived macrocyclic framework and provide a structure–activity framework for further optimization of eEF1A-targeting nannocystin analogues.