Top 20 Frontiers of Global Science: 2025
Using big data mining system iFORA, the HSE Institute for Statistical Studies and Economics of Knowledge (ISSEK) has identified the key global scientific topics for 2025, along with associated promising technologies and products.
Reference:
Since 2021, HSE ISSEK has been mapping global science using big data mining system iFORA. based on the OECD classification of Fields of Research and Development ISSEK has compiled a list comprising 15 fields of science, 144 research areas, and approximately 12,000 specific topics (which encompass exploring various phenomena, processes, or problems, or are linked to the development of promising methods, technologies, products, and services.) The calculations are based on analysis of approximately 30 million scientific publications included in the iFORA database over past five years. Frontiers of science were selected from the thematic areas comprising each scientific field.
Frontiers of science are research topics falling within the top decile (the first 10% of a ranked distribution) in terms of significance level for the preceding year, and characterized by a significance index value for that year that exceeds the average value over the last five years. Super Frontiers are interdisciplinary topics that relate to several fields of science or scientific directions. The topic significance index reflects its relevance (frequency of mention) and its influence both on global science as a whole (overall vector centrality) and on other topics within the specific scientific field to which it belongs (intra-cluster vector centrality). Vector centrality—conceptually similar to the notion of centrality used in graph theory—represents an averaged measure of given topic vector proximity to all other topics included in analysis, calculated using cosine similarity of the corresponding vectors. Additionally, based on the analysis of technology terms specific to frontiers, the technologies, products, and research topics associated with their implementation were identified.
The big data mining system iFORA was developed by ISSEK using artificial intelligence technologies. Its source database already comprises over 850 million documents (scientific publications, patents, regulatory frameworks, market analytics, industry media, materials from international organizations, job vacancies, etc.) and is constantly expanding. In 2020, iFORA was highlighted in the Nature journal as an effective decision-support tool for business and government. The OECD classifies iFORA system as a successful initiative for science digitalization. The system has been highlighted both in UN and UN Secretary-General reports as a cutting-edge tool for technology foresight. Adapted generative model iFORA LLM is included in the Registry of Russian Software (No. 33849 of June 18, 2026). HSE ISSEK publishes a reviews series on the frontiers of global science annually.
By the end of 2025, 890 frontiers of global science had been identified. The topics with the highest significance index values include environmental issues, digital transformation, and human mental health (Table 1).
Climate change issues have long remained central to the global agenda and the subject of comprehensive interdisciplinary projects integrating the natural and social sciences. Such studies involve forecasting temperature changes based on complex mathematical models, analyzing behavioral practices, assessing risks, and developing adaptation strategies. Amid the growing threat of environmental crises and challenge scale comprehension, a fundamental shift from anthropocentric to ecocentric paradigm is taking place. The new model envisions a sustainable interaction with nature, in which human interests cease to take precedence over the need to preserve the natural world integrity.
Urban population growth and urbanized areas expansion are altering land-use patterns and increasing strain on natural resources. Local climate effects emerge in cities—such as heat islands, altered precipitation patterns, and increased vulnerability to extreme weather events. Ecosystem services—specifically carbon sequestration, atmospheric pollutants filtration, and hydrological regimes regulation —can mitigate these changes consequences. Geographic information systems data play a pivotal role in decision-making regarding the urban areas planning, enabling better understanding of urban flow dynamics, assessing green spaces conditions and environmental degradation extension. Developing solutions to minimize the human activity negative impact on nature remains a pressing issue. Carbonization—the process of carbon sequestration and storage—is considered a promising method for reducing carbon dioxide concentrations and minimizing carbon footprint across various industrial sectors. For example, technology of organic waste hydrothermal carbonization enables energy-rich products making and can contribute to long-term carbon storage and circular economy development.
The goal of reducing negative impact on environment drives renewable and hydrogen energy development. In solar energy sector, next-generation photovoltaic cells—such as perovskite and tandem structures—and high-efficiency transparent conductive coatings are taking center stage. Hydrogen energy solutions—specifically electrochemical energy conversion systems, including fuel cells and electrolyzers—depend heavily on the development of efficient catalysts. Noble metal nanoparticles are considered the most promising among them, while transition metal oxides (nickel and cobalt) and iron carbides represent a more cost-effective alternative.
Technological frontiers are defined, on the one hand, by computational capabilities growth and, on the other, by a shift toward tasks requiring simultaneous measurement, development modeling, and control of complex systems and processes at previously unattainable level. Machine learning and neural networks make it possible to identify patterns in data; artificial intelligence (AI) and deep learning enable complex systems management; regularization ensures models stability; and spectroscopy and resonators provide extremely precise measurements. In 2025, synergy between the listed areas intensified.
AI is no longer merely a tool for analysis but also a technology for action, often operating without human involvement. AI is no longer merely tool for analysis but also technology for action, often operating without human involvement. This is most evident in autonomous systems, unmanned systems, and digital energy. To enable decision-making and implementation under conditions of uncertainty, data processing is shifting to edge—directly to devices, sensors, and local modules. At the core of such edge AI systems lies machine learning, which enables interpretation of heterogeneous and noisy data without relying on cloud centers. In 2025, researchers at Osaka University unveiled a self-learning technology for embedded devices that processes data up to 100,000 times faster than traditional deep learning methods and delivers up to 60% higher accuracy.
Neural networks are increasingly being integrated into intelligent control systems, enabling robotic manipulators control in dynamic environments with less effort than classical adaptive controllers. One example is the neuromorphic approach based on spiking neural networks (SNN) and LIF and GIF neuron models. Another emerging area is neural-network-based control of autonomous robots groups for monitoring and industrial automation, which requires decentralized coordination as conditions change.
Alongside these approaches, deep learning is widely used, enabling highly complex and unstructured data processing. It allows to continuously assess the situation based on images and other signals, track sequences of events, and respond to changes. Such functions are particularly in demand in robotics and autonomous control systems. Photonics can significantly accelerate the hardware implementation of such models, although this effect is currently observed only in specific tasks. The more complex models, the more important regularization becomes: without it, they easily overfit and lose ability to handle new data. In critical AI application areas, requirements for robustness are particularly high. Medical diagnostic systems must operate reliably for patients whose demographic characteristics differ from those in the training set, while autonomous vehicles must do so in weather conditions absent in training data.
At AI and physics intersection, resonator technologies are being developed which are the foundation of highly sensitive radio-frequency and sensor devices. They provide signal amplification and stabilization, enabling more accurate measurements and more reliable communication between physical world and digital processing systems. Thus, photonic "electronic noses" based on arrays of microresonators generate a unique "optical fingerprint" of analyzed mixture, while a local AI interface recognizes it directly on the chip. In a photonic extreme learning machine (PELM), an array of 18 gratings coupled to microresonators on a silicon chip functions as hidden layer of a neural network, enabling low-latency, highly energy-efficient data processing.
Spectroscopy, identified as a distinct frontier, allows to see previously hidden matters. In 2025, for the first time, the BASE collaboration at CERN demonstrated coherent spectroscopy of quantum transitions of spin single antiproton. The method can significantly improve comparing precision of the proton and the antiproton magnetic moments, which is crucial for studying matter-antimatter asymmetry.In applied fields, surface-enhanced Raman spectroscopy (SERS) using nanoparticles enables signals detection from individual molecules and combined with AI-based spectral analysis, facilitates the development of methods for cytological diagnostics.
In classical disciplines, such a unified toolkit provides the transition from observation to targeted control of processes at micro- and nanoscale. In chemistry measurements and modeling combination supports molecular design: studying hydrogenation processes and adsorption dynamics under non-steady-state conditions helps create materials and catalysts with tailored properties. In physics, methods controlling light, electrical signals, magnetic states, and quantum effects are being improved. Wave and polarization effects are utilized in satellite communication and navigation systems, liquid-crystal displays, silicon photonic devices, and holographic data storage. Such developments facilitate optical chips miniaturization and information processing speed increase and recording density.
Human potential research and maintenance of both human and animal health constitutes a major, distinct block of leading-edge frontiers. Their common feature is focus on early risks detection from molecular and cellular changes to mental disorders mechanisms and factors influencing child development.
In bio- and veterinary medicine, biomarkers enable changes monitoring long before pronounced clinical signs appearance, facilitating shift from symptomatic to early, accurate, and personalized diagnostics. Serum biomarkers analysis helps to detect inflammatory processes, metabolic disorders, tumors, tissue damage, and other pathological states. Another frontier involves antioxidants that inhibit free-radical oxidation capable triggering and exacerbating various pathological processes. In particular, medicines are being developed based on superoxide dismutase—an enzyme that neutralizes free radicals with high specificity and catalytic efficiency.
Mental health is one of the key areas within this block. Amidst chronic stress, information overload, and the accelerating pace of life, increasing attention is being paid to depressive states, anxiety disorders, and other affective disturbances. These aspects have not only medical but also social, economic, and demographic significance, as mental health largely determines individual's quality of life, as well as their opportunities for professional fulfillment and social adaptation. Importance of research aimed at identifying the biological, psychological, and environmental mechanisms underlying mental disorders development, as well as the factors that foster psychological resilience is growing. Early prevention of mental disorders also extends to child-rearing realm; research findings indicate that quality of parent-child bond in early childhood directly influences DNA methylation and brain's neural network architecture formation.
Resume:
Thus, the map of leading scientific frontiers reflects search for systemic responses to global challenges. Environmental studies aims to preserve the planet; biomedicine focuses on improving the quality of human life; digital technologies provide new, highly effective tools for finding innovative solutions; and advancements in biology, physics, and chemistry drive breakthroughs in fundamental knowledge. The frontiers list can serve as a guide for research organisations, universities, and innovative companies when selecting research topics and development strategies, and for government bodies, development institutions, and funding agencies when determining science and technology priorities for supporting.
Sources: calculations based on the iFORA big data mining system (rights holder: ISSEK HSE); this work utilizes results from project HSE-MS-2025-002, conducted at HSE University.
Review prepared by Danil Yatskin and Anna Grebenyuk
