Drug Discovery Today最新文献

Rare diseases represent a major unmet medical need, yet the development of therapeutics for these diseases is limited by small patient cohorts, fragmented data, and uncertain disease mechanisms. Recent advances that combine patient-derived induced pluripotent stem cell (iPSC) models with multi-omics profiling and artificial intelligence offer a robust approach for phenotype definition and drug target identification that requires only minimal patient material. By integrating transcriptomic, proteomic, and metabolomic data, machine learning can produce disease-relevant molecular signatures, identify convergent pathway drivers, and prioritize druggable targets despite limited sample sizes. This short review highlights emerging in vitro models and AI-enabled workflows, discusses their strengths and limitations, and outlines practical strategies to accelerate drug discovery for rare diseases.

Drug-development programs are often called de-risked after accumulating data, publications, financing or regulatory milestones. These signals may establish activity, credibility or readiness, but not necessarily a reduction in consequential uncertainty. Accumulation alone does not make an asset de-risked; de-risking is a claim that evidence has reduced uncertainty relevant to a decision. Such a claim requires evidence that is reliable, relevant to the choice ahead and available before that choice becomes difficult to reverse. Three oncology histories illustrate how evidence can support progression, narrow an indication or arrive too late. This Editorial proposes a shared uncertainty record to make that reasoning explicit across development.

Artificial intelligence (AI) can support drug discovery, but benchmark performance alone does not establish experimental, operational or clinical benefit. This keynote review distinguishes predictive generalization, experimental confirmation, workflow utility and clinical outcomes, separating demonstrated failures from missing validation. We critically examine quantitative case studies of benchmark bias, prospective hit finding and clinical development, alongside data quality, reproducibility, uncertainty and governance. Practical recommendations link each intended use to appropriate comparators, evaluation designs and acceptance criteria. The aim is not to presume that AI fails to translate but to specify what evidence is needed to demonstrate reliable prospective use in drug discovery.

Pulmonary arterial hypertension (PAH) is a progressive pulmonary vascular disease characterized by increased pulmonary vascular resistance, right ventricular overload and premature death. Current therapies improve symptoms and hemodynamics but do not adequately reverse pulmonary vascular remodeling. Heat shock proteins (HSPs), key regulators of proteostasis and stress adaptation, are increasingly implicated in PAH pathogenesis. Among them, Hsp70, Hsp90 and Hsp110 have attracted particular interest because of their distinct roles in chaperone networks and vascular remodeling. This review summarizes the structural and functional features of these HSPs, their roles in PAH-associated remodeling and the therapeutic potential and challenges of HSP-targeted strategies.

We evaluated the commercial performance of 490 FDA-approved new molecular entities (NMEs) and biologic products approved from 2016 through 2025 against analyst forecasts, testing the hypothesis that only a small proportion of approved drugs have the potential to generate substantial returns for their sponsors. We demonstrate that analyst forecasts are subject to considerable limitations, with systematic overestimation of sales particularly prevalent for drugs with biotech sponsors. We further identify a strong association with large, established pharmaceutical companies and their ability to select or develop commercially successful products relative to smaller competitors and emerging companies. Large pharma achieved blockbuster (BB) or super-blockbuster (SBB) status for approximately 34% of their approved portfolio, compared to approximately 9% for medium pharma and approximately 2% for biotech sponsors.

An industry-wide survey of 26 biopharmaceutical companies benchmarked artificial intelligence and machine learning (AI/ML) adoption in translational and ADME sciences. AI/ML is widely used in early discovery and small-molecule workflows but rarely the primary driver of decision making. Data quality and organizational readiness were identified as key barriers to adoption. Respondents highlighted molecular design and hybrid mechanistic-AI approaches as key opportunities. Beyond benchmarking adoption, this work provides cross-industry perspectives on the scientific, organizational and data foundations needed to scale impact through integrated workflows, shared benchmarks and collaborative translational decision-making.

Real-world evidence (RWE) is increasingly included in marketing authorization applications, and plays a growing role in regulatory decision-making. Although commonly included in product labels as safety information, its use for efficacy claims remains limited and randomized controlled trials (RCTs) remain the gold-standard. Regulatory guidance on the use of RWE is expanding, but a significant gap remains regarding its role in labeling, particularly for communicating drug effectiveness. This review examines how real-world evidence (RWE) has been used in many therapeutic areas to support efficacy claims in drug product labels approved by the US Food & Drug Administration (FDA) and the European Medicines Agency (EMA). A systematic search identified 173 brands that included RWE as primary efficacy evidence, as an external comparator to single-arm trials, or as supportive evidence alongside RCTs. This work offers insights for designing RWE studies that support labeling claims and inform regulatory guidance.

The increasing healthcare burden demands precise diagnostics for early detection and longitudinal disease monitoring. Although conventional assays such as ELISA and PCR offer precision, they require intensive sample preparation, bulkier instrumentation and prolonged assay times. This review examines plasmonic biosensors as a transformative alternative for personalized and point-of-care diagnostics. We comprehensively examine recent advancements, such as the integration of nanomaterials, microfabrication strategies, advanced surface chemistry and prototype miniaturization, and consider examples for the detection of cancer biomarkers (microRNA-21 and alpha-fetoprotein), viral proteins (respiratory syncytial virus and SARS-CoV-2), and inflammatory biomarkers (C-reactive protein). Furthermore, we address the critical aspects that hinder the clinical translation of plasmonic biosensors. The review concludes that, despite significant laboratory-scale progress, standardized manufacturing, scalable nanofabrication and the inclusion of artificial intelligence/machine learning frameworks remain essential for the widespread utilization of plasmonic technology in personalized medicine and diagnostics.

Forkhead box (FOX) proteins are transcription factors whose dysregulation drives tumor initiation, metastasis, and therapy resistance, yet their direction of effect is determined by cellular context rather than by their conserved DNA-binding domain. Applying explicit criteria across all 44 human members, we find context-dependence to be the modal behavior, reversing in some cases within a single indication. This creates an inversion for drug discovery: The one structurally tractable region is common to the whole family, whereas the regions conferring selectivity are the least characterized. We assess small-molecule inhibition, covalent ligand discovery, targeted degradation and indirect pathway modulation against the evidence for direct target engagement and identify nonconserved regulatory surfaces as the more promising route to selective FOX-directed agents.