Expert Opinion on Drug Discovery最新文献

Introduction: In silico technologies are increasingly shaping vaccine development, supporting the field beyond empirical discovery toward rational, data-driven design. Contemporary computational pipelines enable rapid antigen screening, high-precision epitope-MHC binding prediction, structural modeling, and immune response simulations. These approaches are accelerating vaccine discovery not only for infectious diseases but also in oncology, where neoantigen prediction underpins personalized cancer immunotherapy.

Areas covered: This review explores recent advances in computational pipelines for epitope-based vaccine design, covering antigen discovery; B- and T-cell epitope mapping; safety and specificity assessment; vaccine construct assembly with adjuvants and linkers; structural modeling; and immune-response simulations that predict efficacy in specific disease contexts using advanced platforms. It showcases applications in infectious diseases, including SARS-CoV-2, tuberculosis, and influenza, and poxivirus infections, as well as in cancer immunotherapy. It is based on literature obtained through searches utilizing PubMed, Scopus, and Web of Science databases covering publications up to 2025, using combinations of keywords such as epitope-based vaccines, reverse vaccinology, immunoinformatics, and immune system simulation.

Expert opinion: In silico approaches offer a transformative advantage to vaccine research by delivering speed, cost-efficiency, and enhanced precision. Yet the predictive power of current computational pipelines is still constrained by algorithmic limitations and by their incomplete integration of immune-regulatory processes. Progress in artificial intelligence, multi-omics integration, and formal recognition of digital evidence by regulatory agencies will be crucial for narrowing the gap between computational predictions and experimental validation. Ultimately, combining predictive immunoinformatics with advanced immune simulations and rigorous verification could help establish in silico methodologies as a cornerstone of next-generation vaccine development.

Introduction: N-methyl-D-aspartate (NMDA) receptor antagonists such as phencyclidine (PCP) and ketamine can induce schizophrenic features in healthy volunteers and exacerbate the symptoms in schizophrenic patients. Furthermore, the administration of NMDA receptor antagonists to rodents produces hyperlocomotion. The ability of drugs to attenuate this hyperlocomotion correlates with clinical efficacy on positive symptoms. Similarly, social withdrawal is taken as a surrogate of unsociability in schizophrenia. Furthermore, first episode psychosis and chronic schizophrenia can be modeled by acute and subchronic administration of NMDA receptor blockers, respectively. Therefore, the NMDA hypofunction model provides a powerful tool to develop new therapeutic strategies in drug discovery to treat schizophrenia.

Areas covered: This perspective describes the similitudes between schizophrenia in humans and the traits demonstrated by rodent models based upon the hypofunction of NMDA receptors. Comparisons are made in terms of behavioral, neurochemical, neuroimaging and neurophysiological studies. Different therapeutic responses are also discussed.

Expert opinion: Both schizophrenic patients and developed rodent models exhibit many similitudes such as decreased expression of NMDA receptors, enhanced dopaminergic and serotonergic transmission as well as altered gamma oscillations and deficits in cognitive paradigms. The NMDA receptor antagonism model can thus represent an excellent strategy to study the neurobiological underpinnings of schizophrenia and the potential therapeutic role of new antipsychotic drugs.

Introduction: The search for molecular novelty frequently collides with the fact that drug candidates with the best translational prospects are confined to - or concentrated in - defined regions of chemical space. The new possibilities of AI, particularly retrosynthesis prediction and generative AI, allow for the automated or semi-automated exploration of less restricted and unexplored areas of chemical space.

Areas covered: The notion of novelty in drug discovery is discussed, and representative examples of AI-guided de novo drug design, optimization, and retrosynthesis prediction are presented, with a focus on reports on open-source tools published in the last 3 years (2022-2025). Scopus was used to search relevant literature.

Expert opinion: Modern deep learning architectures have been adapted for the de novo design and molecular optimization. These technologies, and especially those based on conditional generation, will possibly have a great impact on expanding the regions of chemical space that are exploited therapeutically. However, there are some persistent challenges in the field that are gradually being addressed, including how to assess the synthetic accessibility of designed molecules without compromising the generation of structural novelty; the need to increase the availability and diversity of benchmark datasets; and the relative scarcity of large-scale experimental validation of the designs.

Introduction: Cheminformatics has become a cornerstone of modern drug discovery, offering the ability to efficiently manage and analyze large volumes of chemical and biological data. Publicly available databases such as PubChem, ZINC, ChEMBL, DrugBank, ChemDiv, natural product databases, among others, are essential for accessing diverse chemical structures, biological activities, and pharmacological properties.

Areas covered: This review provides an overview of recent (2024-2025) trends in mining data from PubChem and other representative public databases for virtual screening. It also discusses the integration of experimental validation and computational tools in drug design and cheminformatics workflows. The article is based on literature retrieved from SciFinder.

Expert opinion: Public chemical databases contain thousands to billions of compounds and various computational strategies have necessitated development to navigate this vast chemical space effectively. These include application programming interfaces, similarity searches, physicochemical filtering, and target-based selection. Such filtering strategies have enabled the extraction of focused compound subsets for evaluation through various cheminformatics tools, ultimately supporting informed decision-making in lead discovery and optimization.

Introduction: Cancer cachexia (CC) is a multifactorial syndrome characterized by progressive weight loss, anorexia, and loss of skeletal muscle and fat mass, resulting in reduced quality of life and poor prognosis. Currently, there are no approved pharmacological treatments for CC, highlighting the urgent need for developing novel experimental models.

Area covered: This review covers recent advancements in preclinical models of CC, highlighting their implications for drug discovery and therapeutic development. The literature search was conducted in PubMed up to April 2025.

Expert opinion: CC remains clinically challenging and requires improved translational research and therapeutic strategies. Improved preclinical models, such as personalized patient-derived xenograft models incorporating patient-specific immune profiles and microbiota, hold promise for precision medicine. Identification of standardized extracellular vesicle (EV) derived biomarkers and effective targeting of EV signaling pathways are critical research directions. In addition, clinical validation of appetite regulators such as glucagon-like peptide-1 and growth differentiation factor-15, along with comprehensive approaches integrating diet, exercise, and targeted pharmacological interventions, will be pivotal. Finally, multidisciplinary collaboration is essential to translate these findings into meaningful therapies that will ultimately improve patient prognosis and quality of life.

Introduction: Follicular lymphoma (FL) is an indolent yet incurable subtype of non-Hodgkin lymphoma characterized by repeated relapses and diminishing responses with each treatment line. Although front-line chemoimmunotherapy achieves high initial response rates, a subset of patients - particularly those with early relapse (POD24) - experience poor outcomes and require alternative therapies. Tisagenlecleucel (tisa-cel), a CD19-directed chimeric antigen receptor (CAR) T-cell therapy, has emerged as a promising option for relapsed or refractory (r/r) FL, offering the potential for deep and durable remissions.

Areas covered: This review covers the scientific rationale, preclinical innovations, and clinical development of tisa-cel, from its origins in 2^nd-generation CAR-T engineering to its pivotal trials in hematologic malignancies. It is based on a literature search using PubMed, Embase, and conference abstracts from major hematology meetings from 1987 to April 2025. The paper deta ils the ELARA trial outcomes, subsequent long-term and real-world data, and the competitive landscape of third-line therapies for r/r FL.

Expert opinion: Tisa-cel has demonstrated high response rates and sustained remissions with a favorable safety profile in heavily pretreated FL, including high-risk populations such as those with POD24. While bispecific antibodies offer convenient outpatient administration, CAR-T cell therapy provides the potential for deep and durable remissions. The 4-1BB costimulatory domain used in tisa-cel and liso-cel is associated with a lower incidence of severe CRS and ICANS compared to CD28-based constructs. the field evolves, careful patient selection and head-to-head trials will be essential to refine therapeutic sequencing in r/r FL.

Introduction: In silico trials represent an unprecedented opportunity for ocular drug development. These trials not only promise significant reductions in costs and development timelines but also meaningful improvements in both patient safety and compliance.

Areas covered: This critical perspective gives discussion to the value of in silico trials for novel ocular drug discovery and development. Discussion includes the potential that these trials hold and the challenges that need to be addressed.

Expert opinion: The ophthalmic community stands at a critical juncture, where transitioning from traditional drug development paradigms to more integrative approaches, including computational methods, may profoundly reshape clinical practice. Nevertheless, there a several important limitations that must be overcome; these limitations include dependency on the quality and completeness of input data, accounting for complex biological systems, particularly in ophthalmology, and the variability in patient responses due to genetic, environmental, or lifestyle factors. The issue of silico model validation is also important, especially where the extensive real-world clinical data is not available for comparison. Another important concern is the limited regulatory acceptance of in silico trials to date while standardized guidelines and validation frameworks are still in development. All these issues will need to be addressed for future meaningful progression in the field.

Background: Promiscuity of drugs and targets plays an important role in drug-target prediction, ranging from the explanation of side effects to their exploitation in drug repositioning. A specific form of promiscuity concerns drugs, which interfere with protein-protein interactions. With the rising importance of such drugs in drug discovery and with the large-scale availability of structural data, the question arises on the structural basis of this form of promiscuity and the commonalities of the underlying protein-ligand (PLI) and protein-protein interactions (PPI).

Research design and methods: The authors applied the protein-ligand interaction profiler, PLIP, to experimental and predicted structures and characterize drugs in clinical trials, which target PPI.

Results: PPIs generally involve more non-covalent interactions than PLI with overlapping interaction patterns and key binding site residues. In contrast to experimental structures, predicted structures fall short in accurately capturing interaction details at the interface.

Conclusion: Taken together, our analysis shows that PPIs and PLIs have sufficient commonalities to merit future work into computational screenings for drugs targeting PPIs. It will be key to further improve structure prediction, specifically for binding site details.

Introduction: Understanding the interactions between functional groups, ligands, molecular fragments, and whole molecules is critical in modern drug discovery. Key to this endeavor is the theoretical development of the fundamental inter-particle forces at play and their implementation in numerous software packages that allow the calculation of interaction energies at varying levels of theory ranging from the entirely classical at one extreme to the fully quantum mechanical at the other.

Areas covered: In this review, the authors consider the concept of an intermolecular potential energy function and its separation into short- and long-range regions. This is followed by a summary of the perturbation theory calculation of the electrostatic, induction, and dispersion energy shifts by expanding the charge distribution in terms of source multipole moments. Next, the authors outline the construction of a typical molecular force field and its parameterization; they also discuss the fundamental background of molecular dynamics (MD) simulations, their implementation in several well-known software packages and their deployment in modern computational drug discovery, including recent work with Artificial Intelligence and Machine Learning techniques. Papers cited by SSC were the result of a literature search conducted using PubMed and Google Scholar during Jan-July 2025 as well as from the authors' personal literature stock.

Expert opinion: While the underlying quantum mechanical theory of intermolecular forces is well-known, their accurate and reliable calculation for an ever-growing variety of increasingly complex systems mirrors the advances in computational hardware on which such simulations are performed. Coupled with emerging machine learning techniques, this allows for the rapid and efficient computer-aided discovery of potential new drug candidates, in the process revolutionizing research and development in both academia and industry.

Background: Cathepsin K (CTSK), a cysteine protease of the papain family, exhibits high expression in activated osteoclasts, making it a key therapeutic target for osteoporosis. However, there are currently no CTSK inhibitors available for clinical use.

Research design and methods: The authors employed a combination of deep learning approaches and experimental methods to identify novel CTSK inhibitors. Firstly, the authors utilized Chemprop to develop a predictive model for predicting CTSK inhibition. Subsequently, the top 100 predicted molecules were selected for experimental validation, with the most potent inhibitors chosen for further analysis, including enzyme kinetics, molecular docking, molecular dynamics simulations, and RANKL-induced osteoclastogenesis assays.

Results: The authors identified six compounds exhibiting concentration-dependent CTSK inhibitory effects, with Quercetin, γ-Linolenic acid (GLA), and Benzyl isothiocyanate (BITC) demonstrating the highest potency. Enzyme kinetics studies revealed that these inhibitors employ distinct mechanisms of CTSK inhibition. Molecular dynamics simulations further showed that Quercetin and BITC form stable interactions at the CTSK active site. Moreover, in-vitro studies demonstrated that Quercetin and GLA significantly inhibit RANKL-induced osteoclastogenesis in RAW264.7 cells.

Conclusions: This study led to the development of a deep learning model capable of predicting CTSK inhibitors and identified Quercetin, GLA, and BITC as promising candidates for the treatment of osteoporosis.