The Emergence of Catalysis-ai: A New Frontier in Catalysis Research?

The Emergence of Catalysis-ai: A New Frontier in Catalysis Research? | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article The Emergence of Catalysis-ai: A New Frontier in Catalysis Research? Samuel Villanueva, Lisbeth Mendoza, Paulino Betancourt, Susana Pinto-Castilla This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9182089/v1 This work is licensed under a CC BY 4.0 License Status: Under Revision Version 1 posted 14 You are reading this latest preprint version Abstract Recent advancements in data infrastructure, computational statistics, and artificial intelligence (AI) have inaugurated a transformative era for chemical sciences. These computational paradigms facilitate the optimization of complex systems at an unprecedented rate, transcending the limitations of traditional trial-and-error methodologies. This innovative convergence of domain-specific scientific knowledge and advanced heuristics is pivotal for the engineering of next-generation, sustainable chemical processes characterized by minimized energy footprints and enhanced selectivity. Significantly, this evolution fosters a deep integration between heterogeneous and homogeneous catalysis and the diverse analytical tools emerging from the digital frontier, potentially catalyzing a new “Catalysis–AI” paradigm. However, comprehensive bibliometric analyses reveal a significant break; despite the proliferation of AI literature, its substantive implementation in experimental catalysis remains nascent. The current landscape is hindered by data silos and a lack of standardized descriptors. Consequently, the formulation of robust explanatory and predictive hypotheses—anchored in both physical chemistry and data-driven insights—is imperative. Establishing such a framework is essential for achieving a fundamental understanding of active sites and reaction mechanisms, ultimately addressing the urgent scientific and technological imperatives of modern industry. Catalysis-AI bibliometric analysis artificial intelligence evolution of scientific catalytic publications Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 Highlights Recent advancements in artificial intelligence have significantly impacted science and technology. Despite the concerns and potential risks associated with artificial intelligence, the responsible integration of these tools is essential for the advancement of science and technology, especially in the field of catalysis. It is necessary to integrate catalysis, with the diverse tools of artificial intelligence, ultimately giving rise to a new paradigm of catalysis–AI. The disconnect between catalysis and artificial intelligence is particularly pronounced in Ibero-American countries, which have served as fundamental pillars in the advancement of catalytic knowledge for decades. Introduction Throughout history, there have been numerous periods characterized by technological advancements that have catalyzed significant scientific discoveries and innovations, fundamentally transforming society. The discourse surrounding artificial intelligence (AI) often emphasizes its potential risks, including algorithmic bias, systemic discrimination, widespread job displacement, and even existential threats to humanity [ 1 , 2 ]. While some commentators express concern over these dystopian scenarios, others advocate for the substantial benefits that AI may offer [ 3 ]. Proponents argue that AI has the capacity to aid humanity in addressing some of its most formidable and intricate challenges by dramatically accelerating the pace of scientific discovery, particularly in fields such as medicine, climate science, sustainable technology [ 4 , 5 ], transforming even the traditional teaching methodologies [ 6 – 8 ]. AI-driven tools are increasingly being integrated across virtually all domains of science [ 9 ]. For instance, they can efficiently identify promising candidates for the synthesis of novel materials and analyze vast datasets to uncover patterns and model complex systems. Within this context, two domains exhibit remarkable potential to revolutionize scientific practice: “literature-based discovery” (LBD) and “autonomous laboratories” or robotic scientists. LBD employs AI language analysis to scrutinize existing scientific literature, enabling the identification of new hypotheses, connections, and insights, thereby fostering interdisciplinary collaboration and innovation. Conversely, self-directed laboratories utilize AI to autonomously generate and test hypotheses through the execution of hundreds or thousands of experimental trials. Moreover, the integration of AI into scientific research has the potential to facilitate widespread access to cutting-edge knowledge and techniques. By automating complex analyses and lowering the barriers to entry in high-tech research environments, AI tools can empower researchers from diverse backgrounds and institutions, including those in developing regions, to contribute to global scientific advancements. These tools can promote a more inclusive scientific community and enhance the diversity of perspectives that drive innovation. Following this line of inquiry, we explored the connection between catalysis and its integration with diverse AI tools over recent decades. Our approach began with an examination of the various definitions associated with this emerging interdisciplinary field. In addition, we conducted a bibliometric analysis using the scientific publication platform The Lens®, applying filters by document type, including journal articles, review papers, letters, and conference proceedings. All bibliometric analysis encompasses publications spanning the period 1991–2024. Methodology A comprehensive search was performed on the scientific publishing platform The Lens® ( https://www.lens.org/ ) using the following query: (catalys OR catalyst OR catalyzed OR catalytic OR biocatalysis OR biocatalytic OR “enzyme technology” OR “enzyme engineering” OR photocatalytic OR photocatalyst OR “photo-catalytic” OR electrocatalysis OR “electro-catalytic activity” OR nanocatalysts OR nanocatalysis) AND (“artificial intelligence” OR “machine intelligence” OR “computational intelligence”). The search was applied to titles, abstracts, keywords, and fields of study. Results were filtered by document type, including journal articles, review articles, letters, and conference proceedings. A total of 725 records were retrieved, between 1991–2024, of which 107 were excluded after manual screening for irrelevance to the study objectives. The remaining records were exported in CSV format, containing metadata such as title, abstract, publication date and year, journal, authors, field of study, DOI, and related information. The search was conducted globally, and bibliometric indicators of scientific output over time were generated, categorized by evolution of scientific publications , countries, most frequent terms in abstracts, and trending topics. Bibliometric study Data processing was carried out using the Biblioshiny interface of the Bibliometrix package, where trivial terms were removed and synonyms consolidated. It is important to note that Bibliometrix R package, (version 4.3.0) is a tool designed for scientometric and quantitative bibliometric analyses. Its implementation in the R programming language leverages the advantages of an open-source environment and a robust ecosystem. VOSviewer (version 1.6.19), in contrast, is specialized software for the visualization of bibliographic networks, enabling the exploration of relationships and patterns among authors, institutions, and keywords through co-occurrence and co-citation analyses. In this study, VOSviewer was employed specifically for visualization of bibliometric maps in both network and density formats. All search, analysis, validation, and data extraction procedures were conducted between July 29 and August 10, 2025. Evolution of the Thematic Map To elucidate the conceptual structure of the study, a thematic map was generated by analyzing bigrams extracted from the abstracts of each publication using the Walktrap algorithm. The map was organized into four quadrants according to Callon’s framework of density and centrality. In this representation, nodes are characterized by two measures: Callon centrality (CC) and Callon density (CD) [ 10 ]. Callon centrality quantifies the importance of a theme or node within the overall network, whereas Callon density reflects the degree of development of the theme. Based on these two measures, research topics can be positioned within a two-dimensional diagram comprising four quadrants: (1) upper right, driving themes; (2) lower right, basic themes; (3) lower left, emerging or declining themes; and (4) upper left, niche or highly specialized themes. Discussion The question of whether machines can think was first posed by Alan Turing in his seminal 1950 paper, “Computing Machinery and Intelligence” [ 11 ]. In this work, Turing contends that addressing this question necessitates a clear definition of the terms: “machine” and “thinking”. However, precisely defining “thinking” presents a challenge, as it is inherently subjective. To circumvent this issue, Turing proposed an indirect approach to assess a machine’s ability to exhibit intelligence indistinguishable from that of a human. This assessment, now known as the Turing Test, involves a human engaging in conversation with both a computer and another human, without knowing which participant is which. If the human is unable to reliably discern the machine from the human interlocutor, the machine is deemed intelligent. The term “Artificial Intelligence” (AI) was formally defined by John McCarthy during the Dartmouth Conference in 1956 as “the science and engineering of making intelligent machines”. This landmark event is widely regarded as the inception of the field of Artificial Intelligence. Presently, the domains of artificial intelligence can be broadly categorized into 16 areas, which include [ 12 ]: reasoning, programming, artificial life, belief revision, data mining, distributed AI, expert systems, genetic algorithms, knowledge representation, machine learning, natural language understanding, neural networks, theorem proving, constraint satisfaction, and the theory of computation. These categories collectively encompass the diverse methodologies and applications that define contemporary AI research and development. AI experienced a significant surge in popularity during the 1980s, marked by the development of “expert systems” by various research institutions and universities. These systems were designed to encapsulate a set of foundational rules derived from expert knowledge, thereby assisting non-experts in making informed decisions and addressing real-world problems for the first time. Notable examples of such systems include XCON (eXpert CONfigurer), developed by Carnegie Mellon University, and MYCIN, created at Stanford University. The XCON program was implemented at Digital Equipment Corporation (DEC), a pioneer in the minicomputer manufacturing sector. By 1986, XCON had processed approximately 80,000 orders, achieving an impressive accuracy rate of nearly 98%. It was estimated that the system saved DEC around $25 million annually. MYCIN, on the other hand, was designed to diagnose infectious diseases and recommend antibiotic dosages based on patients’ weight. Research conducted by Stanford Medical School indicated that MYCIN’s diagnostic success rate was approximately 65%, while human specialists achieved a success rate of about 80%. These early expert systems laid the groundwork for subsequent advancements in AI and its applications in various fields. However, over time, expert systems have disclosed several inherent limitations, including a lack of flexibility, restricted versatility, high maintenance costs, and various ethical and legal dilemmas. For instance, in the event that MYCIN erroneously diagnosed a patient, the question arises: who would bear responsibility—the programmer or the physician? During the same decade, Geoffrey Hinton and his colleagues made significant advancements in the development of the backpropagation algorithm, which facilitates machine learning and underpins nearly all contemporary neural networks. Nevertheless, it was not until 2006 that neural networks based on backpropagation began to exert a substantial influence on the field. Notably, one of the so-called “godfathers” of artificial intelligence resigned from his position about Google, expressing a reconsideration of his stance by stating: “I’ve suddenly changed my mind about whether these systems will surpass human intelligence”. In the 21st century, artificial intelligence has emerged as a critical domain of research, encompassing a diverse array of fields, including engineering, science, education, medicine, marketing, accounting, finance, economics, and law, among others [ 13 ]. The scope of AI has expanded tremendously, as machine intelligence equipped with machine learning capabilities has produced significant impacts across businesses, government entities, and scientific inquiry. In this context, artificial intelligence holds promise in addressing critical issues related to sustainable production, optimizing energy resources, managing supply chains, and reducing waste. For example, in the realm of smart manufacturing, there is a growing trend to integrate AI into green manufacturing processes to adhere to the most stringent environmental standards. Discovering New Research Methodologies: Mesoscience Despite the numerous scientific discoveries and advancements achieved to date, researchers continue to grapple with the enigmatic gaps that persist within the chains of knowledge. Regardless of the extent of information available, our understanding of the complex interrelationships that often exist between a system “as a whole” and its constituent elements remains limited. Such connections can manifest in various phenomena, including violent tornadoes, intricate neural networks, and fluctuating reactive patterns, among others. To the astonishment of scientists, relationships observed in seemingly unrelated domains reveal profound and intriguing similarities. Describing, predicting, and manipulating this complexity poses a significant challenge across nearly all fields of research and represents a bottleneck in numerous industrial processes. Is there a viable approach to decode these phenomena? A group of Chinese researchers is pursuing this inquiry and has introduced the concept of “mesoscience”. This methodology aims to elucidate the mechanisms underlying the multiscale spatiotemporal structures of complex systems and to enhance the analysis of these systems through the use of computational models and paradigms [ 14 , 15 ]. Specifically, mesoscience seeks to bridge the gap between the macroscale (system) and the microscale (unit). To elucidate the common principles and rules within mesoscience, the Energy Minimization Multiscale Model (EMMS), has been developed [ 16 ]. This model is formulated as a multi-objective variational problem, accounting for trade-offs that arise when competing factors are considered. The convergence between artificial intelligence and the mesoscience paradigm offers a promising route to overcome the intrinsic computational limitations of the EMMS model. Historically, solving the multi-objective variational problems associated with this model has demanded massive processing resources, restricting its applicability in complex operational scenarios. In this context, the implementation of Physics-Informed Neural Networks (PINNs) emerges as a disruptive tool. These algorithms do not merely act as universal approximators; they integrate fundamental physical laws directly into the network's loss function, ensuring the model's consistency with conservation principles. By leveraging this capability, it is possible to drastically accelerate calculation convergence. Consequently, AI enables real-time simulations that, until recently, were computationally prohibitive for the design of multiscale catalytic systems. The following examples illustrate how these multiscale challenges are being addressed through the integration of mesoscience and advanced computational tools. Guided by this perspective, the complexity of catalytic systems has garnered increasing attention in recent years and is being investigated from a mesoscience perspective. For instance, the generation of greener materials for the production of clean, efficient, and cost-effective energy necessitates the design of more effective and environmentally friendly energy conversion and storage devices. In this context, electrocatalytic materials with varying morphologies, sizes, and compositions, exhibit distinct catalytic responses in these applications, depending on their specific configurations. Consequently, the relationship between the catalytic behavior of electrocatalytic materials and these factors has begun to be explored at multiple levels through the lens of mesoscience. For example, Peng and Wei [ 17 ] examined the integrated morphology of the electrode, which encompasses the structure of the catalyst’s pores, the reaction interface, and the active site, to demonstrate how mesoscience can aid in the design of more efficient electrocatalytic materials by modulating their geometric and electronic structures. Similarly, Wu and colleagues [ 18 ] investigated the design of supported nanometallic catalysts using mesoscience to understand the equilibrium between high catalytic activity (designated as mechanism A) and high catalyst stability (mechanism B) in the synthesis of hydrogen peroxide (H 2 O 2 ). Among their findings, the researchers identified that the heat of reaction and the enthalpy of fusion of the supported nanometallic catalyst were significant factors influencing mechanism B, thereby providing critical insights for optimizing catalyst design. In the other study about enzymatic systems, integrated micro-macro methodologies had yielded valuable contributions by examining the various factors that modify catalytic responses [ 19 ]. This approach aims to develop a new generation of supramolecular compounds through the assembly of enzymes, resulting in systems with enhanced catalytic selectivity and efficiency for a wide range of applications. Simplifying Calculations with Machine Learning Machine learning, a subfield of artificial intelligence, enables machines to learn autonomously without explicit programming. This technology is predicated on the identification of patterns within data to facilitate predictions. In contrast to traditional computing, which relies on coded instructions crafted by experts, machine learning platforms independently learn through increasing exposure to examples, utilizing algorithms to refine their understanding. In essence, this approach provides the opportunity to train computers on the known catalytic properties of various materials, thereby predicting the most promising potential catalysts for a reaction of interest within an automated machine learning framework for catalyst design [ 20 – 22 ]. A notable study by Ulissi et al. [ 23 ] made significant strides toward establishing such an automated machine learning framework for catalyst design. The systematic methodology employed in this research led to the identification of an active site that had previously been overlooked in nickel and gallium systems during the electrochemical reduction of CO 2 : Ni atoms encircled by Ga atoms at the surface. These newly identified active sites demonstrated enhanced thermodynamic parameters and exhibited step-like kinetic behavior. This discovery substantially advances our understanding of the factors that contribute to the improved design of bimetallic catalysts. Toward Industry 4.0 Smart production systems necessitate innovative solutions to enhance the quality and sustainability of manufacturing activities while concurrently reducing costs. In this context, artificial intelligence AI-driven technologies, in conjunction with key Industry 4.0 (I4.0) advancements, are instrumental in optimizing and monitoring manufacturing processes. These technologies enable the integration of various processes and systems within the plant, heralding a new industrial revolution characterized by the convergence of physical and digital dimensions, which is expected to generate novel industrial paradigms. The term “Industry 4.0” was first introduced at the Hannover Fair in 2011, during the opening ceremony delivered by Professor Wolfgang Wahlster, Director and CEO of the German Research Center for Artificial Intelligence. The array of technological tools encompassed within this paradigm includes virtual and augmented reality, the Internet of Things (IoT), artificial intelligence and computer vision, the implementation of virtual assistants, Big Data management, cloud computing, process design and simulation programs, 3D printing [ 24 ], as well as developments in nanotechnology, biotechnology, and emerging quantum computing [ 25 ]. In a comprehensive study conducted from 1999 to 2019 [ 12 ], the publication of articles addressing the terms “Artificial Intelligence”, “Machine Learning”, and “Application” remained relatively stable until 2013, after which a remarkable increase of approximately 525% was observed in the Scopus database. The projections indicate that AI technologies are poised to enhance efficiency, optimize energy management, improve transparency, and promote the utilization of renewable energy sources. In recent years, advancements in AI technology have significantly transformed the methodologies employed in monitoring data, facilitating communication within energy systems, analyzing input-output relationships, and visualizing data in unprecedented ways. The integration of these AI developments into the energy sector renders new applications increasingly viable. To contextualize the integration of catalysis and its cross-cutting research areas into emerging artificial intelligence tools, we evaluated a series of conceptual links between relevant terms, as outlined in the methodology, and associated with these domains of knowledge. Bibliometric analysis (1991–2024) 1. Evolution of scientific publications in the field of catalysis related to artificial intelligence. The Lens® platform reports 2,413,151 publications related to catalysis and 552,685 related to artificial intelligence. The intersection of these domains accounts for only 618 records (0.02%), yet exhibits a notable annual growth rate of 18.34%, since 2019. These articles were published across 492 journals and involved 2,306 authors, of whom 97 were unique contributors. The earliest publication at the interface of AI and catalysis appeared in 1991, introducing a fuzzy temporal model designed to establish an expert system for fault diagnosis in a fluidized catalytic cracking unit [ 26 ]. Nevertheless, from 1991 to 2019, scientific output in this area remained stagnant. Beginning in 2020, however, a sustained and exponential increase was observed, culminating in a production 29 times higher than 2019 (Fig. 1 ). Between 2020 and 2024, transformative advances in AI reshaped the field of catalysis related to protein studies, it emerges as a transversal axis within this field. These included the application of text mining to large-scale databases, the use of Natural Language Processing (NLP) to establish structure–activity relationships in complex systems, the development of Large-Scale Language Models (LLMs), and the accurate prediction of protein three-dimensional structures from amino acid sequences (AlphaFold 2) as well as protein–ligand interactions (AlphaFold 3). Furthermore, multimodal AI systems emerged, capable of integrating and analyzing diverse types of information. In parallel, autonomous agents and self-driving laboratories were introduced, accelerating and optimizing experimental workflows. 2. Leading countries worldwide and in Ibero-American by number of scientific publications on AI–catalysis. Figure 2 illustrates that, at the global level, the United States and China are the leading contributors, each accounting for 17.80% of publications. They are followed by Germany (5.83%), the United Kingdom (4.53%), and India (4.53%). Collectively, the top ten countries represent 64.40% of all documents addressing applications of artificial intelligence in catalysis. The continental distribution of publications is as follows: Asia (27.02%) > America (21.68%) > Europe (12.62%) > Oceania (3.07%). Within the Ibero-American community, Spain is the leading contributor (1.62%), followed by Brazil (1.13%) and Mexico (0.32%). Colombia, Ecuador, Portugal, and Venezuela each account for 0.16% of the total output (Fig. 3 ). To date, only 7 of the 22 countries in the community have produced publications in this field, representing 3.72% of the overall contribution. 3. Most frequent terms in abstracts of AI–Catalysis publications. To identify the most frequently used concepts in research, an analysis of fixed two-term sequences (bigrams) was conducted in the abstracts of scientific publications. Figure 4 presents the 20 most common concepts, which were classified into five categories: Computer Science (Artificial intelligence, Machine learning, Neural networks, Deep learning, Learning models, Generative AI, Learning algorithms, and Molecular dynamics simulations); Biochemistry (Active site, Kinase activity, Allosteric communication, Outer membrane, Catalytic activity, Protein function, and Protein kinases); Chemistry/Engineering (Experimental data and Reaction conditions); Materials Science/Energy (Solar cells and Energy materials); and Physics/Chemistry (Blue light). Within Computer Science, the terms range from broad concepts to specific methodologies such as molecular dynamics simulations, deep learning, and generative AI. In Biochemistry, the identified terms are associated with protein structure and function, enzymatic activity, and cellular regulatory mechanisms. The category Materials Science/Energy, represented by Solar cells and Energy materials, corresponds to research on materials for solar energy capture and energy storage. In Physics/Chemistry, the term Blue light is potentially linked to photocatalysis and optical devices. Finally, the concepts “Experimental data” and “Reaction conditions” reflect empirical information derived from experiments, which can be used to validate theoretical models or computational simulations. 4. Trending topics at the Catalysis–AI intersection based on occurrences in scientific publication abstracts. The trend graphs were constructed using a Cartesian diagram in which the Y-axis represents the identified trending topics and the X-axis corresponds to the reference year (Fig. 5 ). Each node in the graph denotes a concept, with its size proportional to the number of occurrences. The year assigned to each term reflects the median of its distribution of occurrences over the analyzed period. Concepts were grouped according to their thematic relationships and reference year. Between 2001 and 2006, the terms FCC unit, process variables, catalytic cracking , and petroleum engineers were closely associated with the objectives of the energy and chemical process industries, particularly in relation to the optimization of conventional operations. In 2007, oil reservoirs and microbial communities appeared sporadically, reflecting the gradual incorporation of biotechnological approaches into petroleum applications. During 2009–2017, the emergence and consolidation of computational tools became evident. Terms such as process data, process modeling, cognitive science, genetic algorithms, fuzzy logic, computational intelligence, optimization algorithms, and computational models highlighted the growing use of advanced techniques to model, simulate, and optimize increasingly complex chemical processes. Several of these concepts have a long research history, underscoring the continuous evolution of computational approaches to industrial problem-solving. As shown in Fig. 1 , the period of greatest growth began in 2020, with complex systems, computational models , and operating conditions emerging prominently, signaling the application of mathematical and algorithmic tools in process analysis, simulation, and prediction. In 2021, protein engineering, catalyst discovery , and catalytic reaction became trending topics, indicating a shift in focus from process optimization toward biocatalysis and catalyst design. By 2022, neural networks, outer membranes , and crystal structures gained prominence, with neural networks ranking as the third most frequent term among the 36. Research increasingly relied on computational models enabling image analysis, property prediction, and structural characterization. In 2023, artificial intelligence, machine learning , and deep learning were consolidated as dominant topics, represented by the largest node sizes, reflecting their high frequency of occurrence. These concepts established themselves as the central technological axes for the analysis, optimization, and automation of complex processes. Their prevalence underscores a clear trend toward intensive digitization, the adoption of predictive algorithms, and the integration of data-driven approaches in catalysis research. Finally, in 2024, solar cells, energy materials , and reaction conditions emerged as leading topics, signaling a growing emphasis on renewable energy and the development of more efficient materials. 5. Density map of terms from the abstracts of scientific publications in the Catalysis-AI set. The density map reveals, from left to right, three main co-occurrence subspaces (see Fig. 6 ): the first two correspond to the domain of artificial intelligence, while the third is associated directly with catalysis. In addition, emerging and dispersed clusters highlight research trends in renewable energies and advanced materials (e.g., solar cells, renewable energy, solar power, waste management), consistent with the patterns observed in Fig. 5 for the year 2024. The persistent distance between the catalysis and AI clusters in the density map confirms that “Catalysis-AI” is indeed an emerging frontier rather than a fully consolidated field. The most concentrated area of terms is linked to artificial intelligence, followed by deep learning. The relatively small size of the clusters suggests that the knowledge generated so far has not yet reached a substantial volume. Moreover, the clear separation among the three subspaces indicates that the integration between AI-related topics and catalysis remains limited. As shown in Fig. 1 , convergence themes at the intersection of AI and catalysis are still in their early stages. The physical separation between these clusters reinforces the need for the “bridge” descriptors and multidisciplinary training proposed in earlier sections to foster a truly integrated Catalysis-AI domain. 6. Thematic evolution. A longitudinal thematic map analysis. To determine the conceptual structure of the Catalysis–AI collection, a strategic thematic map (Fig. 7 ) was constructed by analyzing bigrams extracted from the abstracts of each publication in the corpus. The map was organized into four quadrants according to Callon’s definition of density and centrality. Period 1991–2019 Figure 7 a displays nine clusters formed over 29 years (68 documents; 11.00%) and their distribution within the Callon diagram. In the first quadrant, clusters related to Fuzzy logic, Genetic algorithm [ 27 ], Artificial intelligence [ 28 , 29 ] and Machine learning [ 30 , 31 ] were identified, in increasing order of importance. These clusters exhibited high development and relevance, with their associated terms recognized as driving research topics. In contrast, the third quadrant contained clusters such as Cyber-physical systems, Fuzzy inference, Diagnostic devices, and Cognitive science , all characterized by low centrality and density, reflecting minimal development and importance. These themes disappeared in subsequent periods (2020–2021 and 2022–2024), suggesting a decline. Computational Intelligence [ 32 ] was located at the center of the diagram, with density and centrality values that prevent its clear assignment to any quadrant. Period 2020–2021 Despite being 14.5 times shorter, this period exhibited significant activity, producing 77 documents that account for 12.46% of the corpus. Nine thematic groups were identified, six of which are closely related to catalysis and three to artificial intelligence (Fig. 7 b). In the first quadrant, a prominent cluster characterized by high centrality and density includes the terms Learning models [ 33 ], Data-driven materials [ 34 ], and Functional materials . These emerging topics reflect noteworthy development and importance. The second cluster, exhibiting intermediate centrality and density, has remained a driving theme since the previous period. It consists of the terms Chemical reactions [ 35 ], E xperimental data , and Heterogeneous catalysis [ 6 ] A third cluster, demonstrating high relevance and substantial development, includes the terms Domain knowledge [ 36 ], Machine intelligence , and Paradigm shift , among others. This cluster is identified as a key theme that emerged during this period. In the fourth quadrant, another cluster was found with high centrality but lower density compared to the previous three groups. This basic and cross-cutting theme is well developed and constitutes the most frequently occurring group among all clusters: A rtificial intelligenc [ 37 – 41 ], it encompasses 32 terms, which can be divided into two subgroups: those related to catalysis as: P rotein engineering [ 42 , 43 ], A mino acids , C atalysis research , and C atalytic reactions ; as well as related to AI, including Machine learning [ 44 ], Neural networks , Deep learning , AB initio , Convolutional neural networks , Density functionals , and Learning techniques . The Artificial Intelligence group arises from the combination of the Artificial intelligence and Machine learning groups from the previous period. In the second quadrant, a cluster with high density but low centrality was identified, indicating limited importance. This niche topic comprises the following terms: Adsorption energy [ 45 ], Catalytic materials , Catalytic performance , and Reactivity descriptors . Another cluster with slightly lower density and occurrence includes: Enzyme engineering [ 46 ], Complex systems , Active sites , Chemical reactions , and Free energy . Finally, in the third quadrant, we identified three groups with low centrality and density, indicating topics of minimal relevance and development. These subjects are considered either emerging or declining, including Hydrogen production , Catalytic domain , and Organic frameworks . Period 2022–2024 In the last period, the number of publications increased by a factor of 8.5, reaching a total of 473 publications, which constitutes 76.54% of the overall corpus. Concurrently, the cluster structure demonstrated significant consolidation, decreasing from 9 to 3 clusters (Fig. 7 c). In the first quadrant, a cluster composed of 48-word pairs emerged with high density and centrality. The most recurrent pairs included Catalytic activity , Active site , and Protein structure [ 47 – 49 ]. These themes are well-developed and significant, identifying them as driving themes. At the center of the diagram, there exists a group with intermediate density and centrality, comprised of the terms Electron microscopy , Learning approaches , Identify potential , and Molecular mechanisms [ 50 – 52 ]. Due to its placement in the diagram, specific elements cannot be assigned to this group. Finally, in the third quadrant, we find the largest cluster, characterized by low density and centrality, consisting of the terms Artificial intelligence , Machine learning , Neural network, Deep learning , and Learning models , among others [ 21 , 53 – 57 ]. This group is located in the quadrant of emerging or declining topics. However, its evolutionary trajectory reveals a distinct dynamic: it began the study period in the motor skills quadrant with an incipient volume of publications, transitioned to basic and cross-cutting themes, and ultimately shifted to its current category. Given the volume of publications recorded, one can infer that these topics are not in a phase of decline, but rather may be undergoing a process of internal reconfiguration. The following graph provides a dynamic and concise representation of how these research areas have evolved over recent decades, achieved by grouping the different terms. Data Challenges: Standardization and Publication Bias The emergence of a robust “Catalysis-AI” paradigm is critically contingent upon the quality and nature of available data assets. Currently, the discipline faces the significant challenge of a lack of standardization in experimental reporting, which hinders the interoperability of datasets required for training deep learning models. Even more critical is the phenomenon of publication bias, where the exclusive dissemination of experimental successes and optimized yields prevails. For artificial intelligence algorithms to achieve true predictive capability, it is imperative that the scientific community begins to document and publish so-called “negative results” or failed reactions. These data are essential, as they provide the necessary contrast for models to identify reactivity boundaries and prevent overfitting. Ultimately, an AI that learns only from successes is a tool blind to the inherent complexity of the catalytic design space. Perspectives and Future Outlook Although Ibero-America's contribution to the intersection of catalysis and artificial intelligence stands at a mere 3.72%, the region possesses a wealth of fundamental chemical knowledge that could act as a “catalyst” for its own digital transformation. To bridge this gap, it is imperative to transition from a purely experimental research model toward the creation of regional consortia focused on data science. Such alliances would foster the democratization of access to supercomputing infrastructure and open databases—resources that are often prohibitively expensive for individual institutions. Furthermore, the early integration of chemometrics and machine learning into chemistry curricula is essential for cultivating a new generation of “bilingual” scientists proficient in both disciplines. Rather than competing in raw processing power with global leaders, the strategic focus should center on developing algorithms with low computational costs and applying AI to optimize processes of regional significance, such as biomass valorization, new catalytic materials and environmental catalysis. Expanding the reach of AI-Catalysis in these regions will require the strategic measures previously discussed, ensuring that fundamental chemical knowledge is effectively paired with data-driven infrastructures. Conclusions Despite the promising capabilities of the new generation of programs based on artificial intelligence, machine learning, metadata management, and correlation techniques to optimize experimental timelines automatically and efficiently, several challenges persist in the implementation of AI in the field of catalysis. Heterogeneous systems exhibit a few of interconnected parameters that evolve throughout the reaction, complicating the prediction of materials suitable for optimization. Additionally, considerations regarding the cost, abundance, and availability of metals identified by algorithms, as well as the technological resources necessary for the successful scale-up of catalyst synthesis, present further obstacles. However, bibliometric analyses indicate that a robust connection between the diverse areas of catalysis and emerging AI tools, has not yet been established. This gap is even more evident in Ibero-American countries, despite their strong tradition and achievements in catalysis. Such a disconnect may result in a substantial disparity between the pace of advances occurring in other scientific domains and those within catalysis. It is conceivable that the coming decades will herald the advent of a human-machine synergy that fosters scientific advancement, particularly in the design of novel catalytic materials. Abbreviations | AI | Artificial Intelligence | | LBD | Literature-Based Discovery | | CSV | Comma-Separated Values | | DOI | Digital Object Identifier | | CC | Callon Centrality | | CD | Callon Density | | XCON | eXpert CONfigurer | | DEC | Digital Equipment Corporation | | EMMS | Energy Minimization Multiscale Model | | PINNs | Physics-Informed Neural Networks | | I4.0 | Industry 4.0 | | CEO | Chief Executive Officer | | IoT | Internet of Things | | NLP | Natural Language Processing | | LLMs | Large-Scale Language Models | | FCC | Fluid Catalytic Cracking | Declarations If your study is a clinical trial, then please provide the necessary registration details (registry, trial registration number, and data of registration). The study is not a clinical trial: not applicable. Declarations Ethics and Consent to Participate Ethics and Consent to Participate: not applicable. Consent for Publication Consent for Publication: not applicable. Competing Interest There are no Competing Interests. Author Contribution The conceptual framework and the original idea for this study were conceived by S.P.C. The research was subsequently expanded through the collaborative efforts of the entire group. All authors contributed to the design of the figures, manuscript drafting, and critical revision. S.V. conducted the bibliometric analysis. Funding Funding: not applicable- Availability of data and materials The link is not available. The bibliometric data supporting the findings of this research are maintained by the authors and can be provided for review purposes upon request. Acknowledgments We extend our gratitude to the team at the Marcel Roche Regional Library of Science and Technology of the Venezuelan Institute of Scientific Research (IVIC) for their invaluable assistance in retrieving scientific articles pertinent to this study. References Li Z, Wang S, Xin H. Toward artificial intelligence in catalysis. Nat Catal. 2018;1:641–2. https://doi.org/10.1038/s41929-018-0150-1 . Donatien S. Challenges of Artificial Intelligence today and future implications for society and the world. World J Adv Res Rev. 2025;26:3045–54. https://doi.org/10.30574/wjarr.2025.26.1.1380 . Houhou R, Bocklitz T. Trends in artificial intelligence, machine learning, and chemometrics applied to chemical data. Anal Sci Adv. 2021;2:128–41. https://doi.org/10.1002/ansa.202000162 . Gaetani M, Mazwi M, Balaci H, et al. Artificial intelligence in medicine and the pursuit of environmentally responsible science. Lancet Digit Heal. 2024;6:e438–40. https://doi.org/10.1016/S2589-7500(24)00090-6 . Godbole NS, Lamb J. (2018) Research into Making Healthcare Green with Cloud, Green IT, and Data Science to Reduce Healthcare Costs and Combat Climate Change. In: 2018 9th IEEE Annual Ubiquitous Computing, Electronics and Mobile Communication Conference, UEMCON 2018. IEEE, pp 189–195. Bokhimi X. Learning the Use of Artificial Intelligence in Heterogeneous Catalysis. Front Chem Eng. 2021;3:1–8. https://doi.org/10.3389/fceng.2021.740270 . Amali Henadirage NG. Barriers to and Opportunities for the Adoption.pdf. Int J Artif Intell Educ. 2025;35:245–81. https://doi.org/10.1007/s40593-024-00439-5 . Eldar Haber, Jemielniak D, Kurasiński A, Przegalińska A. (2025) Future Trends and Emerging Tools. In: Macmillan P, editor Using AI in Academic Writing and Research. Cham. Fu V. (2025) AI for Science: Opportunities, Challenges, and Future Directions. TechRxiv. https://doi.org/10.36227/techrxiv.173949768.84003950/v1 Callon M, Courtial JP, Laville F. Co-word analysis as a tool for describing the network of interactions between basic and technological research: The case of polymer chemsitry. Scientometrics. 1991;22:155–205. https://doi.org/10.1007/BF02019280 . Turing AM. Computing Machinery and Intelligence. Mind. 1950;59:433–60. https://doi.org/10.1093/mind/xlvi.181.131 . Cioffi R, Travaglioni M, Piscitelli G, et al. Artificial intelligence and machine learning applications in smart production: Progress, trends, and directions. Sustain. 2020;12. https://doi.org/10.3390/su12020492 . Halal WE. Artificial intelligence is almost here. Horiz. 2003;11:37–8. https://doi.org/10.1108/10748120310486771 . Guo L, Wu J, Li J. Complexity at Mesoscales: A Common Challenge in Developing Artificial Intelligence. Engineering. 2019;5:924–9. https://doi.org/10.1016/j.eng.2019.08.005 . LI Xiangyu CA of S. (2018) Mesoscience: Discovering the Unknowns Between the Knowns. EurekAlert 29–31. Li J. Exploring the Logic and Landscape of the Knowledge System: Multilevel Structures, Each Multiscaled with Complexity at the Mesoscale. Engineering. 2016;2:276–85. https://doi.org/10.1016/J.ENG.2016.03.001 . Peng L, Wei Z. Recent progress of mesoscience in design of electrocatalytic materials for hydrogen energy conversion. Particuology. 2020;48:19–33. https://doi.org/10.1016/j.partic.2018.08.013 . Wu N, Ji X, Li L, et al. Mesoscience in supported nano-metal catalysts based on molecular thermodynamic modeling: A mini review and perspective. Chem Eng Sci. 2021;229:116164. https://doi.org/10.1016/j.ces.2020.116164 . Han J, Gong H, Ren X, Yan X. Supramolecular nanozymes based on peptide self-assembly for biomimetic catalysis. Nano Today. 2021;41:101295. https://doi.org/10.1016/j.nantod.2021.101295 . Yang W, Fidelis TT, Sun WH. Machine Learning in Catalysis, from Proposal to Practicing. ACS Omega. 2020;5:83–8. https://doi.org/10.1021/acsomega.9b03673 . Mazheika A, Wang YG, Valero R, et al. Artificial-intelligence-driven discovery of catalyst genes with application to CO2 activation on semiconductor oxides. Nat Commun. 2022;13:1–37. https://doi.org/10.1038/s41467-022-28042-z . Hirst JD, Boobier S, Coughlan J, et al. ML meets MLn: Machine learning in ligand promoted homogeneous catalysis. Artif Intell Chem. 2023;1:100006. https://doi.org/10.1016/j.aichem.2023.100006 . Ulissi ZW, Tang MT, Xiao J, et al. Machine-learning methods enable exhaustive searches for active Bimetallic facets and reveal active site motifs for CO 2 reduction. ACS Catal. 2017;7:6600–8. https://doi.org/10.1021/acscatal.7b01648 . Chen J, Wu P, Bu F, et al. 3D printing enhanced catalysis for energy conversion and environment treatment. DeCarbon. 2023;2:100019. https://doi.org/10.1016/j.decarb.2023.100019 . (2022) Industria 4.0, la cuarta revolución industrial y la inteligencia operacional. In: Consult Informático. https://www.cic.es/industria-40-revolucion-industrial/#:~:text= Daqun Q, Yongzai L. Fuzzy Temporal Knowledge Representation, Reasoning and Their Applications to Dynamic Systems. Acta Autom Sin. 1991;17:559–65. Nandi S, Badhe Y, Lonari J, et al. Hybrid process modeling and optimization strategies integrating neural networks/support vector regression and genetic algorithms: Study of benzene isopropylation on Hbeta catalyst. Chem Eng J. 2004;97:115–29. https://doi.org/10.1016/S1385-8947(03)00150-5 . Asadi S, Hassan M, Nadiri A, Dylla H. Artificial intelligence modeling to evaluate field performance of photocatalytic asphalt pavement for ambient air purification. Environ Sci Pollut Res. 2014;21:8847–57. https://doi.org/10.1007/s11356-014-2821-z . Pathak L, Singh V, Niwas R, et al. Artificial intelligence versus statistical modeling and optimization of cholesterol oxidase production by using Streptomyces sp. PLoS ONE. 2015;10:1–14. https://doi.org/10.1371/journal.pone.0137268 . Schlexer Lamoureux P, Winther KT, Garrido Torres JA, et al. Machine Learning for Computational Heterogeneous Catalysis. ChemCatChem. 2019;11:3581–601. https://doi.org/10.1002/cctc.201900595 . Bonk BM, Weis JW, Tidor B. Machine Learning Identifies Chemical Characteristics That Promote Enzyme Catalysis. J Am Chem Soc. 2019;141:4108–18. https://doi.org/10.1021/jacs.8b13879 . Remagnino P, Shapiro D. Artificial Intelligence Methods for Ambient Intelligence. Comput Intell. 2007;23:393–4. https://doi.org/10.1111/j.1467-8640.2007.00312.x . Kartashov OO, Chernov AV, Polyanichenko DS, Butakova MA. XAS data preprocessing of nanocatalysts for machine learning applications. Mater (Basel). 2021;14:7884. https://doi.org/10.3390/ma14247884 . Cole JM. A Design-to-Device Pipeline for Data-Driven Materials Discovery. Acc Chem Res. 2020;53:599–610. https://doi.org/10.1021/acs.accounts.9b00470 . Foppa L, Ghiringhelli LM, Girgsdies F, et al. Materials genes of heterogeneous catalysis from clean experiments and artificial intelligence. MRS Bull. 2021;46:1016–26. https://doi.org/10.1557/s43577-021-00165-6 . Vijayabaskar S. Harnessing Generative AI for Risk Management and Fraud Detection in Fintech: A New Era of Human-Machine Collaboration. Int J Sci Res Manag. 2020;8:369–79. https://doi.org/10.18535/ijsrm/v8i04.ec01 . Trunschke A, Bellini G, Boniface M, et al. Towards Experimental Handbooks in Catalysis. Top Catal. 2020;63:1683–99. https://doi.org/10.1007/s11244-020-01380-2 . Yi D, Bayer T, Badenhorst CPS, et al. Recent trends in biocatalysis. Chem Soc Rev. 2021;50:8003–49. https://doi.org/10.1039/d0cs01575j . Ge L, Yuan H, Min Y, et al. Predicted Optimal Bifunctional Electrocatalysts for the Hydrogen Evolution Reaction and the Oxygen Evolution Reaction Using Chalcogenide Heterostructures Based on Machine Learning Analysis of in Silico Quantum Mechanics Based High Throughput Screening. J Phys Chem Lett. 2020;11:869–76. https://doi.org/10.1021/acs.jpclett.9b03875 . Li P, Du Z, Chang C, et al. Efficient Catalytic Conversion of Waste Peanut Shells into Liquid Biofuel: An Artificial Intelligence Approach. Energy Fuels. 2020;34:1791–801. https://doi.org/10.1021/acs.energyfuels.9b03433 . Boucheikhchoukh A, Thibault J, Fauteux-Lefebvre C. Catalyst design using artificial intelligence: SO2 to SO3 case study. Can J Chem Eng. 2020;98:2016–31. https://doi.org/10.1002/cjce.23756 . Siedhoff NE, Illig AM, Schwaneberg U, Davari MD. PyPEF—An Integrated Framework for Data-Driven Protein Engineering. J Chem Inf Model. 2021;61:3463–76. https://doi.org/10.1021/acs.jcim.1c00099 . Xiong W, Liu B, Shen Y, et al. Protein engineering design from directed evolution to de novo synthesis. Biochem Eng J. 2021;174:108096. https://doi.org/10.1016/j.bej.2021.108096 . Siedhoff NE, Schwaneberg U, Davari MD. Machine learning-assisted enzyme engineering. Methods Enzymol. 2020;643:281–315. https://doi.org/10.1016/bs.mie.2020.05.005 . Wang B, Zhang F. Main Descriptors To Correlate Structures with the Performances of Electrocatalysts. Angew Int Ed Chemie. 2022;61:e202111026. https://doi.org/10.1002/anie.202111026 . Singh N, Malik S, Gupta A, Srivastava KR. Revolutionizing enzyme engineering through artificial intelligence and machine learning. Emerg Top Life Sci. 2021;5:113–25. https://doi.org/10.1042/ETLS20200257 . Aulakh SS Jr., Epand JCB RM. Exploring the AlphaFold Predicted Conformational Properties of Human Diacylglycerol Kinases. J Phys Chem B. 2022;126:7172–83. https://doi.org/https://doi.org/10.1021/acs.jpcb.2c04533 . Sajjadi E, Frascarelli C, Venetis K, et al. Computational pathology to improve biomarker testing in breast cancer: how close are we ? Eur J Cancer Prev. 2023;32:460–7. https://doi.org/10.1097/CEJ.0000000000000804 . Cochereau B, Strat Y, Le, Ji Q, et al. Heterologous Expression and Biochemical Characterization of a New Chloroperoxidase Isolated from the Deep – Sea Hydrothermal Vent Black Yeast Hortaea werneckii UBOCC – A – 208029. Mar Biotechnol. 2023;25:519–36. https://doi.org/10.1007/s10126-023-10222-7 . Mitchell S, Parés F, Faust Akl D, et al. Automated Image Analysis for Single-Atom Detection in Catalytic Materials by Transmission Electron Microscopy. J Am Chem Soc. 2022;144:8018–29. https://doi.org/10.1021/jacs.1c12466 . Ren F, Wu F, Wu X, et al. Fungal systems for lignocellulose deconstruction: From enzymatic mechanisms to hydrolysis optimization. GCB Bioenergy. 2024;16:e13130. https://doi.org/10.1111/gcbb.13130 . Höppner S, Schröder B, Fluhrer R. Structure and function of SPP/SPPL proteases: insights from biochemical evidence and predictive modeling. FEBS J. 2023;290:5456–74. https://doi.org/10.1111/febs.16968 . Mai H, Le TC, Chen D, et al. Machine Learning for Electrocatalyst and Photocatalyst Design and Discovery. Chem Rev Cite. 2022;122:13478–515. https://doi.org/https://doi.org/10.1021/acs.chemrev.2c00061 . Goodarzi N, Ashrafi-Peyman Z, Khani E, Moshfegh AZ. Recent Progress on Semiconductor Heterogeneous Photocatalysts in Clean Energy Production and Environmental Remediation. Catalysts. 2023;13:1102. https://doi.org/https://doi.org/10.3390/catal13071102 . Kouba P, Kohout P, Haddadi F, et al. Machine Learning-Guided Protein Engineering. ACS Catal. 2023;13. https://doi.org/10.1021/acscatal.3c02743 . :13863 – 13895. Zhang J, Fei Y, Sun L, Zhang QC. Advances and opportunities in RNA structure experimental determination and computational modeling. Nat Methods. 2022;19:19:1193–207. https://doi.org/https://doi.org/10.1038/s41592-022-01623-y . Zhang Z, Zheng Y, Qian L, et al. Emerging Trends in Sustainable CO2-Management Materials. Adv Mater. 2022;34:2201547. https://doi.org/https://doi.org/10.1002/adma.202201547 . Additional Declarations No competing interests reported. Supplementary Files GA.png Graphical Abstract The graphical abstract was generated using NotebookLM, based on the following prompt: Minimalist scientific style with a white background. Display a shower of loose puzzle pieces of different sizes orchestrated from largest to smallest that generate a waveform entering through the left side of a computer that displays basic programming code on its screen. Connected to the right side of this computer by a cable is an android. This android acts as a teacher, writing science-related information on a screen. This whiteboard should contain a reaction coordinate graph, a heterogeneous catalyst model, and chemical molecules. Cite Share Download PDF Status: Under Revision Version 1 posted Editorial decision: Revision requested 18 May, 2026 Reviews received at journal 01 May, 2026 Reviews received at journal 28 Apr, 2026 Reviews received at journal 19 Apr, 2026 Reviewers agreed at journal 18 Apr, 2026 Reviews received at journal 11 Apr, 2026 Reviewers agreed at journal 10 Apr, 2026 Reviewers agreed at journal 10 Apr, 2026 Reviewers agreed at journal 10 Apr, 2026 Reviewers invited by journal 10 Apr, 2026 Editor invited by journal 02 Apr, 2026 Editor assigned by journal 28 Mar, 2026 Submission checks completed at journal 28 Mar, 2026 First submitted to journal 20 Mar, 2026 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-9182089","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":625732840,"identity":"d236a1b9-31f3-4bcc-afef-784da2d224a2","order_by":0,"name":"Samuel Villanueva","email":"","orcid":"","institution":"Instituto Venezolano de Investigaciones Científicas","correspondingAuthor":false,"prefix":"","firstName":"Samuel","middleName":"","lastName":"Villanueva","suffix":""},{"id":625732841,"identity":"2884adf2-9e65-4bd6-b9ed-2e7e011af2a4","order_by":1,"name":"Lisbeth Mendoza","email":"","orcid":"","institution":"Instituto Venezolano de Investigaciones Científicas","correspondingAuthor":false,"prefix":"","firstName":"Lisbeth","middleName":"","lastName":"Mendoza","suffix":""},{"id":625732842,"identity":"0696d1df-4944-4584-b16c-4898e6a13cd7","order_by":2,"name":"Paulino Betancourt","email":"","orcid":"","institution":"Central University of Venezuela","correspondingAuthor":false,"prefix":"","firstName":"Paulino","middleName":"","lastName":"Betancourt","suffix":""},{"id":625732843,"identity":"e012f82d-934e-4dd7-8e33-fdcc55ea48a6","order_by":3,"name":"Susana Pinto-Castilla","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAu0lEQVRIiWNgGAWjYLCCDxU2QJKx8QDROhhnnEkDUQ3Ea2HmbTsMZhCnhb//7MHPvG3n7da2HwbaUmMTTVCLxI28ZMk5524nbzuTCNRyLC23gZAWAwkeM4Y3ZbeTzQ4AtTA2HCZCC/8ZMwYetnPJZucfEquFIceMkaftgJ3ZDWJtAftlxpnkBLMbQFsSiPELKMQ+fKiwszc7n/7wwYcaG8JaGBh4wGQiWGUCYeUILfbEKR4Fo2AUjIIRCQDxqUfnpAGu9gAAAABJRU5ErkJggg==","orcid":"","institution":"Instituto Venezolano de Investigaciones Científicas","correspondingAuthor":true,"prefix":"","firstName":"Susana","middleName":"","lastName":"Pinto-Castilla","suffix":""}],"badges":[],"createdAt":"2026-03-20 21:53:12","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-9182089/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9182089/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":107315888,"identity":"3d2fc6ec-350f-435a-a615-e17aff4ceda0","added_by":"auto","created_at":"2026-04-20 09:50:58","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":380987,"visible":true,"origin":"","legend":"\u003cp\u003eEvolution of scientific publications related to catalysis and artificial intelligence\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-9182089/v1/eb04e53a40126f934a8ea7cc.png"},{"id":107486091,"identity":"ffe60704-fc63-4702-8992-b473cca958df","added_by":"auto","created_at":"2026-04-22 02:37:23","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":227842,"visible":true,"origin":"","legend":"\u003cp\u003eCountries with the highest scientific publications in the field of catalysis related to artificial intelligence\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-9182089/v1/1f6a79169218a0dd8a03c5e2.png"},{"id":107315890,"identity":"b1ad52a6-e347-492f-bc68-20090a57644b","added_by":"auto","created_at":"2026-04-20 09:50:58","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":151773,"visible":true,"origin":"","legend":"\u003cp\u003eIbero-American countries with the highest number of scientific publications in the field of catalysis related to artificial intelligence\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-9182089/v1/4bef76ea5ac38528ff6faf8c.png"},{"id":107484596,"identity":"187c9835-94b8-49ee-85a0-d59553aa92b8","added_by":"auto","created_at":"2026-04-22 02:32:28","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":518885,"visible":true,"origin":"","legend":"\u003cp\u003eMost frequent terms identified in the abstracts of scientific publications on Catalysis and Artificial Intelligence\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-9182089/v1/11f2bd8d481f22d4ae941b17.png"},{"id":107315892,"identity":"6a3a7e4e-212a-4965-83ac-8c9df4ff7c49","added_by":"auto","created_at":"2026-04-20 09:50:58","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":1872004,"visible":true,"origin":"","legend":"\u003cp\u003eTrending Topics at the Catalysis–AI based on occurrences in scientific publication abstracts\u003c/p\u003e","description":"","filename":"5.png","url":"https://assets-eu.researchsquare.com/files/rs-9182089/v1/78bcd6998b67daede295f191.png"},{"id":107315893,"identity":"07f8936f-febd-4ff3-91ed-c8ddfdb8b809","added_by":"auto","created_at":"2026-04-20 09:50:58","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":1657273,"visible":true,"origin":"","legend":"\u003cp\u003eDensity map of terms from the abstracts of scientific publications in the Catalysis–AI dataset\u003c/p\u003e","description":"","filename":"6.png","url":"https://assets-eu.researchsquare.com/files/rs-9182089/v1/26ba56683ce9b2254c901d50.png"},{"id":107488274,"identity":"37ec96df-b3a3-41e0-99cd-2fd284d738b3","added_by":"auto","created_at":"2026-04-22 02:44:04","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":635929,"visible":true,"origin":"","legend":"\u003cp\u003eThematic Map of terms from scientific publications in the Catalysis–AI: a) 1991-2019; b) 2020-2021 y c) 2022-2024\u003c/p\u003e","description":"","filename":"7.png","url":"https://assets-eu.researchsquare.com/files/rs-9182089/v1/6c67c2fe843ad3aae98c4401.png"},{"id":107315895,"identity":"35dd7837-2049-489e-b9c9-e07400206c8c","added_by":"auto","created_at":"2026-04-20 09:50:58","extension":"png","order_by":8,"title":"Figure 8","display":"","copyAsset":false,"role":"figure","size":180686,"visible":true,"origin":"","legend":"\u003cp\u003eThematic evolution in three time slices\u003c/p\u003e","description":"","filename":"8.png","url":"https://assets-eu.researchsquare.com/files/rs-9182089/v1/f2d843278335db88ad52031d.png"},{"id":107489548,"identity":"0da27528-44ff-4c02-9aa3-1428b386c44b","added_by":"auto","created_at":"2026-04-22 02:48:09","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":5266368,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9182089/v1/d3c9653b-c91e-45c5-9445-5c3923882e1a.pdf"},{"id":107315887,"identity":"db4bcfd2-10db-41b3-8c08-d0160b0e2521","added_by":"auto","created_at":"2026-04-20 09:50:57","extension":"png","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":5743909,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eGraphical Abstract\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe graphical abstract was generated using NotebookLM, based on the following prompt: \u003cem\u003eMinimalist scientific style with a white background. Display a shower of loose puzzle pieces of different sizes orchestrated from largest to smallest that generate a waveform entering through the left side of a computer that displays basic programming code on its screen. Connected to the right side of this computer by a cable is an android. This android acts as a teacher, writing science-related information on a screen. This whiteboard should contain a reaction coordinate graph, a heterogeneous catalyst model, and chemical molecules.\u003c/em\u003e\u003c/p\u003e","description":"","filename":"GA.png","url":"https://assets-eu.researchsquare.com/files/rs-9182089/v1/59248437d2945fa001839e9d.png"}],"financialInterests":"No competing interests reported.","formattedTitle":"\u003cp\u003eThe Emergence of Catalysis-ai: A New Frontier in Catalysis Research?\u003c/p\u003e","fulltext":[{"header":"Highlights","content":"\u003cp\u003eRecent advancements in artificial intelligence have significantly impacted science and technology.\u003c/p\u003e\n\u003cp\u003eDespite the concerns and potential risks associated with artificial intelligence, the responsible integration of these tools is essential for the advancement of science and technology, especially in the field of catalysis.\u003c/p\u003e\n\u003cp\u003eIt is necessary to integrate catalysis, with the diverse tools of artificial intelligence, ultimately giving rise to a new paradigm of catalysis–AI.\u003c/p\u003e\n\u003cp\u003eThe disconnect between catalysis and artificial intelligence is particularly pronounced in Ibero-American countries, which have served as fundamental pillars in the advancement of catalytic knowledge for decades.\u003c/p\u003e"},{"header":"Introduction","content":"\u003cp\u003eThroughout history, there have been numerous periods characterized by technological advancements that have catalyzed significant scientific discoveries and innovations, fundamentally transforming society. The discourse surrounding artificial intelligence (AI) often emphasizes its potential risks, including algorithmic bias, systemic discrimination, widespread job displacement, and even existential threats to humanity [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e, \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e]. While some commentators express concern over these dystopian scenarios, others advocate for the substantial benefits that AI may offer [\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]. Proponents argue that AI has the capacity to aid humanity in addressing some of its most formidable and intricate challenges by dramatically accelerating the pace of scientific discovery, particularly in fields such as medicine, climate science, sustainable technology [\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e, \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e], transforming even the traditional teaching methodologies [\u003cspan additionalcitationids=\"CR7\" citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eAI-driven tools are increasingly being integrated across virtually all domains of science [\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]. For instance, they can efficiently identify promising candidates for the synthesis of novel materials and analyze vast datasets to uncover patterns and model complex systems. Within this context, two domains exhibit remarkable potential to revolutionize scientific practice: \u0026ldquo;literature-based discovery\u0026rdquo; (LBD) and \u0026ldquo;autonomous laboratories\u0026rdquo; or robotic scientists. LBD employs AI language analysis to scrutinize existing scientific literature, enabling the identification of new hypotheses, connections, and insights, thereby fostering interdisciplinary collaboration and innovation. Conversely, self-directed laboratories utilize AI to autonomously generate and test hypotheses through the execution of hundreds or thousands of experimental trials.\u003c/p\u003e \u003cp\u003eMoreover, the integration of AI into scientific research has the potential to facilitate widespread access to cutting-edge knowledge and techniques. By automating complex analyses and lowering the barriers to entry in high-tech research environments, AI tools can empower researchers from diverse backgrounds and institutions, including those in developing regions, to contribute to global scientific advancements. These tools can promote a more inclusive scientific community and enhance the diversity of perspectives that drive innovation.\u003c/p\u003e \u003cp\u003eFollowing this line of inquiry, we explored the connection between catalysis and its integration with diverse AI tools over recent decades. Our approach began with an examination of the various definitions associated with this emerging interdisciplinary field. In addition, we conducted a bibliometric analysis using the scientific publication platform The Lens\u0026reg;, applying filters by document type, including journal articles, review papers, letters, and conference proceedings. All bibliometric analysis encompasses publications spanning the period 1991\u0026ndash;2024.\u003c/p\u003e"},{"header":"Methodology","content":"\u003cp\u003eA comprehensive search was performed on the scientific publishing platform The Lens\u0026reg; (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.lens.org/\u003c/span\u003e\u003cspan address=\"https://www.lens.org/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e) using the following query: (catalys OR catalyst OR catalyzed OR catalytic OR biocatalysis OR biocatalytic OR \u0026ldquo;enzyme technology\u0026rdquo; OR \u0026ldquo;enzyme engineering\u0026rdquo; OR photocatalytic OR photocatalyst OR \u0026ldquo;photo-catalytic\u0026rdquo; OR electrocatalysis OR \u0026ldquo;electro-catalytic activity\u0026rdquo; OR nanocatalysts OR nanocatalysis) AND (\u0026ldquo;artificial intelligence\u0026rdquo; OR \u0026ldquo;machine intelligence\u0026rdquo; OR \u0026ldquo;computational intelligence\u0026rdquo;). The search was applied to titles, abstracts, keywords, and fields of study. Results were filtered by document type, including journal articles, review articles, letters, and conference proceedings. A total of 725 records were retrieved, between 1991\u0026ndash;2024, of which 107 were excluded after manual screening for irrelevance to the study objectives. The remaining records were exported in CSV format, containing metadata such as title, abstract, publication date and year, journal, authors, field of study, DOI, and related information. The search was conducted globally, and bibliometric indicators of scientific output over time were generated, categorized by \u003cem\u003eevolution of scientific publications\u003c/em\u003e, \u003cem\u003ecountries, most frequent terms in abstracts, and trending topics.\u003c/em\u003e\u003c/p\u003e \u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eBibliometric study\u003c/h2\u003e \u003cp\u003eData processing was carried out using the Biblioshiny interface of the Bibliometrix package, where trivial terms were removed and synonyms consolidated. It is important to note that Bibliometrix R package, (version 4.3.0) is a tool designed for scientometric and quantitative bibliometric analyses. Its implementation in the R programming language leverages the advantages of an open-source environment and a robust ecosystem. VOSviewer (version 1.6.19), in contrast, is specialized software for the visualization of bibliographic networks, enabling the exploration of relationships and patterns among authors, institutions, and keywords through co-occurrence and co-citation analyses. In this study, VOSviewer was employed specifically for visualization of bibliometric maps in both network and density formats. All search, analysis, validation, and data extraction procedures were conducted between July 29 and August 10, 2025.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eEvolution of the Thematic Map\u003c/h3\u003e\n\u003cp\u003eTo elucidate the conceptual structure of the study, a thematic map was generated by analyzing bigrams extracted from the abstracts of each publication using the Walktrap algorithm. The map was organized into four quadrants according to Callon\u0026rsquo;s framework of density and centrality. In this representation, nodes are characterized by two measures: Callon centrality (CC) and Callon density (CD) [\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e]. Callon centrality quantifies the importance of a theme or node within the overall network, whereas Callon density reflects the degree of development of the theme. Based on these two measures, research topics can be positioned within a two-dimensional diagram comprising four quadrants: (1) upper right, driving themes; (2) lower right, basic themes; (3) lower left, emerging or declining themes; and (4) upper left, niche or highly specialized themes.\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eThe question of whether machines can think was first posed by Alan Turing in his seminal 1950 paper, \u0026ldquo;Computing Machinery and Intelligence\u0026rdquo; [\u003cspan class=\"CitationRef\"\u003e11\u003c/span\u003e]. In this work, Turing contends that addressing this question necessitates a clear definition of the terms: \u0026ldquo;machine\u0026rdquo; and \u0026ldquo;thinking\u0026rdquo;. However, precisely defining \u0026ldquo;thinking\u0026rdquo; presents a challenge, as it is inherently subjective. To circumvent this issue, Turing proposed an indirect approach to assess a machine\u0026rsquo;s ability to exhibit intelligence indistinguishable from that of a human. This assessment, now known as the Turing Test, involves a human engaging in conversation with both a computer and another human, without knowing which participant is which. If the human is unable to reliably discern the machine from the human interlocutor, the machine is deemed intelligent.\u003c/p\u003e\n\u003cp\u003eThe term \u0026ldquo;Artificial Intelligence\u0026rdquo; (AI) was formally defined by John McCarthy during the Dartmouth Conference in 1956 as \u0026ldquo;the science and engineering of making intelligent machines\u0026rdquo;. This landmark event is widely regarded as the inception of the field of Artificial Intelligence. Presently, the domains of artificial intelligence can be broadly categorized into 16 areas, which include [\u003cspan class=\"CitationRef\"\u003e12\u003c/span\u003e]: reasoning, programming, artificial life, belief revision, data mining, distributed AI, expert systems, genetic algorithms, knowledge representation, machine learning, natural language understanding, neural networks, theorem proving, constraint satisfaction, and the theory of computation. These categories collectively encompass the diverse methodologies and applications that define contemporary AI research and development.\u003c/p\u003e\n\u003cp\u003eAI experienced a significant surge in popularity during the 1980s, marked by the development of \u0026ldquo;expert systems\u0026rdquo; by various research institutions and universities. These systems were designed to encapsulate a set of foundational rules derived from expert knowledge, thereby assisting non-experts in making informed decisions and addressing real-world problems for the first time. Notable examples of such systems include XCON (eXpert CONfigurer), developed by Carnegie Mellon University, and MYCIN, created at Stanford University.\u003c/p\u003e\n\u003cp\u003eThe XCON program was implemented at Digital Equipment Corporation (DEC), a pioneer in the minicomputer manufacturing sector. By 1986, XCON had processed approximately 80,000 orders, achieving an impressive accuracy rate of nearly 98%. It was estimated that the system saved DEC around $25\u0026nbsp;million annually. MYCIN, on the other hand, was designed to diagnose infectious diseases and recommend antibiotic dosages based on patients\u0026rsquo; weight. Research conducted by Stanford Medical School indicated that MYCIN\u0026rsquo;s diagnostic success rate was approximately 65%, while human specialists achieved a success rate of about 80%. These early expert systems laid the groundwork for subsequent advancements in AI and its applications in various fields.\u003c/p\u003e\n\u003cp\u003eHowever, over time, expert systems have disclosed several inherent limitations, including a lack of flexibility, restricted versatility, high maintenance costs, and various ethical and legal dilemmas. For instance, in the event that MYCIN erroneously diagnosed a patient, the question arises: who would bear responsibility\u0026mdash;the programmer or the physician?\u003c/p\u003e\n\u003cp\u003eDuring the same decade, Geoffrey Hinton and his colleagues made significant advancements in the development of the backpropagation algorithm, which facilitates machine learning and underpins nearly all contemporary neural networks. Nevertheless, it was not until 2006 that neural networks based on backpropagation began to exert a substantial influence on the field. Notably, one of the so-called \u0026ldquo;godfathers\u0026rdquo; of artificial intelligence resigned from his position about Google, expressing a reconsideration of his stance by stating: \u0026ldquo;I\u0026rsquo;ve suddenly changed my mind about whether these systems will surpass human intelligence\u0026rdquo;.\u003c/p\u003e\n\u003cp\u003eIn the 21st century, artificial intelligence has emerged as a critical domain of research, encompassing a diverse array of fields, including engineering, science, education, medicine, marketing, accounting, finance, economics, and law, among others [\u003cspan class=\"CitationRef\"\u003e13\u003c/span\u003e]. The scope of AI has expanded tremendously, as machine intelligence equipped with machine learning capabilities has produced significant impacts across businesses, government entities, and scientific inquiry. In this context, artificial intelligence holds promise in addressing critical issues related to sustainable production, optimizing energy resources, managing supply chains, and reducing waste. For example, in the realm of smart manufacturing, there is a growing trend to integrate AI into green manufacturing processes to adhere to the most stringent environmental standards.\u003c/p\u003e\n\u003ch3\u003eDiscovering New Research Methodologies: Mesoscience\u003c/h3\u003e\n\u003cp\u003eDespite the numerous scientific discoveries and advancements achieved to date, researchers continue to grapple with the enigmatic gaps that persist within the chains of knowledge. Regardless of the extent of information available, our understanding of the complex interrelationships that often exist between a system \u0026ldquo;as a whole\u0026rdquo; and its constituent elements remains limited. Such connections can manifest in various phenomena, including violent tornadoes, intricate neural networks, and fluctuating reactive patterns, among others. To the astonishment of scientists, relationships observed in seemingly unrelated domains reveal profound and intriguing similarities. Describing, predicting, and manipulating this complexity poses a significant challenge across nearly all fields of research and represents a bottleneck in numerous industrial processes.\u003c/p\u003e\n\u003cp\u003eIs there a viable approach to decode these phenomena? A group of Chinese researchers is pursuing this inquiry and has introduced the concept of \u0026ldquo;mesoscience\u0026rdquo;. This methodology aims to elucidate the mechanisms underlying the multiscale spatiotemporal structures of complex systems and to enhance the analysis of these systems through the use of computational models and paradigms [\u003cspan class=\"CitationRef\"\u003e14\u003c/span\u003e, \u003cspan class=\"CitationRef\"\u003e15\u003c/span\u003e]. Specifically, mesoscience seeks to bridge the gap between the macroscale (system) and the microscale (unit).\u003c/p\u003e\n\u003cp\u003eTo elucidate the common principles and rules within mesoscience, the Energy Minimization Multiscale Model (EMMS), has been developed [\u003cspan class=\"CitationRef\"\u003e16\u003c/span\u003e]. This model is formulated as a multi-objective variational problem, accounting for trade-offs that arise when competing factors are considered.\u003c/p\u003e\n\u003cp\u003eThe convergence between artificial intelligence and the mesoscience paradigm offers a promising route to overcome the intrinsic computational limitations of the EMMS model. Historically, solving the multi-objective variational problems associated with this model has demanded massive processing resources, restricting its applicability in complex operational scenarios. In this context, the implementation of Physics-Informed Neural Networks (PINNs) emerges as a disruptive tool. These algorithms do not merely act as universal approximators; they integrate fundamental physical laws directly into the network's loss function, ensuring the model's consistency with conservation principles. By leveraging this capability, it is possible to drastically accelerate calculation convergence. Consequently, AI enables real-time simulations that, until recently, were computationally prohibitive for the design of multiscale catalytic systems. The following examples illustrate how these multiscale challenges are being addressed through the integration of mesoscience and advanced computational tools.\u003c/p\u003e\n\u003cp\u003eGuided by this perspective, the complexity of catalytic systems has garnered increasing attention in recent years and is being investigated from a mesoscience perspective. For instance, the generation of greener materials for the production of clean, efficient, and cost-effective energy necessitates the design of more effective and environmentally friendly energy conversion and storage devices. In this context, electrocatalytic materials with varying morphologies, sizes, and compositions, exhibit distinct catalytic responses in these applications, depending on their specific configurations. Consequently, the relationship between the catalytic behavior of electrocatalytic materials and these factors has begun to be explored at multiple levels through the lens of mesoscience.\u003c/p\u003e\n\u003cp\u003eFor example, Peng and Wei [\u003cspan class=\"CitationRef\"\u003e17\u003c/span\u003e] examined the integrated morphology of the electrode, which encompasses the structure of the catalyst\u0026rsquo;s pores, the reaction interface, and the active site, to demonstrate how mesoscience can aid in the design of more efficient electrocatalytic materials by modulating their geometric and electronic structures. Similarly, Wu and colleagues [\u003cspan class=\"CitationRef\"\u003e18\u003c/span\u003e] investigated the design of supported nanometallic catalysts using mesoscience to understand the equilibrium between high catalytic activity (designated as mechanism A) and high catalyst stability (mechanism B) in the synthesis of hydrogen peroxide (H\u003csub\u003e2\u003c/sub\u003eO\u003csub\u003e2\u003c/sub\u003e). Among their findings, the researchers identified that the heat of reaction and the enthalpy of fusion of the supported nanometallic catalyst were significant factors influencing mechanism B, thereby providing critical insights for optimizing catalyst design.\u003c/p\u003e\n\u003cp\u003eIn the other study about enzymatic systems, integrated micro-macro methodologies had yielded valuable contributions by examining the various factors that modify catalytic responses [\u003cspan class=\"CitationRef\"\u003e19\u003c/span\u003e]. This approach aims to develop a new generation of supramolecular compounds through the assembly of enzymes, resulting in systems with enhanced catalytic selectivity and efficiency for a wide range of applications.\u003c/p\u003e\n\u003ch3\u003eSimplifying Calculations with Machine Learning\u003c/h3\u003e\n\u003cp\u003eMachine learning, a subfield of artificial intelligence, enables machines to learn autonomously without explicit programming. This technology is predicated on the identification of patterns within data to facilitate predictions. In contrast to traditional computing, which relies on coded instructions crafted by experts, machine learning platforms independently learn through increasing exposure to examples, utilizing algorithms to refine their understanding.\u003c/p\u003e\n\u003cp\u003eIn essence, this approach provides the opportunity to train computers on the known catalytic properties of various materials, thereby predicting the most promising potential catalysts for a reaction of interest within an automated machine learning framework for catalyst design [\u003cspan class=\"CitationRef\"\u003e20\u003c/span\u003e\u0026ndash;\u003cspan class=\"CitationRef\"\u003e22\u003c/span\u003e]. A notable study by Ulissi et al. [\u003cspan class=\"CitationRef\"\u003e23\u003c/span\u003e] made significant strides toward establishing such an automated machine learning framework for catalyst design. The systematic methodology employed in this research led to the identification of an active site that had previously been overlooked in nickel and gallium systems during the electrochemical reduction of CO\u003csub\u003e2\u003c/sub\u003e: Ni atoms encircled by Ga atoms at the surface. These newly identified active sites demonstrated enhanced thermodynamic parameters and exhibited step-like kinetic behavior. This discovery substantially advances our understanding of the factors that contribute to the improved design of bimetallic catalysts.\u003c/p\u003e\n\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e\n\u003ch2\u003eToward Industry 4.0\u003c/h2\u003e\n\u003cp\u003eSmart production systems necessitate innovative solutions to enhance the quality and sustainability of manufacturing activities while concurrently reducing costs. In this context, artificial intelligence AI-driven technologies, in conjunction with key Industry 4.0 (I4.0) advancements, are instrumental in optimizing and monitoring manufacturing processes. These technologies enable the integration of various processes and systems within the plant, heralding a new industrial revolution characterized by the convergence of physical and digital dimensions, which is expected to generate novel industrial paradigms.\u003c/p\u003e\n\u003cp\u003eThe term \u0026ldquo;Industry 4.0\u0026rdquo; was first introduced at the Hannover Fair in 2011, during the opening ceremony delivered by Professor Wolfgang Wahlster, Director and CEO of the German Research Center for Artificial Intelligence. The array of technological tools encompassed within this paradigm includes virtual and augmented reality, the Internet of Things (IoT), artificial intelligence and computer vision, the implementation of virtual assistants, Big Data management, cloud computing, process design and simulation programs, 3D printing [\u003cspan class=\"CitationRef\"\u003e24\u003c/span\u003e], as well as developments in nanotechnology, biotechnology, and emerging quantum computing [\u003cspan class=\"CitationRef\"\u003e25\u003c/span\u003e].\u003c/p\u003e\n\u003cp\u003eIn a comprehensive study conducted from 1999 to 2019 [\u003cspan class=\"CitationRef\"\u003e12\u003c/span\u003e], the publication of articles addressing the terms \u0026ldquo;Artificial Intelligence\u0026rdquo;, \u0026ldquo;Machine Learning\u0026rdquo;, and \u0026ldquo;Application\u0026rdquo; remained relatively stable until 2013, after which a remarkable increase of approximately 525% was observed in the Scopus database.\u003c/p\u003e\n\u003cp\u003eThe projections indicate that AI technologies are poised to enhance efficiency, optimize energy management, improve transparency, and promote the utilization of renewable energy sources. In recent years, advancements in AI technology have significantly transformed the methodologies employed in monitoring data, facilitating communication within energy systems, analyzing input-output relationships, and visualizing data in unprecedented ways. The integration of these AI developments into the energy sector renders new applications increasingly viable.\u003c/p\u003e\n\u003cp\u003eTo contextualize the integration of catalysis and its cross-cutting research areas into emerging artificial intelligence tools, we evaluated a series of conceptual links between relevant terms, as outlined in the methodology, and associated with these domains of knowledge.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eBibliometric analysis (1991\u0026ndash;2024)\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cem\u003e1. Evolution of scientific publications in the field of catalysis related to artificial intelligence.\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eThe Lens\u0026reg; platform reports 2,413,151 publications related to catalysis and 552,685 related to artificial intelligence. The intersection of these domains accounts for only 618 records (0.02%), yet exhibits a notable annual growth rate of 18.34%, since 2019. These articles were published across 492 journals and involved 2,306 authors, of whom 97 were unique contributors. The earliest publication at the interface of AI and catalysis appeared in 1991, introducing a fuzzy temporal model designed to establish an expert system for fault diagnosis in a fluidized catalytic cracking unit [\u003cspan class=\"CitationRef\"\u003e26\u003c/span\u003e]. Nevertheless, from 1991 to 2019, scientific output in this area remained stagnant. Beginning in 2020, however, a sustained and exponential increase was observed, culminating in a production 29 times higher than 2019 (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e).\u003c/p\u003e\n\u003cp\u003eBetween 2020 and 2024, transformative advances in AI reshaped the field of catalysis related to protein studies, it emerges as a transversal axis within this field. These included the application of text mining to large-scale databases, the use of Natural Language Processing (NLP) to establish structure\u0026ndash;activity relationships in complex systems, the development of Large-Scale Language Models (LLMs), and the accurate prediction of protein three-dimensional structures from amino acid sequences (AlphaFold 2) as well as protein\u0026ndash;ligand interactions (AlphaFold 3). Furthermore, multimodal AI systems emerged, capable of integrating and analyzing diverse types of information. In parallel, autonomous agents and self-driving laboratories were introduced, accelerating and optimizing experimental workflows.\u003c/p\u003e\n\u003cp\u003e2.\u0026nbsp;\u003cem\u003eLeading countries worldwide and in Ibero-American by number of scientific publications on AI\u0026ndash;catalysis.\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eFigure \u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003e illustrates that, at the global level, the United States and China are the leading contributors, each accounting for 17.80% of publications. They are followed by Germany (5.83%), the United Kingdom (4.53%), and India (4.53%). Collectively, the top ten countries represent 64.40% of all documents addressing applications of artificial intelligence in catalysis. The continental distribution of publications is as follows: Asia (27.02%) \u0026gt; America (21.68%) \u0026gt; Europe (12.62%) \u0026gt; Oceania (3.07%).\u003c/p\u003e\n\u003cp\u003eWithin the Ibero-American community, Spain is the leading contributor (1.62%), followed by Brazil (1.13%) and Mexico (0.32%). Colombia, Ecuador, Portugal, and Venezuela each account for 0.16% of the total output (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003e). To date, only 7 of the 22 countries in the community have produced publications in this field, representing 3.72% of the overall contribution.\u003c/p\u003e\n\u003cp\u003e3.\u0026nbsp;\u003cem\u003eMost frequent terms in abstracts of AI\u0026ndash;Catalysis publications.\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eTo identify the most frequently used concepts in research, an analysis of fixed two-term sequences (bigrams) was conducted in the abstracts of scientific publications. Figure\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e4\u003c/span\u003e presents the 20 most common concepts, which were classified into five categories: \u003cstrong\u003eComputer Science\u003c/strong\u003e (Artificial intelligence, Machine learning, Neural networks, Deep learning, Learning models, Generative AI, Learning algorithms, and Molecular dynamics simulations); \u003cstrong\u003eBiochemistry\u003c/strong\u003e (Active site, Kinase activity, Allosteric communication, Outer membrane, Catalytic activity, Protein function, and Protein kinases); \u003cstrong\u003eChemistry/Engineering\u003c/strong\u003e (Experimental data and Reaction conditions); \u003cstrong\u003eMaterials Science/Energy\u003c/strong\u003e (Solar cells and Energy materials); and \u003cstrong\u003ePhysics/Chemistry\u003c/strong\u003e (Blue light).\u003c/p\u003e\n\u003cp\u003eWithin Computer Science, the terms range from broad concepts to specific methodologies such as molecular dynamics simulations, deep learning, and generative AI. In Biochemistry, the identified terms are associated with protein structure and function, enzymatic activity, and cellular regulatory mechanisms. The category Materials Science/Energy, represented by Solar cells and Energy materials, corresponds to research on materials for solar energy capture and energy storage. In Physics/Chemistry, the term Blue light is potentially linked to photocatalysis and optical devices. Finally, the concepts \u0026ldquo;Experimental data\u0026rdquo; and \u0026ldquo;Reaction conditions\u0026rdquo; reflect empirical information derived from experiments, which can be used to validate theoretical models or computational simulations.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003e4. Trending topics at the Catalysis\u0026ndash;AI intersection based on occurrences in scientific publication abstracts.\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eThe trend graphs were constructed using a Cartesian diagram in which the Y-axis represents the identified trending topics and the X-axis corresponds to the reference year (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e5\u003c/span\u003e). Each node in the graph denotes a concept, with its size proportional to the number of occurrences. The year assigned to each term reflects the median of its distribution of occurrences over the analyzed period. Concepts were grouped according to their thematic relationships and reference year.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eBetween 2001 and 2006, the terms \u003cem\u003eFCC unit, process variables, catalytic cracking\u003c/em\u003e, and \u003cem\u003epetroleum engineers\u003c/em\u003e were closely associated with the objectives of the energy and chemical process industries, particularly in relation to the optimization of conventional operations. In 2007, \u003cem\u003eoil reservoirs\u003c/em\u003e and \u003cem\u003emicrobial communities\u003c/em\u003e appeared sporadically, reflecting the gradual incorporation of biotechnological approaches into petroleum applications.\u003c/p\u003e\n\u003cp\u003eDuring 2009\u0026ndash;2017, the emergence and consolidation of computational tools became evident. Terms such as \u003cem\u003eprocess data, process modeling, cognitive science, genetic algorithms, fuzzy logic, computational intelligence, optimization algorithms, and computational models\u003c/em\u003e highlighted the growing use of advanced techniques to model, simulate, and optimize increasingly complex chemical processes. Several of these concepts have a long research history, underscoring the continuous evolution of computational approaches to industrial problem-solving.\u003c/p\u003e\n\u003cp\u003eAs shown in Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e, the period of greatest growth began in 2020, with \u003cem\u003ecomplex systems, computational models\u003c/em\u003e, and \u003cem\u003eoperating conditions\u003c/em\u003e emerging prominently, signaling the application of mathematical and algorithmic tools in process analysis, simulation, and prediction. In 2021, \u003cem\u003eprotein engineering, catalyst discovery\u003c/em\u003e, and catalytic \u003cem\u003ereaction\u003c/em\u003e became trending topics, indicating a shift in focus from process optimization toward biocatalysis and catalyst design. By 2022, \u003cem\u003eneural networks, outer membranes\u003c/em\u003e, and \u003cem\u003ecrystal structures\u003c/em\u003e gained prominence, with \u003cem\u003eneural networks\u003c/em\u003e ranking as the third most frequent term among the 36. Research increasingly relied on computational models enabling image analysis, property prediction, and structural characterization.\u003c/p\u003e\n\u003cp\u003eIn 2023, \u003cem\u003eartificial intelligence, machine learning\u003c/em\u003e, and \u003cem\u003edeep learning\u003c/em\u003e were consolidated as dominant topics, represented by the largest node sizes, reflecting their high frequency of occurrence. These concepts established themselves as the central technological axes for the analysis, optimization, and automation of complex processes. Their prevalence underscores a clear trend toward intensive digitization, the adoption of predictive algorithms, and the integration of data-driven approaches in catalysis research.\u003c/p\u003e\n\u003cp\u003eFinally, in 2024, \u003cem\u003esolar cells, energy materials\u003c/em\u003e, and \u003cem\u003ereaction conditions\u003c/em\u003e emerged as leading topics, signaling a growing emphasis on renewable energy and the development of more efficient materials.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003e5. Density map of terms from the abstracts of scientific publications in the Catalysis-AI set.\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eThe density map reveals, from left to right, three main co-occurrence subspaces (see Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e6\u003c/span\u003e): the first two correspond to the domain of artificial intelligence, while the third is associated directly with catalysis. In addition, emerging and dispersed clusters highlight research trends in renewable energies and advanced materials (e.g., solar cells, renewable energy, solar power, waste management), consistent with the patterns observed in Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e5\u003c/span\u003e for the year 2024. The persistent distance between the catalysis and AI clusters in the density map confirms that \u0026ldquo;Catalysis-AI\u0026rdquo; is indeed an emerging frontier rather than a fully consolidated field.\u003c/p\u003e\n\u003cp\u003eThe most concentrated area of terms is linked to artificial intelligence, followed by deep learning. The relatively small size of the clusters suggests that the knowledge generated so far has not yet reached a substantial volume. Moreover, the clear separation among the three subspaces indicates that the integration between AI-related topics and catalysis remains limited. As shown in Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e, convergence themes at the intersection of AI and catalysis are still in their early stages. The physical separation between these clusters reinforces the need for the \u0026ldquo;bridge\u0026rdquo; descriptors and multidisciplinary training proposed in earlier sections to foster a truly integrated Catalysis-AI domain.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003e6. Thematic evolution. A longitudinal thematic map analysis.\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eTo determine the conceptual structure of the Catalysis\u0026ndash;AI collection, a strategic thematic map (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e7\u003c/span\u003e) was constructed by analyzing bigrams extracted from the abstracts of each publication in the corpus. The map was organized into four quadrants according to Callon\u0026rsquo;s definition of density and centrality.\u003c/p\u003e\n\u003c/div\u003e\n\u003ch3\u003ePeriod 1991\u0026ndash;2019\u003c/h3\u003e\n\u003cp\u003eFigure \u003cspan class=\"InternalRef\"\u003e7\u003c/span\u003ea displays nine clusters formed over 29 years (68 documents; 11.00%) and their distribution within the Callon diagram. In the first quadrant, clusters related to \u003cem\u003eFuzzy logic, Genetic algorithm\u003c/em\u003e [\u003cspan class=\"CitationRef\"\u003e27\u003c/span\u003e], \u003cem\u003eArtificial intelligence\u003c/em\u003e [\u003cspan class=\"CitationRef\"\u003e28\u003c/span\u003e, \u003cspan class=\"CitationRef\"\u003e29\u003c/span\u003e] and \u003cem\u003eMachine learning\u003c/em\u003e [\u003cspan class=\"CitationRef\"\u003e30\u003c/span\u003e, \u003cspan class=\"CitationRef\"\u003e31\u003c/span\u003e] were identified, in increasing order of importance. These clusters exhibited high development and relevance, with their associated terms recognized as driving research topics.\u003c/p\u003e\n\u003cp\u003eIn contrast, the third quadrant contained clusters such as \u003cem\u003eCyber-physical systems, Fuzzy inference, Diagnostic devices, and Cognitive science\u003c/em\u003e, all characterized by low centrality and density, reflecting minimal development and importance. These themes disappeared in subsequent periods (2020\u0026ndash;2021 and 2022\u0026ndash;2024), suggesting a decline. \u003cem\u003eComputational Intelligence\u003c/em\u003e [\u003cspan class=\"CitationRef\"\u003e32\u003c/span\u003e] was located at the center of the diagram, with density and centrality values that prevent its clear assignment to any quadrant.\u003c/p\u003e\n\u003ch3\u003ePeriod 2020\u0026ndash;2021\u003c/h3\u003e\n\u003cp\u003eDespite being 14.5 times shorter, this period exhibited significant activity, producing 77 documents that account for 12.46% of the corpus. Nine thematic groups were identified, six of which are closely related to catalysis and three to artificial intelligence (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e7\u003c/span\u003eb).\u003c/p\u003e\n\u003cp\u003eIn the first quadrant, a prominent cluster characterized by high centrality and density includes the terms \u003cem\u003eLearning models\u003c/em\u003e [\u003cspan class=\"CitationRef\"\u003e33\u003c/span\u003e], \u003cem\u003eData-driven materials\u003c/em\u003e [\u003cspan class=\"CitationRef\"\u003e34\u003c/span\u003e], and \u003cem\u003eFunctional materials\u003c/em\u003e. These emerging topics reflect noteworthy development and importance.\u003c/p\u003e\n\u003cp\u003eThe second cluster, exhibiting intermediate centrality and density, has remained a driving theme since the previous period. It consists of the terms \u003cem\u003eChemical reactions\u003c/em\u003e [\u003cspan class=\"CitationRef\"\u003e35\u003c/span\u003e], E\u003cem\u003experimental data\u003c/em\u003e, and \u003cem\u003eHeterogeneous catalysis\u003c/em\u003e [\u003cspan class=\"CitationRef\"\u003e6\u003c/span\u003e]\u003c/p\u003e\n\u003cp\u003eA third cluster, demonstrating high relevance and substantial development, includes the terms \u003cem\u003eDomain knowledge\u003c/em\u003e [\u003cspan class=\"CitationRef\"\u003e36\u003c/span\u003e], \u003cem\u003eMachine intelligence\u003c/em\u003e, and \u003cem\u003eParadigm shift\u003c/em\u003e, among others. This cluster is identified as a key theme that emerged during this period.\u003c/p\u003e\n\u003cp\u003eIn the fourth quadrant, another cluster was found with high centrality but lower density compared to the previous three groups. This basic and cross-cutting theme is well developed and constitutes the most frequently occurring group among all clusters: A\u003cem\u003ertificial intelligenc\u003c/em\u003e [\u003cspan class=\"CitationRef\"\u003e37\u003c/span\u003e\u0026ndash;\u003cspan class=\"CitationRef\"\u003e41\u003c/span\u003e], it encompasses 32 terms, which can be divided into two subgroups: those related to catalysis as: P\u003cem\u003erotein engineering\u003c/em\u003e [\u003cspan class=\"CitationRef\"\u003e42\u003c/span\u003e, \u003cspan class=\"CitationRef\"\u003e43\u003c/span\u003e], A\u003cem\u003emino acids\u003c/em\u003e, C\u003cem\u003eatalysis research\u003c/em\u003e, and C\u003cem\u003eatalytic reactions\u003c/em\u003e; as well as related to AI, including \u003cem\u003eMachine learning\u003c/em\u003e [\u003cspan class=\"CitationRef\"\u003e44\u003c/span\u003e], \u003cem\u003eNeural networks\u003c/em\u003e, \u003cem\u003eDeep learning\u003c/em\u003e, \u003cem\u003eAB initio\u003c/em\u003e, \u003cem\u003eConvolutional neural networks\u003c/em\u003e, \u003cem\u003eDensity functionals\u003c/em\u003e, and \u003cem\u003eLearning techniques\u003c/em\u003e. The \u003cem\u003eArtificial Intelligence\u003c/em\u003e group arises from the combination of the \u003cem\u003eArtificial intelligence\u003c/em\u003e and \u003cem\u003eMachine learning\u003c/em\u003e groups from the previous period.\u003c/p\u003e\n\u003cp\u003eIn the second quadrant, a cluster with high density but low centrality was identified, indicating limited importance. This niche topic comprises the following terms: \u003cem\u003eAdsorption energy\u003c/em\u003e [\u003cspan class=\"CitationRef\"\u003e45\u003c/span\u003e], \u003cem\u003eCatalytic materials\u003c/em\u003e, \u003cem\u003eCatalytic performance\u003c/em\u003e, and \u003cem\u003eReactivity descriptors\u003c/em\u003e. Another cluster with slightly lower density and occurrence includes: \u003cem\u003eEnzyme engineering\u003c/em\u003e [\u003cspan class=\"CitationRef\"\u003e46\u003c/span\u003e], \u003cem\u003eComplex systems\u003c/em\u003e, \u003cem\u003eActive sites\u003c/em\u003e, \u003cem\u003eChemical reactions\u003c/em\u003e, and \u003cem\u003eFree energy\u003c/em\u003e.\u003c/p\u003e\n\u003cp\u003eFinally, in the third quadrant, we identified three groups with low centrality and density, indicating topics of minimal relevance and development. These subjects are considered either emerging or declining, including \u003cem\u003eHydrogen production\u003c/em\u003e, \u003cem\u003eCatalytic domain\u003c/em\u003e, and \u003cem\u003eOrganic frameworks\u003c/em\u003e.\u003c/p\u003e\n\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e\n\u003ch2\u003ePeriod 2022\u0026ndash;2024\u003c/h2\u003e\n\u003cp\u003eIn the last period, the number of publications increased by a factor of 8.5, reaching a total of 473 publications, which constitutes 76.54% of the overall corpus. Concurrently, the cluster structure demonstrated significant consolidation, decreasing from 9 to 3 clusters (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e7\u003c/span\u003ec).\u003c/p\u003e\n\u003cp\u003eIn the first quadrant, a cluster composed of 48-word pairs emerged with high density and centrality. The most recurrent pairs included \u003cem\u003eCatalytic activity\u003c/em\u003e, \u003cem\u003eActive site\u003c/em\u003e, and \u003cem\u003eProtein structure\u003c/em\u003e [\u003cspan class=\"CitationRef\"\u003e47\u003c/span\u003e\u0026ndash;\u003cspan class=\"CitationRef\"\u003e49\u003c/span\u003e]. These themes are well-developed and significant, identifying them as driving themes.\u003c/p\u003e\n\u003cp\u003eAt the center of the diagram, there exists a group with intermediate density and centrality, comprised of the terms \u003cem\u003eElectron microscopy\u003c/em\u003e, \u003cem\u003eLearning approaches\u003c/em\u003e, \u003cem\u003eIdentify potential\u003c/em\u003e, and \u003cem\u003eMolecular mechanisms\u003c/em\u003e [\u003cspan class=\"CitationRef\"\u003e50\u003c/span\u003e\u0026ndash;\u003cspan class=\"CitationRef\"\u003e52\u003c/span\u003e]. Due to its placement in the diagram, specific elements cannot be assigned to this group.\u003c/p\u003e\n\u003cp\u003eFinally, in the third quadrant, we find the largest cluster, characterized by low density and centrality, consisting of the terms \u003cem\u003eArtificial intelligence\u003c/em\u003e, \u003cem\u003eMachine learning\u003c/em\u003e, Neural network, \u003cem\u003eDeep learning\u003c/em\u003e, and \u003cem\u003eLearning models\u003c/em\u003e, among others [\u003cspan class=\"CitationRef\"\u003e21\u003c/span\u003e, \u003cspan class=\"CitationRef\"\u003e53\u003c/span\u003e\u0026ndash;\u003cspan class=\"CitationRef\"\u003e57\u003c/span\u003e]. This group is located in the quadrant of emerging or declining topics. However, its evolutionary trajectory reveals a distinct dynamic: it began the study period in the motor skills quadrant with an incipient volume of publications, transitioned to basic and cross-cutting themes, and ultimately shifted to its current category. Given the volume of publications recorded, one can infer that these topics are not in a phase of decline, but rather may be undergoing a process of internal reconfiguration.\u003c/p\u003e\n\u003cp\u003eThe following graph provides a dynamic and concise representation of how these research areas have evolved over recent decades, achieved by grouping the different terms.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec12\" class=\"Section2\"\u003e\n\u003ch2\u003eData Challenges: Standardization and Publication Bias\u003c/h2\u003e\n\u003cp\u003eThe emergence of a robust \u0026ldquo;Catalysis-AI\u0026rdquo; paradigm is critically contingent upon the quality and nature of available data assets. Currently, the discipline faces the significant challenge of a lack of standardization in experimental reporting, which hinders the interoperability of datasets required for training deep learning models. Even more critical is the phenomenon of publication bias, where the exclusive dissemination of experimental successes and optimized yields prevails. For artificial intelligence algorithms to achieve true predictive capability, it is imperative that the scientific community begins to document and publish so-called \u0026ldquo;negative results\u0026rdquo; or failed reactions. These data are essential, as they provide the necessary contrast for models to identify reactivity boundaries and prevent overfitting. Ultimately, an AI that learns only from successes is a tool blind to the inherent complexity of the catalytic design space.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec13\" class=\"Section2\"\u003e\n\u003ch2\u003ePerspectives and Future Outlook\u003c/h2\u003e\n\u003cp\u003eAlthough Ibero-America's contribution to the intersection of catalysis and artificial intelligence stands at a mere 3.72%, the region possesses a wealth of fundamental chemical knowledge that could act as a \u0026ldquo;catalyst\u0026rdquo; for its own digital transformation. To bridge this gap, it is imperative to transition from a purely experimental research model toward the creation of regional consortia focused on data science. Such alliances would foster the democratization of access to supercomputing infrastructure and open databases\u0026mdash;resources that are often prohibitively expensive for individual institutions. Furthermore, the early integration of chemometrics and machine learning into chemistry curricula is essential for cultivating a new generation of \u0026ldquo;bilingual\u0026rdquo; scientists proficient in both disciplines. Rather than competing in raw processing power with global leaders, the strategic focus should center on developing algorithms with low computational costs and applying AI to optimize processes of regional significance, such as biomass valorization, new catalytic materials and environmental catalysis.\u003c/p\u003e\n\u003cp\u003eExpanding the reach of AI-Catalysis in these regions will require the strategic measures previously discussed, ensuring that fundamental chemical knowledge is effectively paired with data-driven infrastructures.\u003c/p\u003e\n\u003c/div\u003e"},{"header":"Conclusions","content":"\u003cp\u003eDespite the promising capabilities of the new generation of programs based on artificial intelligence, machine learning, metadata management, and correlation techniques to optimize experimental timelines automatically and efficiently, several challenges persist in the implementation of AI in the field of catalysis. Heterogeneous systems exhibit a few of interconnected parameters that evolve throughout the reaction, complicating the prediction of materials suitable for optimization. Additionally, considerations regarding the cost, abundance, and availability of metals identified by algorithms, as well as the technological resources necessary for the successful scale-up of catalyst synthesis, present further obstacles.\u003c/p\u003e \u003cp\u003eHowever, bibliometric analyses indicate that a robust connection between the diverse areas of catalysis and emerging AI tools, has not yet been established. This gap is even more evident in Ibero-American countries, despite their strong tradition and achievements in catalysis. Such a disconnect may result in a substantial disparity between the pace of advances occurring in other scientific domains and those within catalysis. It is conceivable that the coming decades will herald the advent of a human-machine synergy that fosters scientific advancement, particularly in the design of novel catalytic materials.\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003cp\u003e| \u003cstrong\u003eAI\u003c/strong\u003e | Artificial Intelligence |\u003c/p\u003e\n\u003cp\u003e| \u003cstrong\u003eLBD\u003c/strong\u003e | Literature-Based Discovery |\u003c/p\u003e\n\u003cp\u003e| \u003cstrong\u003eCSV\u003c/strong\u003e | Comma-Separated Values |\u003c/p\u003e\n\u003cp\u003e| \u003cstrong\u003eDOI\u003c/strong\u003e | Digital Object Identifier |\u003c/p\u003e\n\u003cp\u003e| \u003cstrong\u003eCC\u003c/strong\u003e | Callon Centrality |\u003c/p\u003e\n\u003cp\u003e| \u003cstrong\u003eCD\u003c/strong\u003e | Callon Density |\u003c/p\u003e\n\u003cp\u003e| \u003cstrong\u003eXCON\u003c/strong\u003e | eXpert CONfigurer |\u003c/p\u003e\n\u003cp\u003e| \u003cstrong\u003eDEC\u003c/strong\u003e | Digital Equipment Corporation |\u003c/p\u003e\n\u003cp\u003e| \u003cstrong\u003eEMMS\u003c/strong\u003e | Energy Minimization Multiscale Model |\u003c/p\u003e\n\u003cp\u003e| \u003cstrong\u003ePINNs\u003c/strong\u003e | Physics-Informed Neural Networks |\u003c/p\u003e\n\u003cp\u003e|\u003cstrong\u003e\u0026nbsp;I4.0\u0026nbsp;\u003c/strong\u003e| Industry 4.0 |\u003c/p\u003e\n\u003cp\u003e| \u003cstrong\u003eCEO\u003c/strong\u003e | Chief Executive Officer |\u003c/p\u003e\n\u003cp\u003e| \u003cstrong\u003eIoT\u003c/strong\u003e | Internet of Things |\u003c/p\u003e\n\u003cp\u003e| \u003cstrong\u003eNLP\u003c/strong\u003e | Natural Language Processing |\u003c/p\u003e\n\u003cp\u003e| \u003cstrong\u003eLLMs\u003c/strong\u003e | Large-Scale Language Models |\u003c/p\u003e\n\u003cp\u003e| \u003cstrong\u003eFCC\u003c/strong\u003e | Fluid Catalytic Cracking |\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003eIf your study is a clinical trial, then please provide the necessary registration details (registry, trial registration number, and data of registration).\u003c/p\u003e\n\u003cp\u003eThe study is not a clinical trial: not applicable.\u003c/p\u003e\n\u003cp\u003eDeclarations\u003c/p\u003e\n\u003cp\u003eEthics and Consent to Participate\u003c/p\u003e\n\u003cp\u003eEthics and Consent to Participate: not applicable.\u003c/p\u003e\n\u003cp\u003eConsent for Publication\u003c/p\u003e\n\u003cp\u003eConsent for Publication: not applicable.\u003c/p\u003e\n\u003cp\u003eCompeting Interest\u003c/p\u003e\n\u003cp\u003eThere are no Competing Interests.\u003c/p\u003e\n\u003cp\u003eAuthor Contribution\u003c/p\u003e\n\u003cp\u003eThe conceptual framework and the original idea for this study were conceived by S.P.C. The research was subsequently expanded through the collaborative efforts of the entire group. All authors contributed to the design of the figures, manuscript drafting, and critical revision. S.V. conducted the bibliometric analysis.\u003c/p\u003e\n\u003cp\u003eFunding\u003c/p\u003e\n\u003cp\u003eFunding: not applicable-\u003c/p\u003e\n\u003cp\u003eAvailability of data and materials\u003c/p\u003e\n\u003cp\u003eThe link is not available. The bibliometric data supporting the findings of this research are maintained by the authors and can be provided for review purposes upon request.\u003c/p\u003e\n\u003cp\u003eAcknowledgments\u003c/p\u003e\n\u003cp\u003eWe extend our gratitude to the team at the Marcel Roche Regional Library of Science and Technology of the Venezuelan Institute of Scientific Research (IVIC) for their invaluable assistance in retrieving scientific articles pertinent to this study.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eLi Z, Wang S, Xin H. Toward artificial intelligence in catalysis. Nat Catal. 2018;1:641\u0026ndash;2. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s41929-018-0150-1\u003c/span\u003e\u003cspan address=\"10.1038/s41929-018-0150-1\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDonatien S. Challenges of Artificial Intelligence today and future implications for society and the world. World J Adv Res Rev. 2025;26:3045\u0026ndash;54. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.30574/wjarr.2025.26.1.1380\u003c/span\u003e\u003cspan address=\"10.30574/wjarr.2025.26.1.1380\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHouhou R, Bocklitz T. Trends in artificial intelligence, machine learning, and chemometrics applied to chemical data. Anal Sci Adv. 2021;2:128\u0026ndash;41. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1002/ansa.202000162\u003c/span\u003e\u003cspan address=\"10.1002/ansa.202000162\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGaetani M, Mazwi M, Balaci H, et al. Artificial intelligence in medicine and the pursuit of environmentally responsible science. Lancet Digit Heal. 2024;6:e438\u0026ndash;40. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/S2589-7500(24)00090-6\u003c/span\u003e\u003cspan address=\"10.1016/S2589-7500(24)00090-6\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGodbole NS, Lamb J. (2018) Research into Making Healthcare Green with Cloud, Green IT, and Data Science to Reduce Healthcare Costs and Combat Climate Change. In: 2018 9th IEEE Annual Ubiquitous Computing, Electronics and Mobile Communication Conference, UEMCON 2018. IEEE, pp 189\u0026ndash;195.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBokhimi X. Learning the Use of Artificial Intelligence in Heterogeneous Catalysis. Front Chem Eng. 2021;3:1\u0026ndash;8. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3389/fceng.2021.740270\u003c/span\u003e\u003cspan address=\"10.3389/fceng.2021.740270\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAmali Henadirage NG. Barriers to and Opportunities for the Adoption.pdf. Int J Artif Intell Educ. 2025;35:245\u0026ndash;81. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s40593-024-00439-5\u003c/span\u003e\u003cspan address=\"10.1007/s40593-024-00439-5\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eEldar Haber, Jemielniak D, Kurasiński A, Przegalińska A. (2025) Future Trends and Emerging Tools. In: Macmillan P, editor Using AI in Academic Writing and Research. Cham.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFu V. (2025) AI for Science: Opportunities, Challenges, and Future Directions. TechRxiv. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.36227/techrxiv.173949768.84003950/v1\u003c/span\u003e\u003cspan address=\"10.36227/techrxiv.173949768.84003950/v1\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCallon M, Courtial JP, Laville F. Co-word analysis as a tool for describing the network of interactions between basic and technological research: The case of polymer chemsitry. Scientometrics. 1991;22:155\u0026ndash;205. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/BF02019280\u003c/span\u003e\u003cspan address=\"10.1007/BF02019280\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTuring AM. Computing Machinery and Intelligence. Mind. 1950;59:433\u0026ndash;60. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1093/mind/xlvi.181.131\u003c/span\u003e\u003cspan address=\"10.1093/mind/xlvi.181.131\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCioffi R, Travaglioni M, Piscitelli G, et al. Artificial intelligence and machine learning applications in smart production: Progress, trends, and directions. Sustain. 2020;12. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/su12020492\u003c/span\u003e\u003cspan address=\"10.3390/su12020492\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHalal WE. Artificial intelligence is almost here. Horiz. 2003;11:37\u0026ndash;8. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1108/10748120310486771\u003c/span\u003e\u003cspan address=\"10.1108/10748120310486771\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGuo L, Wu J, Li J. Complexity at Mesoscales: A Common Challenge in Developing Artificial Intelligence. Engineering. 2019;5:924\u0026ndash;9. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.eng.2019.08.005\u003c/span\u003e\u003cspan address=\"10.1016/j.eng.2019.08.005\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLI Xiangyu CA of S. (2018) Mesoscience: Discovering the Unknowns Between the Knowns. EurekAlert 29\u0026ndash;31.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLi J. Exploring the Logic and Landscape of the Knowledge System: Multilevel Structures, Each Multiscaled with Complexity at the Mesoscale. Engineering. 2016;2:276\u0026ndash;85. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/J.ENG.2016.03.001\u003c/span\u003e\u003cspan address=\"10.1016/J.ENG.2016.03.001\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePeng L, Wei Z. Recent progress of mesoscience in design of electrocatalytic materials for hydrogen energy conversion. Particuology. 2020;48:19\u0026ndash;33. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.partic.2018.08.013\u003c/span\u003e\u003cspan address=\"10.1016/j.partic.2018.08.013\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWu N, Ji X, Li L, et al. Mesoscience in supported nano-metal catalysts based on molecular thermodynamic modeling: A mini review and perspective. Chem Eng Sci. 2021;229:116164. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.ces.2020.116164\u003c/span\u003e\u003cspan address=\"10.1016/j.ces.2020.116164\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHan J, Gong H, Ren X, Yan X. Supramolecular nanozymes based on peptide self-assembly for biomimetic catalysis. Nano Today. 2021;41:101295. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.nantod.2021.101295\u003c/span\u003e\u003cspan address=\"10.1016/j.nantod.2021.101295\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYang W, Fidelis TT, Sun WH. Machine Learning in Catalysis, from Proposal to Practicing. ACS Omega. 2020;5:83\u0026ndash;8. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1021/acsomega.9b03673\u003c/span\u003e\u003cspan address=\"10.1021/acsomega.9b03673\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMazheika A, Wang YG, Valero R, et al. Artificial-intelligence-driven discovery of catalyst genes with application to CO2 activation on semiconductor oxides. Nat Commun. 2022;13:1\u0026ndash;37. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s41467-022-28042-z\u003c/span\u003e\u003cspan address=\"10.1038/s41467-022-28042-z\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHirst JD, Boobier S, Coughlan J, et al. ML meets MLn: Machine learning in ligand promoted homogeneous catalysis. Artif Intell Chem. 2023;1:100006. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.aichem.2023.100006\u003c/span\u003e\u003cspan address=\"10.1016/j.aichem.2023.100006\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eUlissi ZW, Tang MT, Xiao J, et al. Machine-learning methods enable exhaustive searches for active Bimetallic facets and reveal active site motifs for CO\u003csub\u003e2\u003c/sub\u003e reduction. ACS Catal. 2017;7:6600\u0026ndash;8. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1021/acscatal.7b01648\u003c/span\u003e\u003cspan address=\"10.1021/acscatal.7b01648\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChen J, Wu P, Bu F, et al. 3D printing enhanced catalysis for energy conversion and environment treatment. DeCarbon. 2023;2:100019. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.decarb.2023.100019\u003c/span\u003e\u003cspan address=\"10.1016/j.decarb.2023.100019\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e(2022) Industria 4.0, la cuarta revoluci\u0026oacute;n industrial y la inteligencia operacional. In: Consult Inform\u0026aacute;tico. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.cic.es/industria-40-revolucion-industrial/#:~:text=\u003c/span\u003e\u003cspan address=\"https://www.cic.es/industria-40-revolucion-industrial/#:~:text=\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDaqun Q, Yongzai L. Fuzzy Temporal Knowledge Representation, Reasoning and Their Applications to Dynamic Systems. Acta Autom Sin. 1991;17:559\u0026ndash;65.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNandi S, Badhe Y, Lonari J, et al. Hybrid process modeling and optimization strategies integrating neural networks/support vector regression and genetic algorithms: Study of benzene isopropylation on Hbeta catalyst. Chem Eng J. 2004;97:115\u0026ndash;29. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/S1385-8947(03)00150-5\u003c/span\u003e\u003cspan address=\"10.1016/S1385-8947(03)00150-5\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAsadi S, Hassan M, Nadiri A, Dylla H. Artificial intelligence modeling to evaluate field performance of photocatalytic asphalt pavement for ambient air purification. Environ Sci Pollut Res. 2014;21:8847\u0026ndash;57. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s11356-014-2821-z\u003c/span\u003e\u003cspan address=\"10.1007/s11356-014-2821-z\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePathak L, Singh V, Niwas R, et al. Artificial intelligence versus statistical modeling and optimization of cholesterol oxidase production by using Streptomyces sp. PLoS ONE. 2015;10:1\u0026ndash;14. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1371/journal.pone.0137268\u003c/span\u003e\u003cspan address=\"10.1371/journal.pone.0137268\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSchlexer Lamoureux P, Winther KT, Garrido Torres JA, et al. Machine Learning for Computational Heterogeneous Catalysis. ChemCatChem. 2019;11:3581\u0026ndash;601. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1002/cctc.201900595\u003c/span\u003e\u003cspan address=\"10.1002/cctc.201900595\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBonk BM, Weis JW, Tidor B. Machine Learning Identifies Chemical Characteristics That Promote Enzyme Catalysis. J Am Chem Soc. 2019;141:4108\u0026ndash;18. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1021/jacs.8b13879\u003c/span\u003e\u003cspan address=\"10.1021/jacs.8b13879\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRemagnino P, Shapiro D. Artificial Intelligence Methods for Ambient Intelligence. Comput Intell. 2007;23:393\u0026ndash;4. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1111/j.1467-8640.2007.00312.x\u003c/span\u003e\u003cspan address=\"10.1111/j.1467-8640.2007.00312.x\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKartashov OO, Chernov AV, Polyanichenko DS, Butakova MA. XAS data preprocessing of nanocatalysts for machine learning applications. Mater (Basel). 2021;14:7884. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/ma14247884\u003c/span\u003e\u003cspan address=\"10.3390/ma14247884\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCole JM. A Design-to-Device Pipeline for Data-Driven Materials Discovery. Acc Chem Res. 2020;53:599\u0026ndash;610. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1021/acs.accounts.9b00470\u003c/span\u003e\u003cspan address=\"10.1021/acs.accounts.9b00470\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFoppa L, Ghiringhelli LM, Girgsdies F, et al. Materials genes of heterogeneous catalysis from clean experiments and artificial intelligence. MRS Bull. 2021;46:1016\u0026ndash;26. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1557/s43577-021-00165-6\u003c/span\u003e\u003cspan address=\"10.1557/s43577-021-00165-6\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVijayabaskar S. Harnessing Generative AI for Risk Management and Fraud Detection in Fintech: A New Era of Human-Machine Collaboration. Int J Sci Res Manag. 2020;8:369\u0026ndash;79. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.18535/ijsrm/v8i04.ec01\u003c/span\u003e\u003cspan address=\"10.18535/ijsrm/v8i04.ec01\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTrunschke A, Bellini G, Boniface M, et al. Towards Experimental Handbooks in Catalysis. Top Catal. 2020;63:1683\u0026ndash;99. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s11244-020-01380-2\u003c/span\u003e\u003cspan address=\"10.1007/s11244-020-01380-2\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYi D, Bayer T, Badenhorst CPS, et al. Recent trends in biocatalysis. Chem Soc Rev. 2021;50:8003\u0026ndash;49. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1039/d0cs01575j\u003c/span\u003e\u003cspan address=\"10.1039/d0cs01575j\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGe L, Yuan H, Min Y, et al. Predicted Optimal Bifunctional Electrocatalysts for the Hydrogen Evolution Reaction and the Oxygen Evolution Reaction Using Chalcogenide Heterostructures Based on Machine Learning Analysis of in Silico Quantum Mechanics Based High Throughput Screening. J Phys Chem Lett. 2020;11:869\u0026ndash;76. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1021/acs.jpclett.9b03875\u003c/span\u003e\u003cspan address=\"10.1021/acs.jpclett.9b03875\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLi P, Du Z, Chang C, et al. Efficient Catalytic Conversion of Waste Peanut Shells into Liquid Biofuel: An Artificial Intelligence Approach. Energy Fuels. 2020;34:1791\u0026ndash;801. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1021/acs.energyfuels.9b03433\u003c/span\u003e\u003cspan address=\"10.1021/acs.energyfuels.9b03433\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBoucheikhchoukh A, Thibault J, Fauteux-Lefebvre C. Catalyst design using artificial intelligence: SO2 to SO3 case study. Can J Chem Eng. 2020;98:2016\u0026ndash;31. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1002/cjce.23756\u003c/span\u003e\u003cspan address=\"10.1002/cjce.23756\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSiedhoff NE, Illig AM, Schwaneberg U, Davari MD. PyPEF\u0026mdash;An Integrated Framework for Data-Driven Protein Engineering. J Chem Inf Model. 2021;61:3463\u0026ndash;76. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1021/acs.jcim.1c00099\u003c/span\u003e\u003cspan address=\"10.1021/acs.jcim.1c00099\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eXiong W, Liu B, Shen Y, et al. Protein engineering design from directed evolution to de novo synthesis. Biochem Eng J. 2021;174:108096. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.bej.2021.108096\u003c/span\u003e\u003cspan address=\"10.1016/j.bej.2021.108096\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSiedhoff NE, Schwaneberg U, Davari MD. Machine learning-assisted enzyme engineering. Methods Enzymol. 2020;643:281\u0026ndash;315. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/bs.mie.2020.05.005\u003c/span\u003e\u003cspan address=\"10.1016/bs.mie.2020.05.005\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang B, Zhang F. Main Descriptors To Correlate Structures with the Performances of Electrocatalysts. Angew Int Ed Chemie. 2022;61:e202111026. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1002/anie.202111026\u003c/span\u003e\u003cspan address=\"10.1002/anie.202111026\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSingh N, Malik S, Gupta A, Srivastava KR. Revolutionizing enzyme engineering through artificial intelligence and machine learning. Emerg Top Life Sci. 2021;5:113\u0026ndash;25. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1042/ETLS20200257\u003c/span\u003e\u003cspan address=\"10.1042/ETLS20200257\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAulakh SS Jr., Epand JCB RM. Exploring the AlphaFold Predicted Conformational Properties of Human Diacylglycerol Kinases. J Phys Chem B. 2022;126:7172\u0026ndash;83. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/https://doi.org/10.1021/acs.jpcb.2c04533\u003c/span\u003e\u003cspan address=\"10.1021/acs.jpcb.2c04533\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSajjadi E, Frascarelli C, Venetis K, et al. Computational pathology to improve biomarker testing in breast cancer: how close are we ? Eur J Cancer Prev. 2023;32:460\u0026ndash;7. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1097/CEJ.0000000000000804\u003c/span\u003e\u003cspan address=\"10.1097/CEJ.0000000000000804\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCochereau B, Strat Y, Le, Ji Q, et al. Heterologous Expression and Biochemical Characterization of a New Chloroperoxidase Isolated from the Deep \u0026ndash; Sea Hydrothermal Vent Black Yeast Hortaea werneckii UBOCC \u0026ndash; A \u0026ndash; 208029. Mar Biotechnol. 2023;25:519\u0026ndash;36. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s10126-023-10222-7\u003c/span\u003e\u003cspan address=\"10.1007/s10126-023-10222-7\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMitchell S, Par\u0026eacute;s F, Faust Akl D, et al. Automated Image Analysis for Single-Atom Detection in Catalytic Materials by Transmission Electron Microscopy. J Am Chem Soc. 2022;144:8018\u0026ndash;29. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1021/jacs.1c12466\u003c/span\u003e\u003cspan address=\"10.1021/jacs.1c12466\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRen F, Wu F, Wu X, et al. Fungal systems for lignocellulose deconstruction: From enzymatic mechanisms to hydrolysis optimization. GCB Bioenergy. 2024;16:e13130. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1111/gcbb.13130\u003c/span\u003e\u003cspan address=\"10.1111/gcbb.13130\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eH\u0026ouml;ppner S, Schr\u0026ouml;der B, Fluhrer R. Structure and function of SPP/SPPL proteases: insights from biochemical evidence and predictive modeling. FEBS J. 2023;290:5456\u0026ndash;74. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1111/febs.16968\u003c/span\u003e\u003cspan address=\"10.1111/febs.16968\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMai H, Le TC, Chen D, et al. Machine Learning for Electrocatalyst and Photocatalyst Design and Discovery. Chem Rev Cite. 2022;122:13478\u0026ndash;515. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/https://doi.org/10.1021/acs.chemrev.2c00061\u003c/span\u003e\u003cspan address=\"10.1021/acs.chemrev.2c00061\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGoodarzi N, Ashrafi-Peyman Z, Khani E, Moshfegh AZ. Recent Progress on Semiconductor Heterogeneous Photocatalysts in Clean Energy Production and Environmental Remediation. Catalysts. 2023;13:1102. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/https://doi.org/10.3390/catal13071102\u003c/span\u003e\u003cspan address=\"10.3390/catal13071102\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKouba P, Kohout P, Haddadi F, et al. Machine Learning-Guided Protein Engineering. ACS Catal. 2023;13. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1021/acscatal.3c02743\u003c/span\u003e\u003cspan address=\"10.1021/acscatal.3c02743\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. :13863\u0026thinsp;\u0026ndash;\u0026thinsp;13895.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhang J, Fei Y, Sun L, Zhang QC. Advances and opportunities in RNA structure experimental determination and computational modeling. Nat Methods. 2022;19:19:1193\u0026ndash;207. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/https://doi.org/10.1038/s41592-022-01623-y\u003c/span\u003e\u003cspan address=\"10.1038/s41592-022-01623-y\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhang Z, Zheng Y, Qian L, et al. Emerging Trends in Sustainable CO2-Management Materials. Adv Mater. 2022;34:2201547. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/https://doi.org/10.1002/adma.202201547\u003c/span\u003e\u003cspan address=\"10.1002/adma.202201547\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"discover-chemistry","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"","sideBox":"Learn more about [Discover Chemistry](https://link.springer.com/journal/44371)","snPcode":"44371","submissionUrl":"https://submission.nature.com/new-submission/44371/3","title":"Discover Chemistry","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Discover Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Catalysis-AI, bibliometric analysis, artificial intelligence, evolution of scientific catalytic publications","lastPublishedDoi":"10.21203/rs.3.rs-9182089/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9182089/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eRecent advancements in data infrastructure, computational statistics, and artificial intelligence (AI) have inaugurated a transformative era for chemical sciences. These computational paradigms facilitate the optimization of complex systems at an unprecedented rate, transcending the limitations of traditional trial-and-error methodologies. This innovative convergence of domain-specific scientific knowledge and advanced heuristics is pivotal for the engineering of next-generation, sustainable chemical processes characterized by minimized energy footprints and enhanced selectivity. Significantly, this evolution fosters a deep integration between heterogeneous and homogeneous catalysis and the diverse analytical tools emerging from the digital frontier, potentially catalyzing a new \u0026ldquo;Catalysis\u0026ndash;AI\u0026rdquo; paradigm. However, comprehensive bibliometric analyses reveal a significant break; despite the proliferation of AI literature, its substantive implementation in experimental catalysis remains nascent. The current landscape is hindered by data silos and a lack of standardized descriptors. Consequently, the formulation of robust explanatory and predictive hypotheses\u0026mdash;anchored in both physical chemistry and data-driven insights\u0026mdash;is imperative. Establishing such a framework is essential for achieving a fundamental understanding of active sites and reaction mechanisms, ultimately addressing the urgent scientific and technological imperatives of modern industry.\u003c/p\u003e","manuscriptTitle":"The Emergence of Catalysis-ai: A New Frontier in Catalysis Research?","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-04-20 09:50:53","doi":"10.21203/rs.3.rs-9182089/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2026-05-18T22:06:53+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-05-01T14:23:06+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-04-28T16:39:15+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-04-19T13:58:15+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"202871445775266029283815985321095876736","date":"2026-04-18T21:04:32+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-04-12T01:33:50+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"84975007486898235069440667761885474779","date":"2026-04-10T18:57:34+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"274702659276712327040793546566679521094","date":"2026-04-10T16:38:59+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"166637383286228193010085607842510281980","date":"2026-04-10T14:21:18+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2026-04-10T13:36:11+00:00","index":"","fulltext":""},{"type":"editorInvited","content":"","date":"2026-04-02T13:01:06+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2026-03-29T02:26:45+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2026-03-29T02:26:01+00:00","index":"","fulltext":""},{"type":"submitted","content":"Discover Chemistry","date":"2026-03-20T21:39:48+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"discover-chemistry","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"","sideBox":"Learn more about [Discover Chemistry](https://link.springer.com/journal/44371)","snPcode":"44371","submissionUrl":"https://submission.nature.com/new-submission/44371/3","title":"Discover Chemistry","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Discover Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"94313585-9172-4335-9d65-aa876ab306bf","owner":[],"postedDate":"April 20th, 2026","published":true,"recentEditorialEvents":[{"type":"decision","content":"Revision requested","date":"2026-05-18T22:06:53+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-05-01T14:23:06+00:00","index":28,"fulltext":""}],"rejectedJournal":[],"revision":"","amendment":"","status":"in-revision","subjectAreas":[],"tags":[],"updatedAt":"2026-05-18T22:08:56+00:00","versionOfRecord":[],"versionCreatedAt":"2026-04-20 09:50:53","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9182089","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9182089","identity":"rs-9182089","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}