Drug development remains a critical issue in the field of biomedicine. With the rapid advancement of information technologies such as artificial intelligence (AI) and the advent of the big data era, AI-assisted drug development has become a new trend, particularly in predicting drug-target associations. To address the challenge of drug-target prediction, AI-driven models have emerged as powerful tools, offering innovative solutions by effectively extracting features from complex biological data, accurately modeling molecular interactions, and precisely predicting potential drug-target outcomes. Traditional machine learning (ML), network-based, and advanced deep learning architectures such as convolutional neural networks (CNNs), graph convolutional networks (GCNs), and transformers play a pivotal role. This review systematically compiles and evaluates AI algorithms for drug- and drug combination-target predictions, highlighting their theoretical frameworks, strengths, and limitations. CNNs effectively identify spatial patterns and molecular features critical for drug-target interactions. GCNs provide deep insights into molecular interactions via relational data, whereas transformers increase prediction accuracy by capturing complex dependencies within biological sequences. Network-based models offer a systematic perspective by integrating diverse data sources, and traditional ML efficiently handles large datasets to improve overall predictive accuracy. Collectively, these AI-driven methods are transforming drug-target predictions and advancing the development of personalized therapy. This review summarizes the application of AI in drug development, particularly in drug-target prediction, and offers recommendations on models and algorithms for researchers engaged in biomedical research. It also provides typical cases to better illustrate how AI can further accelerate development in the fields of biomedicine and drug discovery.

RNA modifications constitute a crucial class of post-transcriptional chemical alterations that profoundly influence RNA stability and translational efficiency, thereby shaping cellular protein expression profiles. These diverse chemical marks are ubiquitously involved in key biological processes, including cell proliferation, differentiation, apoptosis, and metastatic potential, and they exert precise regulatory control over these functions. A major advance in the field is the recognition that RNA modifications do not act in isolation. Instead, they participate in complex, dynamic interactions—through synergistic enhancement, antagonism, competitive binding, and functional crosstalk—forming what is now termed the “RNA modification interactome” or “RNA modification interaction network.” The formation and functional operation of this interactome rely on a multilayered regulatory framework orchestrated by RNA-modifying enzymes—commonly referred to as “writers,” “erasers,” and “readers.” These enzymes exhibit hierarchical organization within signaling cascades, often functioning in upstream-downstream sequences and converging at critical regulatory nodes. Their integration is further mediated through shared regulatory elements or the assembly into multi-enzyme complexes. This intricate enzymatic network directly governs and shapes the interdependent relationships among various RNA modifications. This review systematically elucidates the molecular mechanisms underlying both direct and indirect interactions between RNA modifications. Building upon this foundation, we introduce novel quantitative assessment frameworks and predictive disease models designed to leverage these interaction patterns. Importantly, studies across multiple disease contexts have identified core downstream signaling axes driven by specific constellations of interacting RNA modifications. These findings not only deepen our understanding of how RNA modification crosstalk contributes to disease initiation and progression, but also highlight its translational potential. This potential is exemplified by the discovery of diagnostic biomarkers based on interaction signatures and the development of therapeutic strategies targeting pathogenic modification networks. Together, these insights provide a conceptual framework for understanding the dynamic and multidimensional regulatory roles of RNA modifications in cellular systems. In conclusion, the emerging concept of RNA modification crosstalk reveals the extraordinary complexity of post-transcriptional regulation and opens new research avenues. It offers critical insights into the central question of how RNA-modifying enzymes achieve substrate specificity—determining which nucleotides within specific RNA transcripts are selectively modified during defined developmental or pathological stages. Decoding these specificity determinants, shaped in large part by the modification interactome, is essential for fully understanding the biological and pathological significance of the epitranscriptome.

BACKGROUND: In the interim analysis of MONARCH plus, adding abemaciclib to endocrine therapy (ET) improved progression-free survival (PFS) and objective response rate (ORR) in predominantly Chinese postmenopausal women with HR+/HER2- advanced breast cancer (ABC). This study presents the final pre-planned PFS analysis. METHODS: In the phase III MONARCH plus study, postmenopausal women in China, India, Brazil, and South Africa with HR+/HER2- ABC without prior systemic therapy in an advanced setting (cohort A) or progression on prior ET (cohort B) were randomized (2:1) to abemaciclib (150 mg twice daily [BID]) or placebo plus: anastrozole (1.0 mg/day) or letrozole (2.5 mg/day) (cohort A) or fulvestrant (500 mg on days 1 and 15 of cycle 1 and then on day 1 of each subsequent cycle) (cohort B). The primary endpoint was PFS of cohort A. Secondary endpoints included cohort B PFS (key secondary endpoint), ORR, overall survival (OS), safety, and health-related quality of life (HRQoL). RESULTS: In cohort A (abemaciclib: n  = 207; placebo: n  = 99), abemaciclib plus a non-steroidal aromatase inhibitor improved median PFS vs . placebo (28.27 months vs . 14.73 months, hazard ratio [HR]: 0.476; 95% confidence interval [95% CI]: 0.348-0.649). In cohort B (abemaciclib: n  = 104; placebo: n  = 53), abemaciclib plus fulvestrant improved median PFS vs . placebo (11.41 months vs . 5.59 months, HR: 0.480; 95% CI: 0.322-0.715). Abemaciclib numerically improved ORR. Although immature, a trend toward OS benefit with abemaciclib was observed (cohort A: HR: 0.893, 95% CI: 0.553-1.443; cohort B: HR: 0.512, 95% CI: 0.281-0.931). The most frequent grade ≥3 adverse events in the abemaciclib arms were neutropenia, leukopenia, anemia (both cohorts), and lymphocytopenia (cohort B). Abemaciclib did not cause clinically meaningful changes in patient-reported global health, functioning, or most symptoms vs . placebo. CONCLUSIONS: Abemaciclib plus ET led to improvements in PFS and ORR, a manageable safety profile, and sustained HRQoL, providing clinical benefit without a high toxicity burden or reduced quality of life. TRIAL REGISTRATION ClinicalTrials.gov (NCT02763566).
Humans ; Female ; Fulvestrant/therapeutic use* ; Breast Neoplasms/metabolism* ; Aminopyridines/therapeutic use* ; Benzimidazoles/therapeutic use* ; Middle Aged ; Aromatase Inhibitors/therapeutic use* ; Aged ; Receptor, ErbB-2/metabolism* ; Adult ; Letrozole/therapeutic use* ; Antineoplastic Combined Chemotherapy Protocols/therapeutic use* ; Anastrozole/therapeutic use*

Transarterial radioembolization (TARE) is a widely utilized therapeutic approach for hepatocellular carcinoma (HCC), however, the clinical implementation is constrained by the stringent preparation conditions of radioembolization agents. Herein, we incorporated the superstable homogeneous iodinated formulation technology (SHIFT), simultaneously utilizing an enhanced solvent form in a carbon dioxide supercritical fluid environment, to encapsulate radionuclides (such as ¹³¹I,¹⁷⁷Lu, or ¹⁸F) with lipiodol for the preparation of radiolipiodol. The resulting radiolipiodol exhibited exceptional stability and ultra-high labeling efficiency (≥99%) and displayed notable intratumoral radionuclide retention and in vivo stability more than 2 weeks following locoregional injection in subcutaneous tumors in mice and orthotopic liver tumors in rats and rabbits. Given these encouraging findings, ¹⁸F was authorized as a radiotracer in radiolipiodol for clinical trials in HCC patients, and showed a favorable tumor accumulation, with a tumor-to-liver uptake ratio of ≥50 and minimal radionuclide leakage, confirming the feasibility of SHIFT for TARE applications. In the context of transforming from preclinical to clinical screening, the preparation of radiolipiodol by SHIFT represents an innovative physical strategy for radionuclide encapsulation. Hence, this work offers a reliable and efficient approach for TARE in HCC, showing considerable promise for clinical application (ChiCTR2400087731).

OBJECTIVES: We propose a multi-feature fusion model based on manually extracted features and deep learning features from endoscopic images for grading rebleeding risk of peptic ulcers. METHODS: Based on the endoscopic appearance of peptic ulcers, color features were extracted to distinguish active bleeding (Forrest I) from non-bleeding ulcers (Forrest II and III). The edge and texture features were used to describe the morphology and appearance of the ulcers in different grades. By integrating deep features extracted from a deep learning network with manually extracted visual features, a multi-feature representation of endoscopic images was created to predict the risk of rebleeding of peptic ulcers. RESULTS: In a dataset consisting of 3573 images from 708 patients with Forrest classification, the proposed multi-feature fusion model achieved an accuracy of 74.94% in the 6-level rebleeding risk classification task, outperforming the experienced physicians who had a classification accuracy of 59.9% (P<0.05). The F1 scores of the model for identifying Forrest Ib, IIa, and III ulcers were 90.16%, 75.44%, and 77.13%, respectively, demonstrating particularly good performance of the model for Forrest Ib ulcers. Compared with the first model for peptic ulcer rebleeding classification, the proposed model had improved F1 scores by 5.8%. In the simplified 3-level risk (high-risk, low-risk, and non-endoscopic treatment) classification task, the model achieved F1 scores of 93.74%, 81.30%, and 73.59%, respectively. CONCLUSIONS The proposed multi-feature fusion model integrating deep features from CNNs with manually extracted visual features effectively improves the accuracy of rebleeding risk classification for peptic ulcers, thus providing an efficient diagnostic tool for clinical assessment of rebleeding risks of peptic ulcers.
Humans ; Deep Learning ; Peptic Ulcer ; Risk Assessment ; Peptic Ulcer Hemorrhage ; Recurrence