Motivation: Virtual reality technology holds significant potential for applications in biomedicine, particularly in the visualization and manipulation of protein molecular structures. To facilitate the study of protein molecules and enable the state-of-the-art VR hardware, we developed a novel VR software named VisionMol, which allows users to engage in immersive exploration and analysis of three-dimensional molecular structures using a range of virtual reality platforms (such as Rhino X Pro, Meta’s Oculus Quest Pro/3) as well as personal computers. Results: Built on the Unity engine and programmed using C#, VisionMol incorporates custom scripts to enable a variety of molecular operations. Users can rotate, scale, and translate molecular models using gestures, controllers, or other input devices. Furthermore, VisionMol offers rich visualization and interactive features, including multi-model molecular display, distance measurement between molecular components, and molecular alignment and docking. Summary: These capabilities facilitate a more intuitive understanding of molecular interactions and chemical properties. The real-time interactive effects and clear visual representations allow users to delve deeper into the relationships between molecular structures and their properties, thereby accelerating research progress and promoting scientific discovery. We believe that this VR-based protein molecule analysis has significant application value in several fields, including biomedicine, life science education, drug design and optimization, biotechnology, and engineering applications. Availability: The code is at https://github.com/WangLabforComputationalBiology/VisionMol. The v1.1 code (for Oculus Quest) could also be found at https://doi.org/10.5281/zenodo.14705790. The v1.0 code (for Rhino X Pro) could also be found at https://doi.org/10.5281/zenodo.14865216. Detailed documentation could be found at https://visionmol.surge.sh/#/en-us/README. Supplementary information: Supplementary are available at Bioinformatics online.

In the era of artificial intelligence (AI) and Industry 4.0, pilot-scale platforms for small molecule chemical drugs are undergoing a transformative digital evolution. These platforms serve as a critical link between early-stage laboratory research and full-scale pharmaceutical manufacturing, ensuring process feasibility, scalability, and regulatory compliance. This review offers a comprehensive and forward-looking analysis of the structure, function, and strategic importance of pilot-scale systems within the modern pharmaceutical landscape. Focusing on the integration of AI and intelligent automation, the study highlights innovations such as AI-driven process optimization, predictive maintenance, data integration, digital twin technologies, and continuous manufacturing. These technologies are reshaping conventional production paradigms by enhancing efficiency, improving quality control, and reducing environmental impact. The convergence of computational pharmaceutics and green chemistry is also examined as a key driver of sustainable and intelligent drug development. Moreover, the review addresses the industry’s transition toward Industry 5.0, characterized by human-machine collaboration, data-centric innovation, and an emphasis on sustainability. Persistent challenges such as equipment standardization gaps, data-sharing limitations, and outdated infrastructure are critically discussed. Drawing from industrial case studies, academic research, and best practices, this paper explores both the opportunities and constraints associated with AI-enabled pilot platforms. Ultimately, the review aims to inform future strategies in digital pharmaceutical manufacturing by underscoring the importance of technological innovation, regulatory alignment, and collaborative ecosystems in advancing the development, efficiency, and sustainability of small molecule drug production.

Learning low-dimensional continuous vector representation for short k-mers divided from long DNA sequences is key to DNA sequence modeling that can be utilized in many bioinformatics investigations, such as DNA sequence classification and retrieval. DNA2Vec is the most widely used method for DNA sequence embedding. However, it poorly scales to large data sets due to its extremely long training time in kmer embedding. In this paper, we propose a novel efficient graph-based kmer embedding method, named Kmer-Node2Vec, to tackle this concern. Our method converts the large DNA corpus into one kmer co-occurrence graph and extracts kmer relation on the graph by random walks to learn fast and high-quality kmer embedding. Extensive experiments show that our method is faster than DNA2Vec by 29 times for training on a 4GB data set, and on par with DNA2Vec in terms of task-specific accuracy of sequence retrieval and classification.

Learning low-dimensional continuous vector representation for short pitch segment extracted from songs is has been confirmed to contain tonal features of music, which is key to melody modeling that can be utilized in many music investigations, such as genre classification, emotion classification, and music retrieval, and so on. The skip-gram version of Word2Vec is ubiquitous, and widely used approach for music pitch segment embedding, but it poorly scales to large data sets due to its extremely long training time. In this paper, we propose a novel efficient graph-based embedding method, named Music-Graph2Vec, to tackle this concern. This approach converts music files into graphs, extracts the rhythmic sequence through random walking, and trains the rhythmic embedding model using skip-gram. Experimental results demonstrate that Music-Graph2Vec outperforms Word2Vec in training rhythmic embedding, with the advantage of being 55 times faster on the top-MAGD dataset (2,134.7s for Word2Vec and 38.9s for Music-Graph2Vec), with the same accuracy for Word2Vec in terms of music genre classification.

Background: Wastewater surveillance has emerged as a cost-effective and powerful approach for monitoring infectious diseases at the community level, as demonstrated during the COVID-19 pandemic. However, translating wastewater viral signals into reliable epidemic forecasts remains challenging because measurements are noisy, environmental conditions vary over time, and disease transmission is inherently non-linear. Mechanistic models such as the Susceptible-Exposed-Infectious-Recovered-Virus (SEIR-V) framework can bridge wastewater signals with infection dynamics, but their performance depends heavily on accurate parameter estimation, and conventional least-squares optimization is prone to convergence to local optima. Methods: We developed a Genetic Algorithm-optimized SEIR-V (GA-SEIR-V) model that applies evolutionary optimization for robust multi-parameter estimation. To improve biological realism, the model incorporates a temperature-dependent viral decay rate represented by a sinusoidal function fitted to regional climate data, enabling dynamic simulation of seasonal viral persistence in wastewater. The framework was calibrated and evaluated using wastewater viral load, clinical case, and temperature data from California (September 2020-November 2022), Greater Boston (September 2020-May 2021), and Switzerland (January 2022-December 2022). Results: Using a unified evaluation protocol with 30 independent genetic algorithm runs, we explicitly separated in-sample calibration from out-of-sample prediction. During calibration, GA-SEIR-V reproduced the reported model fit for California and consistently outperformed a carefully re-tuned least-squares baseline on large, heterogeneous datasets, demonstrating greater robustness and more reliable parameter estimation where gradient-based optimization became trapped in local optima. In contrast, long-term prediction beyond the calibration period remained difficult for all methods, particularly across multiple epidemic waves, resulting in limited out-of-sample forecasting accuracy. Conclusion: GA-SEIR-V combines evolutionary optimization with temperature-aware viral decay modeling to improve the robustness and identifiability of SEIR-V parameter estimation, particularly for complex wastewater datasets. By explicitly distinguishing calibration from prediction, the study shows that excellent in-sample fitting does not necessarily translate into reliable long-term forecasting. The proposed framework is therefore most suitable for robust parameter estimation and short-term epidemic tracking, while providing a foundation for future improvements toward more accurate long-range prediction.