Due to its high prevalence, screening for hepatic fibrosis in the low-risk population is called for action in the primary care clinic. However, current guidelines provide conflicting recommendations on populations to be screened. We aimed to identify the target populations that would most benefit from screening for hepatic fibrosis in clinical practice. This study examined 1288 subjects who underwent magnetic resonance elastography. The diagnostic performance of the Fibrosis-4 (FIB-4) index and NAFLD fibrosis score was compared in the following groups: (1) ultrasonography (USG)-diagnosed NAFLD, (2) elevated liver enzyme, (3) metabolic syndrome, (4) impaired fasting glucose, and (5) type 2 diabetes regardless of fatty liver. Decision curve analysis was performed to express the net benefit of groups over a range of probability thresholds (Pts). The diabetes group showed a better area under the receiver operating characteristic curve (AUROC: 0.69) compared with subjects in the USG-diagnosed NAFLD (AUROC: 0.57) and elevated liver enzyme (AUROC: 0.55) groups based on the FIB-4 index. In decision curve analysis, the diabetes group showed the highest net benefit for the detection of significant fibrosis across a wide range of Pts. Patients with diabetes, even in the absence of fatty liver, would be preferable for hepatic fibrosis screening in low-risk populations.

Background/Aims: The 2030 hepatitis C virus (HCV) elimination targets of the World Health Organization are an 80% reduction in incidence and 65% reduction in mortality compared to the 2015 rates. However, information on the nationwide incidence and treatment rates of HCV infection are limited. We aimed to investigate the nationwide incidence and status of the care cascade for HCV infection in Korea.Methods: This study used data from the Korea Disease Control and Prevention Agency linked with the data of the Korea National Health Insurance Service. Linkage to care was defined as visiting hospitals twice or more due to HCV infection within 1.5 years of the index date. The treatment rate was the number who had been prescribed antiviral medication within 1.5 years from the index date out of patients newly diagnosed with HCV. Results:The new HCV infection rate was 17.2 per 100,000 person-years (n=8,810) in 2019. The number of new HCV infections was the highest in patients aged 50 to 59 years (n=2,480), and the new HCV infection rate significantly increased with age (p<0.001). Among newly infected patients with HCV, the linkage to care rate was 78.2% (78.2% men, 78.2% women) and the treatment rate was 58.1% (56.8% men, 59.3% women) within 1.5 years. Conclusions:The new HCV infection rate was 17.2 per 100,000 person-years in Korea. It is necessary to continuously monitor the incidence and care cascade of HCV to establish proper strategies to reach the goal of HCV elimination by 2030.

Recent advances in artificial intelligence (AI) have led to the development of transformer-based models that have shown success in identifying potential drug molecules for therapeutic purposes. However, for a molecule to be considered a viable drug candidate, it must exhibit certain desirable properties such as low toxicity, high druggability, and synthesizability. To address this, we propose an approach that incorporates prior knowledge about these properties during the model training process. In this study, we utilized the PubChem database, which contains 100 million molecules, to filter drug-like molecules based on the quantity of drug-likeliness (QED) score and the Pfizer rule. We then used this filtered dataset of drug-like molecules to train both molecular representation (ChemBERTa) and molecular generation models (MolGPT). To assess the performance of the molecular representation model, we fine-tuned the results on the MoleculeNet benchmark datasets. Meanwhile, we evaluated the performance of the molecular generation model based on the generated samples comprising 10,000 molecules. Despite the limited diversity of the pretraining dataset, the models for molecular representation were able to retain at least 90% of their original performance on benchmark datasets, with an additional improvement of 6% in predicting clinical toxicology. In the domain of molecular generation, the model pre-trained on drug-like molecules exhibited a high rate of desirable molecule properties in the unconditionally generated outputs. Additionally, the diversity of generated structures demonstrated notable performance compared to the conditional generation approach. Moreover, the drug-like molecule pre-training strategy is not limited to a specific model or training method, making it a flexible approach that can be easily modified based on the research interests and criteria of interest.INDEX TERMS AI-based drug discovery, pre-training, quantity of drug-likeliness, Pfizer rule.