Mapping the Molecular Renaissance of Ayurveda: A Bibliometric and Scientometric Analysis of Genomics, Pharmacology and Precision Health Research, 1991–2024

1. Introduction

Ayurveda, one of the oldest and most comprehensive systems of medicine, emerged over 5,000 years ago in India [1]. It offers a distinctive understanding of health and disease, emphasising a harmonious balance between body (shareer), mind (mana) and spirit (atma) [2]. At the core of Ayurvedic philosophy lies the proposition that the body functions optimally in a state of equilibrium in which physical, mental and spiritual elements are balanced [3]. Ayurvedic principles emphasise the individualised nature of health and disease, holding that each person is born with a unique constitution, or Prakriti, which shapes susceptibility to disease, mental and physical traits, and responses to environmental and lifestyle factors.

Prakriti refers to an individual’s inherent constitution or biological makeup [4] and is held to be determined at conception by the specific combination of the three doshas, Vata, Pitta and Kapha [5]. These doshas derive from five elements and govern physiological and psychological function [6]. Ayurveda holds that Prakriti is established at conception and remains largely stable through life, although modifiable by diet (ahara), lifestyle (vihar), place (desh) and time (kala) [5]. This framework parallels the modern concept of personalised medicine, in which treatment is designed around an individual’s genetic, environmental and lifestyle profile rather than a uniform standard [6].

The principle of Tridosha forms the cornerstone of Ayurvedic physiology [7]. Vata, comprising space and air, governs movement, communication and elimination [8]; Pitta, comprising fire and water, governs metabolism, digestion and transformation; and Kapha, comprising earth and water, governs stability, growth and strength [9]. Health and disease are interpreted through doshic imbalance, with each dosha contributing to homeostasis [11]. The framework also accounts for why individuals of differing constitution respond differently to identical environmental or lifestyle exposures [12].

The Ayurvedic concept of Shoth (inflammation) is of particular contemporary relevance [11]. Shoth represents the body’s response to doshic imbalance [13], and is classified as Nij Shoth, arising from internal vitiation and associated with chronic conditions including arthritis, autoimmune disease and metabolic disorders, and Aagantuj Shoth, arising from external causes such as infection or trauma [11]. Modern immunology describes inflammation as a complex process involving immune cell activation and cytokine release [12], and the correspondence between these frameworks has motivated substantial mechanistic investigation.

Despite this rich foundation, integration of Ayurvedic concepts with modern molecular biology has been limited [14]. Advances in Ayurgenomics, applying genomic methods to the correlation between Prakriti and molecular phenotype, have begun to bridge this gap [12]. Studies by Prasher et al. and Govindaraj et al. have reported that individuals of differing Prakriti exhibit distinct gene expression profiles, particularly in genes associated with metabolism and inflammation, suggesting a molecular basis for classical constitutional classification [15].

Bibliometric and scientometric methods provide a quantitative means of characterising how a research field is structured and how it evolves. Beyond descriptive counts of publications and citations, formal bibliometric analysis permits testing of specific structural hypotheses: whether growth is exponential or saturating, whether author productivity conforms to Lotka’s inverse power law, whether sources scatter according to Bradford’s law, whether output concentration satisfies Price’s law, and where thematic clusters sit on the centrality–density plane that distinguishes motor from peripheral research themes. These methods have been applied extensively across biomedical disciplines but have not previously been applied with this level of formality to the intersection of Ayurveda and molecular science.

The present study therefore undertakes a comprehensive bibliometric and scientometric analysis of Ayurveda–genomics research indexed in Scopus between 1991 and 2024. The objectives are: (i) to characterise the temporal development of the field and to determine, by formal model selection, whether its growth is exponential or saturating; (ii) to test conformity to the classical bibliometric laws of Lotka, Bradford and Price; (iii) to quantify citation concentration and the distribution of scholarly impact; (iv) to characterise collaboration structure at author and country level, including the degree of internationalisation; (v) to reconstruct the conceptual structure of the field through network-based community detection and thematic mapping; (vi) to identify emerging and declining research themes through temporal keyword analysis; and (vii) to identify the historical intellectual foundations of the field through reference publication year spectroscopy.

2. Materials and Methods

2.1. Search Strategy and Corpus Construction

A structured literature search was conducted in Scopus. The Boolean query combined Ayurvedic terminology in the title, abstract or keyword fields (TITLE-ABS-KEY: "Ayurveda" OR "Prakriti" OR "Deha-Prakriti" OR "Tridosha" OR "Vata" OR "Pitta" OR "Kapha") with biomedical descriptors (TITLE-ABS-KEY: "inflammation" OR "genomics" OR "Ayurgenomics" OR "pharmacogenomics" OR "systems biology" OR "gene expression" OR "phenotype" OR "genotype" OR "biomarkers" OR "personalized medicine" OR "pharmacology" OR "herbal medicine" OR "phytopharmacology" OR "drug efficacy" OR "drug safety" OR "cytochrome P450"), constrained by PUBYEAR > 1990 AND PUBYEAR < 2025. Records were retrieved on 24 January 2026 and exported in full, comprising bibliographic metadata, abstracts, author and index keywords, affiliations, funding information and cited references.

Screening and selection followed the PRISMA framework (Figure 1). Identification returned 3,488 records; 27 duplicates were removed during screening; and the eligibility stage retained document type "Article" only, on the grounds that original research articles are citable, methodologically consistent primary contributions. The final corpus comprised 2,027 research articles published between 1991 and 2024, corresponding exactly to the PUBYEAR constraint of the search string. Records in languages other than English were retained where they carried English-language metadata sufficient for bibliometric processing; these represent 1.9% of the corpus (38 records, predominantly Chinese and German).

2.2. Network Construction and Visualisation

Co-citation, bibliographic coupling and co-authorship networks at the source, author, institution and country levels were constructed in VOSviewer version 1.6.20 [16] using full counting, with the thresholds reported in the corresponding figure legends. These networks underpin Figures 4 to 9.

2.3. Quantitative Bibliometric Analysis

All quantitative indicators reported in Section 3.4 onwards were computed directly from the exported record set in Python 3.12 using NumPy, pandas, SciPy, scikit-learn and NetworkX. The complete analysis script is provided as Supplementary File S1 and reproduces every table and figure from the exported dataset without modification.

Growth modelling. Cumulative publication output was fitted to exponential, three-parameter logistic and Gompertz models by non-linear least squares. Models were compared by the Akaike information criterion, with ΔAIC values reported relative to the best-supported model. The logistic carrying capacity K was taken as the modelled asymptote. Relative growth rate was computed as the difference in natural logarithm of cumulative output across each period divided by its duration, and doubling time as ln(2) divided by the relative growth rate. Developmental phases were delimited by binary segmentation changepoint detection on the annual output series using a variance-based cost function, rather than by visual inspection.

Bibliometric laws. Author productivity was tested against Lotka’s inverse power law by log–log regression of the frequency distribution, with the fitted exponent compared to the idealised value of 2 and goodness of fit assessed by the Kolmogorov–Smirnov statistic against the critical value 1.36/√N. Price’s law was evaluated by determining whether the √A most productive authors account for at least half of all authorships. Bradford’s law was assessed by partitioning ranked sources into three zones of approximately equal document yield and computing the Bradford multipliers between successive zones.

Citation structure. The h-index was computed as the largest h for which h documents each received at least h citations, the g-index as the largest g for which the top g documents jointly received at least g² citations, and the m-index as h divided by the number of years elapsed. Citation inequality was quantified by the Gini coefficient and visualised as a Lorenz curve. Distributional shape was characterised by skewness and excess kurtosis.

Collaboration. The degree of collaboration was computed as the proportion of multi-authored documents, the collaborative coefficient following Ajiferuke, and the collaboration index as mean authors per document. Country attribution was obtained by parsing the Scopus affiliation field against a controlled country vocabulary, achieving attribution for 98.9% of records. Documents were classified as single-country (SCP) or multi-country (MCP) publications according to whether more than one distinct country appeared among the affiliations.

Conceptual structure. An author keyword co-occurrence network was constructed over the 60 most frequent keywords after removal of the search terms themselves and generic indexing terms, retaining edges of weight ≥ 3. Communities were detected using the Louvain algorithm with modularity reported. For each community, Callon centrality was computed as the sum of external link weights and Callon density as the mean internal link weight scaled by community size; communities were assigned to quadrants of the thematic map relative to the median of each axis. Trend topics were characterised by the first quartile, median and third quartile of the publication years of documents carrying each keyword.

Reference publication year spectroscopy. Publication years were extracted from all cited references in the corpus by regular expression, yielding 18,862 dated references. The annual frequency distribution was computed and the deviation of each year from the five-year running median was plotted, following the standard RPYS procedure, in which positive deviations identify years disproportionately represented among cited references and therefore constitute the historical foundations of the field.

Open access analysis. Documents were classified by the presence of a Scopus open-access designation. Citation counts of open-access and non-open-access documents were compared by the Mann–Whitney U test with the rank-biserial correlation reported as effect size.

3. Results and Discussion

3.1. Corpus Characteristics

The final corpus comprised 2,027 research articles published between 1991 and 2024 across 762 distinct sources, authored by 7,541 unique individuals contributing 10,067 authorships. The corpus accumulated 51,735 citations at a mean of 25.52 and a median of 9 citations per document, with a field h-index of 97 and g-index of 168. Only 5.8% of documents were single-authored, and mean team size was 4.97 authors. Cited references numbered 19,494 in total, and 44.8% of documents carried an open-access designation. Full descriptive indicators are given in Table 1.

3.2. Temporal Distribution and Subject Composition

Figure 2 presents documents published per year disaggregated by constituent search keyword, and Figure 3 the distribution across Scopus subject areas. Medicine is the leading contributor at 24.3%, underscoring the clinical orientation of Ayurveda-based translational research and reflecting sustained effort to position constitutional concepts such as Prakriti within evidence-based frameworks. Biochemistry, Genetics and Molecular Biology follow at 20.3%, indicating the central role of molecular investigation in decoding the biological basis of Prakriti phenotypes through gene expression signatures, epigenetic modulation, cytokine profiling and cytochrome P450 variability. Agricultural and Biological Sciences account for 15.5%, reflecting the importance of plant-based therapeutics and phytochemical characterisation, while Pharmacology, Toxicology and Pharmaceutics contribute 14.0%, underscoring emphasis on drug metabolism, safety profiling and herb–drug interaction. Immunology and Microbiology (4.9%) reflect growing attention to immune modulation in constitution-based disease predisposition. Contributions from Multidisciplinary Sciences (2.5%), Computer Science (2.2%), Environmental Science (2.1%), Chemistry (1.8%) and Engineering (1.7%) demonstrate the expanding methodological scope of the field, the computational component in particular reflecting increasing use of bioinformatics, machine learning and network biology.

3.3. Networks of Literature, Institutions, Countries, Keywords, Authors and Journals

Co-citation analysis (Figure 4) reveals an intellectual architecture anchored by foundational complementary and alternative medicine scholarship, surrounded by a dense cluster of classical botanical compendia and Indian medicinal plant documentation, with translational bridges formed by integrative medicine researchers. Institutional analysis (Figure 5) identifies Banaras Hindu University as the most prolific contributor, followed by affiliated institutes and Patanjali Yog Peeth, with substantial representation from AIIMS, Savitribai Phule Pune University, Manipal Academy of Higher Education and the Ministry of AYUSH, the latter functioning as a bridging node between academic institutions and policy-driven initiatives. Country-level networks (Figure 6) show India as the central hub with dense connections to the United Kingdom, China, Malaysia, Saudi Arabia and several European nations. The author keyword network (Figure 7) places Ayurveda at the centre, surrounded by herbal medicine, inflammation, antioxidant activity and COVID-19. Author-level networks (Figure 8) identify a productive core including Balkrishna, Varshney, Patwardhan, Mukherjee and Prasher, with Patwardhan occupying a bridging position across clusters. Journal-level networks (Figure 9) place the Journal of Ayurveda and Integrative Medicine at the centre of the coupling network and the Journal of Ethnopharmacology at the centre of the co-citation network, with high-impact biomedical and clinical journals forming distinct peripheral clusters.

3.4. Growth Dynamics: The Field Is Consolidating, Not Expanding Exponentially

Annual output rose from 5 documents in 1991 to 170 in 2024 (Figure 10A). Binary segmentation changepoint detection identified three statistically supported breakpoints, at 2005, 2010 and 2021, delimiting four developmental phases (Table 4). Phase 1 (1991–2004) comprises 65 documents at 4.6 per year and 3.2% of the corpus, representing a foundational period of sparse, largely descriptive scholarship. Phase 2 (2005–2009) comprises 188 documents at 37.6 per year, marking the emergence of interdisciplinary engagement. Phase 3 (2010–2020) is the dominant expansion period, comprising 1,139 documents at 103.5 per year and 56.2% of the entire corpus. Phase 4 (2021–2024) comprises 635 documents at 158.8 per year, the highest absolute intensity of the series.

Formal model selection substantially revises the interpretation of this trajectory. Cumulative output was fitted to exponential, logistic and Gompertz models (Table 2; Figure 10B). The Gompertz model provides the best fit (R² = 0.998), the logistic model is materially worse (ΔAIC = 28.4), and the exponential model is decisively rejected (ΔAIC = 84.7). Both saturating models therefore outperform unbounded exponential growth by a wide margin. The fitted logistic carrying capacity is 2,549 documents, of which the observed corpus represents 79.5%, and the modelled inflexion point falls at 2018.6. On this evidence the field passed its maximum rate of expansion in the late 2010s and is now in a decelerating, consolidating regime.

Growth rate analysis corroborates this directly (Table 3; Figure 10C). The relative growth rate peaked at 0.252 yr⁻¹ during 2001–2010 and has since fallen to 0.130 yr⁻¹ in 2011–2020 and 0.090 yr⁻¹ in 2021–2024. Correspondingly, the doubling time has lengthened from 2.75 years to 5.33 and then 7.68 years. This is a monotonic and substantial deceleration. It is important to note that this does not indicate decline: absolute annual output remains at its highest recorded level, and the compound annual growth rate across the full series is 11.28%. Rather, it indicates that the field has matured past its phase of explosive expansion into one of sustained high-volume production, a pattern characteristic of research domains approaching institutional consolidation.

Table 2. Comparison of growth models fitted to cumulative publication output.

Model	Parameters	R²	AIC	ΔAIC
Gompertz	4473, 11.69, 0.08063	0.9982	228.3	0.0
Logistic	2549, 0.215, 27.57	0.9959	256.7	28.4
Exponential	38.76, 0.1222	0.9771	313.0	84.7

Models were fitted by non-linear least squares and compared by the Akaike information criterion. Logistic parameters are the carrying capacity K, the intrinsic rate r and the inflexion point x₀ expressed in years from 1991. ΔAIC is the difference from the best-supported model; differences exceeding 10 are conventionally regarded as decisive.

Table 2. Comparison of growth models fitted to cumulative publication output.

Model	Parameters	R²	AIC	ΔAIC
Gompertz	4473, 11.69, 0.08063	0.9982	228.3	0.0
Logistic	2549, 0.215, 27.57	0.9959	256.7	28.4
Exponential	38.76, 0.1222	0.9771	313.0	84.7

Models were fitted by non-linear least squares and compared by the Akaike information criterion. Logistic parameters are the carrying capacity K, the intrinsic rate r and the inflexion point x₀ expressed in years from 1991. ΔAIC is the difference from the best-supported model; differences exceeding 10 are conventionally regarded as decisive.

Table 3. Relative growth rate and doubling time by period.

Period	Documents	Share of corpus (%)	Relative growth rate (yr⁻¹)	Doubling time (years)
1991 –2000	28	1.4	0.1914	3.62
2001–2010	309	15.2	0.2516	2.75
2011–2020	1055	52.0	0.13	5.33
2021–2024	635	31.3	0.0903	7.68

Relative growth rate is the difference in natural logarithm of cumulative output across the period divided by its duration; doubling time is ln(2) divided by the relative growth rate.

Table 3. Relative growth rate and doubling time by period.

Period	Documents	Share of corpus (%)	Relative growth rate (yr⁻¹)	Doubling time (years)
1991 –2000	28	1.4	0.1914	3.62
2001–2010	309	15.2	0.2516	2.75
2011–2020	1055	52.0	0.13	5.33
2021–2024	635	31.3	0.0903	7.68

Relative growth rate is the difference in natural logarithm of cumulative output across the period divided by its duration; doubling time is ln(2) divided by the relative growth rate.

Table 4. Developmental phases identified by binary segmentation changepoint detection.

Phase	Years	Documents	Mean per year	Share of corpus (%)
Phase 1	1991–2004	65	4.6	3.2
Phase 2	2005–2009	188	37.6	9.3
Phase 3	2010–2020	1139	103.5	56.2
Phase 4	2021–2024	635	158.8	31.3

Changepoints were detected on the annual output series using a variance-based cost function, without reference to prior expectations regarding phase boundaries.

Table 4. Developmental phases identified by binary segmentation changepoint detection.

Phase	Years	Documents	Mean per year	Share of corpus (%)
Phase 1	1991–2004	65	4.6	3.2
Phase 2	2005–2009	188	37.6	9.3
Phase 3	2010–2020	1139	103.5	56.2
Phase 4	2021–2024	635	158.8	31.3

Changepoints were detected on the annual output series using a variance-based cost function, without reference to prior expectations regarding phase boundaries.

Figure 10. Quantitative growth analysis of the corpus. (A) Annual output with the four developmental phases identified by binary segmentation; dashed lines mark detected changepoints at 2005, 2010 and 2021. (B) Comparison of exponential, logistic and Gompertz models fitted to cumulative output, with the logistic carrying capacity indicated. (C) Relative growth rate and doubling time by period. (D) International collaboration rate and mean authors per document across the four phases. Model selection used the Akaike information criterion; the Gompertz model is preferred (ΔAIC = 0) over the logistic (ΔAIC = 28.4) and exponential (ΔAIC = 84.7) alternatives. Numerical parameters are given in Table 2, Table 3, Table 4 and Table 9.

Figure 10. Quantitative growth analysis of the corpus. (A) Annual output with the four developmental phases identified by binary segmentation; dashed lines mark detected changepoints at 2005, 2010 and 2021. (B) Comparison of exponential, logistic and Gompertz models fitted to cumulative output, with the logistic carrying capacity indicated. (C) Relative growth rate and doubling time by period. (D) International collaboration rate and mean authors per document across the four phases. Model selection used the Akaike information criterion; the Gompertz model is preferred (ΔAIC = 0) over the logistic (ΔAIC = 28.4) and exponential (ΔAIC = 84.7) alternatives. Numerical parameters are given in Table 2, Table 3, Table 4 and Table 9.

3.5. Conformity to the Classical Bibliometric Laws

Lotka’s law. The distribution of author productivity was fitted by log–log regression, yielding an exponent of n = 2.615 (standard error 0.184, R² = 0.914) against the idealised inverse-square value of 2. The Kolmogorov–Smirnov statistic D = 0.064 exceeds the critical value of 0.016 at the 5% level, indicating statistically significant deviation from the idealised distribution (Figure 11A). The direction of this deviation is informative: an exponent above 2 denotes a steeper decline than Lotka predicted, meaning the field contains proportionally more transient contributors and proportionally fewer sustained ones than a canonical scientific discipline. Consistent with this, 6,304 of 7,541 authors (83.6%) contributed exactly one article.

Price’s law. Price’s law posits that the square root of the author population produces half of all output. Here the 86 most prolific authors, corresponding to √7,541, account for 842 of 10,067 authorships, or 8.4%. Price’s law is therefore not merely unmet but missed by a very wide margin. Taken together with the Lotka result, this establishes that Ayurveda–genomics research is not structured around a small elite of dominant producers. It is instead a highly dispersed field with an unusually large periphery of occasional contributors. This finding materially qualifies the common characterisation of the field as driven by a core group of leading scholars: while such a core exists and is visible in the co-authorship networks of Figure 8, it accounts for a small minority of total output.

Bradford’s law. Source scattering conforms well to Bradford’s law (Table 5; Figure 11B). Partitioning the 762 sources into three zones of approximately equal document yield gives a core of 23 journals publishing 33.3% of all articles, a second zone of 150 journals and a third of 589 journals. The Bradford multipliers between successive zones are 6.52 and 3.93. The core is led by the Journal of Ayurveda and Integrative Medicine (136 documents) and the Journal of Ethnopharmacology (102 documents), the latter carrying markedly higher citation impact (mean 65.0 versus 11.7 citations, h-index 36 versus 21). A pronounced impact heterogeneity exists within the core: several high-output journals, including the International Journal of Research in Ayurveda and Pharmacy and the Journal of Natural Remedies, exhibit mean citation counts below 2, indicating that publication volume and scholarly influence are only loosely coupled in this field (Table 6).

3.6. Citation Structure and Impact Concentration

Citation distribution is extremely unequal (Table 7; Figure 11C). The Gini coefficient is 0.728, comparable to the most unequal national income distributions. The top 1% of documents capture 24.0% of all citations, the top 5% capture 44.8%, the top 10% capture 59.0%, and the top 25% capture 80.1%. Meanwhile 12.4% of documents remain entirely uncited. The distribution is severely right-skewed (skewness 14.36, excess kurtosis 279.22), and the mean of 25.52 citations exceeds the median of 9 by a factor of 2.8, confirming that mean citation counts are a poor summary statistic for this field and that median-based reporting is preferable.

The most cited document is Goel et al. (2008) in Biochemical Pharmacology with 2,057 citations, followed by Grover et al. (2002) in the Journal of Ethnopharmacology with 1,485 and Ahmad et al. (2013) with 1,134 (Table 12). Normalising by document age identifies Goel et al. as also the highest-intensity paper at 121.0 citations per year, indicating that its influence is not merely an artefact of seniority.

3.7. Geographical Structure and Internationalisation

Country attribution succeeded for 98.9% of records. India dominates output by a very wide margin (Table 8; Figure 12A). Beyond raw counts, the single- and multi-country publication decomposition reveals a structural feature not visible in the network visualisations of Figure 6: India exhibits one of the lowest international collaboration rates among leading contributors, whereas smaller contributors such as Saudi Arabia, the United Kingdom, Germany and Australia derive a much larger proportion of their output from cross-national teams (Figure 12B). The field is therefore geographically concentrated and, at its centre, comparatively inward-facing.

The relationship between productivity and impact is inverse at the country level (Figure 12C). Countries with lower output frequently exhibit higher mean citation counts, a pattern consistent with selective international participation in higher-visibility work. Longitudinal analysis (Table 9; Figure 10D) nevertheless shows steady internationalisation: the proportion of multi-country documents rose across the four phases, and mean team size rose in parallel from 3.99 to 5.71 authors per document. Both trends indicate a field becoming progressively more collaborative even as its geographical centre of gravity remains stable.

The collaboration indices confirm a strongly collaborative field overall: the degree of collaboration is 0.942, the Ajiferuke collaborative coefficient 0.681 and the collaboration index 4.97 authors per document.

Table 8. Fifteen most productive countries with collaboration and impact indicators.

Country	Documents	SCP	MCP	MCP ratio (%)	Total citations	Mean citations	h-index
India	1429	1242	187	13.1	30148	21.1	75
United States	215	102	113	52.6	12981	60.4	55
China	125	87	38	30.4	3738	29.9	30
United Kingdom	69	14	55	79.7	2327	33.7	26
Germany	49	22	27	55.1	1506	30.7	21
Malaysia	39	12	27	69.2	1765	45.3	14
Australia	39	16	23	59.0	1479	37.9	20
Italy	35	14	21	60.0	1334	38.1	20
Saudi Arabia	32	3	29	90.6	1903	59.5	14
Japan	32	16	16	50.0	1150	35.9	14
Bangladesh	24	10	14	58.3	509	21.2	12
South Korea	23	7	16	69.6	703	30.6	15
Russia	18	7	11	61.1	519	28.8	10
Sri Lanka	17	14	3	17.6	594	34.9	8
Thailand	16	5	11	68.8	395	24.7	9

SCP denotes single-country publications and MCP multi-country publications. The MCP ratio is the proportion of a country’s output involving international co-authorship. Country attribution was derived by parsing Scopus affiliation strings.

Table 8. Fifteen most productive countries with collaboration and impact indicators.

Country	Documents	SCP	MCP	MCP ratio (%)	Total citations	Mean citations	h-index
India	1429	1242	187	13.1	30148	21.1	75
United States	215	102	113	52.6	12981	60.4	55
China	125	87	38	30.4	3738	29.9	30
United Kingdom	69	14	55	79.7	2327	33.7	26
Germany	49	22	27	55.1	1506	30.7	21
Malaysia	39	12	27	69.2	1765	45.3	14
Australia	39	16	23	59.0	1479	37.9	20
Italy	35	14	21	60.0	1334	38.1	20
Saudi Arabia	32	3	29	90.6	1903	59.5	14
Japan	32	16	16	50.0	1150	35.9	14
Bangladesh	24	10	14	58.3	509	21.2	12
South Korea	23	7	16	69.6	703	30.6	15
Russia	18	7	11	61.1	519	28.8	10
Sri Lanka	17	14	3	17.6	594	34.9	8
Thailand	16	5	11	68.8	395	24.7	9

SCP denotes single-country publications and MCP multi-country publications. The MCP ratio is the proportion of a country’s output involving international co-authorship. Country attribution was derived by parsing Scopus affiliation strings.

Table 9. Collaboration indicators across the four developmental phases.

Period	Documents	International collaboration (%)	Mean authors per document	Mean countries per document
1991 –2004	65	10.8	3.77	1.12
2005–2009	182	15.4	4.07	1.22
2010–2020	1119	15.6	4.76	1.22
2021–2024	630	17.6	5.84	1.27

Table 9. Collaboration indicators across the four developmental phases.

Period	Documents	International collaboration (%)	Mean authors per document	Mean countries per document
1991 –2004	65	10.8	3.77	1.12
2005–2009	182	15.4	4.07	1.22
2010–2020	1119	15.6	4.76	1.22
2021–2024	630	17.6	5.84	1.27

3.8. Conceptual Structure and Thematic Mapping

Louvain community detection on the author keyword co-occurrence network resolved eight communities with a modularity of Q = 0.394 (Figure 13A). A modularity in this range indicates genuine but moderate thematic partitioning: the field is organised into recognisable subdomains, yet these remain substantially interconnected rather than fragmented into isolated specialties. The communities correspond to inflammation and oxidative stress; antioxidant and anti-inflammatory pharmacology; herbal and traditional medicine; analytical standardisation; COVID-19 and computational pharmacology; complementary and integrative medicine; phytochemistry; and cancer biology.

Positioning these communities on the Callon centrality–density plane (Table 11; Figure 13B) identifies three motor themes, characterised by both high external connectivity and high internal development. The first comprises antioxidant, anti-inflammatory, analgesic and antimicrobial activity, the pharmacological core of the field. The second comprises COVID-19, Withania somnifera, molecular docking and network pharmacology, a computationally oriented cluster of recent origin that has achieved unusually high internal density (6.29) despite its youth. The third comprises complementary, alternative and integrative medicine, reflecting the field’s continuing engagement with the broader integrative healthcare discourse.

Inflammation and oxidative stress, together with herbal and traditional medicine, occupy the basic and transversal quadrant: they are highly central to the field but comparatively less internally developed, which is the expected position for foundational themes that supply shared vocabulary across many subdomains. Phytochemistry constitutes a niche theme, internally coherent but weakly connected to the wider network. Analytical standardisation and cancer biology fall in the emerging or declining quadrant, with low centrality and low density.

3.9. Temporal Evolution of Research Themes

Trend analysis based on the median publication year of documents carrying each keyword identifies a clear generational sequence (Table 10; Figure 14A). The newest themes are molecular docking and network pharmacology, both with median year 2023, followed by COVID-19 (2022) and SARS-CoV-2 (2022). The appearance of molecular docking and network pharmacology at the leading edge is notable: it indicates that the computational turn in Ayurveda research is not merely an adjunct to experimental work but currently the fastest-growing methodological front. Intermediate themes include ashwagandha (2021), Prakriti (2020), traditional medicine (2020) and phytochemical analysis (2020). The oldest themes are antioxidants (2012), Ayurvedic medicine (2012), arthritis (2013), Siddha (2014), antimicrobial activity (2014) and analgesic activity (2014).

This sequence provides quantitative support for the qualitative claim that the field has shifted from descriptive characterisation toward mechanistic and computational investigation. It also indicates that Prakriti, the concept most central to Ayurgenomics, entered the mainstream literature comparatively recently, with a median year of 2020 and an interquartile range of 2014 to 2021.

Table 10. Trend topics ranked by median publication year.

Keyword	Occurrences	Q1 year	Median year	Q3 year
molecular docking	28	2022	2023	2024
network pharmacology	25	2021	2023	2024
covid-19	39	2021	2022	2023
sars-cov-2	18	2022	2022	2022
ashwagandha	16	2018	2021	2023
traditional medicine	40	2015	2020	2022
prakriti	16	2014	2020	2021
obesity	15	2013	2020	2022
phytochemical	15	2013	2020	2021
natural products	12	2015	2020	2021
ethnopharmacology	12	2014	2020	2021
inflammation	63	2014	2019	2021
oxidative stress	38	2014	2019	2022
hptlc	33	2012	2019	2023
apoptosis	14	2014	2019	2020
magnaporthe oryzae	13	2015	2019	2021
flavonoids	13	2013	2019	2020
rice blast	12	2016	2019	2022
rheumatoid arthritis	33	2011	2018	2022
standardization	23	2013	2018	2020

Q1 and Q3 denote the first and third quartiles of the publication years of documents carrying each keyword. Only keywords occurring in at least eight documents are shown.

Table 10. Trend topics ranked by median publication year.

Keyword	Occurrences	Q1 year	Median year	Q3 year
molecular docking	28	2022	2023	2024
network pharmacology	25	2021	2023	2024
covid-19	39	2021	2022	2023
sars-cov-2	18	2022	2022	2022
ashwagandha	16	2018	2021	2023
traditional medicine	40	2015	2020	2022
prakriti	16	2014	2020	2021
obesity	15	2013	2020	2022
phytochemical	15	2013	2020	2021
natural products	12	2015	2020	2021
ethnopharmacology	12	2014	2020	2021
inflammation	63	2014	2019	2021
oxidative stress	38	2014	2019	2022
hptlc	33	2012	2019	2023
apoptosis	14	2014	2019	2020
magnaporthe oryzae	13	2015	2019	2021
flavonoids	13	2013	2019	2020
rice blast	12	2016	2019	2022
rheumatoid arthritis	33	2011	2018	2022
standardization	23	2013	2018	2020

Q1 and Q3 denote the first and third quartiles of the publication years of documents carrying each keyword. Only keywords occurring in at least eight documents are shown.

Table 11. Thematic clusters with Callon centrality and density.

Cluster	Representative keywords	Keywords	Total occurrences	Callon centrality	Callon density	Quadrant
1	inflammation; oxidative stress; rheumatoid arthritis; antioxidants	6	179	3	3.0	Basic/Transversal
2	antioxidant; anti-inflammatory; analgesic; antimicrobial	4	118	9	4.25	Motor
3	herbal medicine; traditional medicine; ayurvedic medicine; ethnopharmacology	5	131	7	3.2	Basic/Transversal
4	hptlc; standardization	2	56	0	1.5	Emerging/Declining
5	covid-19; withania somnifera; molecular docking; network pharmacology	7	175	3	6.29	Motor
6	complementary and alternative medicine; alternative medicine; complementary medicine; integrative medicine	5	109	4	5.2	Motor
7	phytochemistry; pharmacology	2	44	0	4.0	Niche
8	apoptosis; breast cancer	2	26	0	1.5	Emerging/Declining

Communities were detected by the Louvain algorithm (modularity Q = 0.394). Callon centrality measures external connectivity and Callon density internal development. Quadrant assignment is relative to the median of each axis.

Table 11. Thematic clusters with Callon centrality and density.

Cluster	Representative keywords	Keywords	Total occurrences	Callon centrality	Callon density	Quadrant
1	inflammation; oxidative stress; rheumatoid arthritis; antioxidants	6	179	3	3.0	Basic/Transversal
2	antioxidant; anti-inflammatory; analgesic; antimicrobial	4	118	9	4.25	Motor
3	herbal medicine; traditional medicine; ayurvedic medicine; ethnopharmacology	5	131	7	3.2	Basic/Transversal
4	hptlc; standardization	2	56	0	1.5	Emerging/Declining
5	covid-19; withania somnifera; molecular docking; network pharmacology	7	175	3	6.29	Motor
6	complementary and alternative medicine; alternative medicine; complementary medicine; integrative medicine	5	109	4	5.2	Motor
7	phytochemistry; pharmacology	2	44	0	4.0	Niche
8	apoptosis; breast cancer	2	26	0	1.5	Emerging/Declining

Communities were detected by the Louvain algorithm (modularity Q = 0.394). Callon centrality measures external connectivity and Callon density internal development. Quadrant assignment is relative to the median of each axis.

3.10. Historical Foundations: Reference Publication Year Spectroscopy

Reference publication year spectroscopy across 18,862 dated cited references identifies the historical strata on which the field rests (Figure 14B). Beyond the expected recent maxima reflecting ordinary citation recency, the deviation spectrum reveals distinct positive peaks at 1935, 1956, 1969, 1971, 1976 and 1984. These peaks are substantial relative to the low absolute reference volume of those years, indicating specific foundational works that continue to be cited disproportionately. Their distribution across the mid-twentieth century corresponds to the period of systematic pharmacognostic documentation of Indian medicinal plants and the compilation of the classical botanical compendia that recur prominently in the co-citation network of Figure 4. Among more recent years, 2000 and 2005 show elevated deviations consistent with the foundational Ayurgenomics and integrative medicine literature.

This analysis makes explicit a structural feature that citation counts alone obscure. The field draws simultaneously on two temporally distinct intellectual bases: a mid-twentieth-century pharmacognostic and botanical literature, and a twenty-first-century molecular and integrative medicine literature. The relative sparsity of peaks between these strata indicates limited continuous intellectual transmission between them, and suggests that the modern field has reconnected with its documentary foundations rather than developing from them without interruption.

3.11. Open Access and the Absence of a Citation Advantage

Of the corpus, 908 documents (44.8%) carry an open-access designation. Comparison of citation counts between open-access and non-open-access documents found no statistically significant difference (Mann–Whitney U = 505,553; p = 0.850), with a negligible rank-biserial effect size of 0.005. Median citations were 8 for open-access and 9 for non-open-access documents. The frequently asserted open-access citation advantage is therefore not detectable in this field. This null result is reported explicitly because it is informative: it suggests that in a domain where readership is concentrated within a specialist community with established access pathways, open-access status confers little additional visibility. It also cautions against advocating open-access publication in this field on citation-maximisation grounds.

3.12. Author-Level Structure

The most productive authors are Kumar, A. (33 documents), Balkrishna, A. (31) and Varshney, A. (30), the latter two publishing exclusively since 2019 and reflecting concentrated recent institutional output (Table 13). Productivity and influence are, however, only weakly coupled. Patwardhan, B. contributes 21 documents but attracts 1,267 citations at a mean of 60.3 and an h-index of 17, the highest among the leading authors, while Mukherjee, P.K. attains a mean of 48.8 citations across 16 documents. By contrast, several of the most prolific contributors exhibit mean citation counts near 15. This divergence between volume and influence mirrors the pattern observed at journal level in Section 3.5 and is consistent with the low concentration indicated by the failure of Price’s law.

Table 12. Fifteen most cited documents in the corpus.

First author	Year	Source	Citations	Citations per year
Goel , A. et al.	2008	Biochemical Pharmacology	2057	121.0
Grover, J.K. et al.	2002	Journal of Ethnopharmacology	1485	64.6
Ahmad, A. et al.	2013	Asian Pacific Journal of Tropical Biomedicine	1134	94.5
Chainani-Wu, N. et al.	2003	Journal of Alternative and Complementary Medicine	999	45.4
Jia, Y. et al.	2000	EMBO Journal	916	36.6
Aparna, V. et al.	2012	Chemical Biology and Drug Design	707	54.4
Dugasani, S. et al.	2010	Journal of Ethnopharmacology	641	42.7
Ji, H.-F. et al.	2009	EMBO Reports	564	35.2
Muthu, C. et al.	2006	Journal of Ethnobiology and Ethnomedicine	524	27.6
Kronenberg, F. et al.	2002	Annals of Internal Medicine	511	22.2
Saper, R.B. et al.	2008	JAMA	434	25.5
Naik, G.H. et al.	2003	Phytochemistry	376	17.1
Dharmasiri, M.G. et al.	2003	Journal of Ethnopharmacology	303	13.8
Bhattacharya, S.K. et al.	2000	Phytomedicine	300	12.0
Bag, A. et al.	2013	Asian Pacific Journal of Tropical Biomedicine	292	24.3

Citations per year normalises total citations by the number of years elapsed since publication, and identifies sustained rather than merely accumulated influence.

Table 12. Fifteen most cited documents in the corpus.

First author	Year	Source	Citations	Citations per year
Goel , A. et al.	2008	Biochemical Pharmacology	2057	121.0
Grover, J.K. et al.	2002	Journal of Ethnopharmacology	1485	64.6
Ahmad, A. et al.	2013	Asian Pacific Journal of Tropical Biomedicine	1134	94.5
Chainani-Wu, N. et al.	2003	Journal of Alternative and Complementary Medicine	999	45.4
Jia, Y. et al.	2000	EMBO Journal	916	36.6
Aparna, V. et al.	2012	Chemical Biology and Drug Design	707	54.4
Dugasani, S. et al.	2010	Journal of Ethnopharmacology	641	42.7
Ji, H.-F. et al.	2009	EMBO Reports	564	35.2
Muthu, C. et al.	2006	Journal of Ethnobiology and Ethnomedicine	524	27.6
Kronenberg, F. et al.	2002	Annals of Internal Medicine	511	22.2
Saper, R.B. et al.	2008	JAMA	434	25.5
Naik, G.H. et al.	2003	Phytochemistry	376	17.1
Dharmasiri, M.G. et al.	2003	Journal of Ethnopharmacology	303	13.8
Bhattacharya, S.K. et al.	2000	Phytomedicine	300	12.0
Bag, A. et al.	2013	Asian Pacific Journal of Tropical Biomedicine	292	24.3

Citations per year normalises total citations by the number of years elapsed since publication, and identifies sustained rather than merely accumulated influence.

Table 13. Fifteen most productive authors with impact indicators.

Author	Documents	Total citations	Mean citations	h-index	First publication	Latest publication
Kumar , A.	33	483	14.6	12	1995	2024
Balkrishna, A.	31	470	15.2	13	2019	2024
Varshney, A.	30	466	15.5	13	2019	2024
Kumar, S.	21	256	13.5	11	2008	2024
Patwardhan, B.	21	1267	60.3	17	2000	2022
Singh, S.	17	248	14.6	6	2000	2024
Singh, R.	16	200	13.3	10	2012	2024
Singh, A.	16	268	16.8	9	1991	2024
Mukherjee, P.K.	16	781	48.8	14	2006	2024
Sharma, V.	16	491	30.7	10	2011	2024
Kumar, V.	15	455	30.3	10	2002	2024
Sharma, S.	15	160	10.7	6	2011	2024
Kumar, D.	14	149	10.6	6	2010	2024
Verma, S.	13	183	15.2	9	2007	2023
Tillu, G.	13	418	32.2	10	2012	2022

The h-index is computed within the corpus only and is therefore not the author’s overall h-index. Author names are as indexed in Scopus; homonym disambiguation was not performed, and common surnames may aggregate multiple individuals.

Table 13. Fifteen most productive authors with impact indicators.

Author	Documents	Total citations	Mean citations	h-index	First publication	Latest publication
Kumar , A.	33	483	14.6	12	1995	2024
Balkrishna, A.	31	470	15.2	13	2019	2024
Varshney, A.	30	466	15.5	13	2019	2024
Kumar, S.	21	256	13.5	11	2008	2024
Patwardhan, B.	21	1267	60.3	17	2000	2022
Singh, S.	17	248	14.6	6	2000	2024
Singh, R.	16	200	13.3	10	2012	2024
Singh, A.	16	268	16.8	9	1991	2024
Mukherjee, P.K.	16	781	48.8	14	2006	2024
Sharma, V.	16	491	30.7	10	2011	2024
Kumar, V.	15	455	30.3	10	2002	2024
Sharma, S.	15	160	10.7	6	2011	2024
Kumar, D.	14	149	10.6	6	2010	2024
Verma, S.	13	183	15.2	9	2007	2023
Tillu, G.	13	418	32.2	10	2012	2022

The h-index is computed within the corpus only and is therefore not the author’s overall h-index. Author names are as indexed in Scopus; homonym disambiguation was not performed, and common surnames may aggregate multiple individuals.

4. Limitations

Several limitations should be borne in mind. The analysis relies exclusively on Scopus; coverage differences between Scopus, Web of Science, PubMed and Dimensions are well documented, and a substantial body of Ayurveda research published in regional and non-indexed journals falls outside this corpus. Restriction to document type "Article" excludes reviews, conference papers and book chapters, which in this field carry appreciable intellectual weight. Author names were used as indexed by Scopus without homonym disambiguation; given the prevalence of common surnames in the dominant national contributor, the author-level indicators in Table 13 may aggregate distinct individuals, and this affects the Lotka and Price analyses to an unknown but probably modest degree. Country attribution was derived by parsing affiliation strings and, although successful for 98.9% of records, remains subject to parsing error. Citation counts are cumulative to the export date and disadvantage recent publications; age-normalised measures were reported alongside raw counts to mitigate this. The growth models are extrapolations beyond the observed range and the carrying capacity estimate in particular should be treated as indicative rather than predictive. Finally, keyword-based conceptual mapping depends on author-supplied terms, available for 86.4% of records, and inherits any inconsistency in indexing practice.

5. Conclusion

This study provides a quantitative characterisation of the intellectual, thematic, institutional and geographical structure of Ayurveda–genomics research across 2,027 articles published between 1991 and 2024. Four principal findings emerge. First, the field is no longer in exponential expansion: formal model selection decisively favours saturating growth, the relative growth rate has fallen by nearly two-thirds since its peak, doubling time has lengthened from 2.75 to 7.68 years, and the corpus stands at approximately 79.5% of its modelled carrying capacity. The field is consolidating rather than accelerating. Second, output is far less concentrated among elite authors than is typical: Price’s law fails by a wide margin and Lotka’s exponent of 2.615 indicates an unusually large periphery of single-contribution authors, so the field is better described as broadly dispersed than as elite-driven. Third, citation impact is nevertheless extremely concentrated, with a Gini coefficient of 0.728 and a quarter of all citations accruing to 1% of documents, and volume and influence are only loosely coupled at both journal and author level. Fourth, thematic mapping identifies computational pharmacology, comprising molecular docking and network pharmacology, as simultaneously the newest and one of the most internally developed themes in the field, marking a decisive methodological turn.

Reference publication year spectroscopy further reveals that the field draws on two temporally separated intellectual foundations, a mid-twentieth-century pharmacognostic literature and a twenty-first-century molecular one, with limited continuous transmission between them. Together with the finding that the dominant national contributor exhibits among the lowest international collaboration rates, and that no open-access citation advantage is detectable, these results identify the specific structural constraints that will shape the field’s next phase. Strengthening cross-national collaboration, improving the coupling between publication volume and scholarly influence, and consolidating the emerging computational front represent the most tractable routes to continued maturation of Ayurveda within integrative and precision medicine.