Data of the single botanical drugs were collected from the Part I of API, including 9 volumes as shown in Table 1. For each of the herbal monographs, information on the herbal name, biological origin(s), medicinal plant-part(s), therapeutic use(s) were extracted. The scientific names of the biological origins were then checked against World Flora Online (WFO) using R package “U.Taxonstand” (Zhang and Qian, 2022) with R 4.2.0 (R Core Team, 2022). Since the therapeutic uses of the herbal drugs in volume one to five were only available in Latinized Sanskrit, these terms were translated as follows: 1) to build a Sanskrit-English dictionary with the therapeutic use data in volume six to nine; 2) to translate the therapeutic uses into English with the above dictionary; 3) for those terms which were not included in the above dictionary, search in Wisdom Library (https://www.wisdomlib.org/) manually and cross-checked with the standardization of non-clinical terminologies of ayurveda (National Institute of Ayurveda, 2011). As a result, data including the drug names, biological origins, medicinal plant-parts, as well as the therapeutic uses of all the single botanical drugs in API were compiled (Supplementary Materials S1, S2).

While understanding traditional practice clearly and essential aim of any ethnopharmacological study, there now is a long tradition of using an etic categorization in order to allow an understanding of biomedical uses and the future potential of these species (Staub et al., 2015). In anthropological terms “etic” approaches (borrowed from the term phonetic in linguistics) investigate elements of a culture from an outsider’s perspective focusing on cross-cultural differences and similarities. Since it provides a global standard, it also enables a comparative analysis (Yao et al., 2021). Additionally, we suggest that the categorization should also consider the traditional concept and the pragmatic category in modern hospital. Finally, these therapeutic uses were sorted into 20 categories, including AND (andrology and male reproductive system), ANT (antidote and problems due to poisoning), CAR (cardiovascular disorders, blood-related diseases), DER (dermatologic disorders, skin problems), EAR (ear and hearing problems), EYE (eye and vision disorders), GAS (gastrointestinal and digestive disorders), GYN (gynecology and female reproductive system), GNR (general complaints including cold, fever, inflammation, headache, etc.), IMM (immune disorders and rheumatism), MET (metabolic and endocrine disorders), NER (nervous system, memory and sleeping problems), ORA (oral and dental problems), PAR (parasitic disease), PED (pediatrics disorders), PSY (psychiatric disorders and mental illness), RES (respiratory system disorders), SKE (skeleto-muscular system disorders), TRA (traditional uses that were not related to modern indications), URO (urinary disorders).

To generate a phylogenetic tree of all species, R package “V.PhyloMaker2” (Jin and Qian, 2022) was used with R 4.2.0 (R Core Team, 2022). And then, with formatted tables (i.e., “sample” and “traits”) generated from the data obtained above, this phylogenetic tree was loaded into Phylocom 4.2 (Webb et al., 2008). According to the manual, the following indices for each therapeutic use category were calculated: Faith’s phylogenetic diversity (PD), diversity weighted by interspecific phylogenetic distances (Dp), mean pairwise trait distance among taxa (MPD), net relatedness index (NRI), mean distance to nearest neighbor trait distance (MNTD) and nearest taxon index (NTI). Additionally, “hot nodes,” which mean the nodes with significantly more taxa than random probability for specific therapeutic uses were detected as marked with “SIGMORE” (Webb et al., 2008). Data visualization was performed using iTOL (Letunic and Bork, 2021).

To evaluate the therapeutic similarity of drugs sourced from the same species, the pairwise Jaccard Index (JI) was calculated based on their categorized therapeutic uses data, using the R package “sets.” And then, those JI of the same species were selected manually (as highlighted in Supplementary Material S3), thereafter, a distribution histogram of these JI values were plotted with interval of 0.1 was plotted using Microsoft Excel 365. A Cullen and Frey graph was then mapped to test the distribution pattern using R package “fitdistrplus” (Delignette-Muller & Dutang, 2015) with R 4.2.0 (R Core Team, 2022).

Part I of API consists of 645 monographs for single drugs or extracts, and almost all of these are single botanical drugs except the following 27: 1) Volume VII records 21 drugs sourced exclusively from minerals or metals; 2) volume VI includes three non-botanical drugs (i.e., butter, water and honey; 3) volume VI includes two drugs which are the mixture of plant extracts from two families thus they are not “single”; 4) volume III includes one drug which is the fungal part of a lichen. Besides, different parts of three botanical drugs were with distinguishable therapeutic uses, so these different parts were treated as individual botanical drugs. As a result, 621 single botanical drugs were identified in API in total (A list of these 621 drugs is provided in Supplementary Material S1).

The 621 single botanical drugs in API are sourced from 393 species in 323 genera and 115 families based on APG IV—the current Angiosperm Phylogeny Group’s classification (Figure 1). With 47 species, Fabaceae is the largest source family of these botanical drugs, which is followed by Apiaceae (16), Poaceae (15), Euphorbiaceae (13), Lamiaceae (13), Asteraceae (13), Apocynaceae (13), Cucurbitaceae (11), Malvaceae (10), Zingiberaceae (10), Solanaceae (9), and Moraceae (8). The top 12 families together contribute 178 species which constitute 45.3% of the total source species. Other families such as Acanthaceae, Arecaceae, Anacardiaceae, Rubiaceae, Amaranthaceae, Araceae, Combretaceae, Convolvulaceae and Piperaceae contribute less species to the Ayurvedic medical flora.

As expected, some large families with distinct species like the Fabaceae, Euphorbiaceae, Lamiaceae and Asteraceae are commonly included, but some with a less distinct characteristics like the Poaceae and some smaller but very characteristic families like the Zingiberaceae are also present. It is worth noting that as many as 75 families contribute only one or two species, and these families have increased the phylogenetic diversity of the Ayurvedic medical flora largely.

In terms of the medicinal part, the 621 single botanical drugs (Figure 2) include125 roots, (20%) making it the most commonly used medicinal part. Fruit rank second (85 drugs; 14%). These are followed by whole plant (13%), stem (12%), leaf (11%), bark (10%), and seed (10%).

Taking the traditional concepts, biomedical uses and the pragmatic disease classification into account, therapeutic uses of the botanical drugs were sorted into 20 categories (Figure 3 ). There are 293 species of 95 families used in GAS (75% of all species and 82% of all families). This is followed by that of GNR (255 species, 89 families), DER (250 species, 91 families), MET (227 species, 81 families), CAR (201 species, 81 families), RES (181 species, 68 families), PAR (134 species, 58 families), and URO (125 species, 61 families). Interestingly, the above-mentioned therapeutic uses, which are linked to more species or families than the rest, are with a higher average species per family, ranging from 3.1 to 2.0; conversely, for those uses linking to less species and families, the average species per family are below 2.0.

Different parts of 96 plant species are used as two or more drugs. For example, the root and the seed of Abrus precatorius L. are used as two drugs, separately; the ripe and unripe fruit of Diospyros malabarica (Desr.) Kostel are recorded as one drug but with different therapeutic uses, so we treat them as two single drugs. Volumes VIII and IX of API include 15 botanical drugs and their processed products (i.e., powder, hydro-alcoholic extract, and water extract) respectively, and we treat these powders and extracts as the same drug with their original drug. In total, 238 single botanical drugs of API are sourced from these 96 species. The pairwise Jaccard Index (JI) was calculated to evaluate the similarity between every two drugs (Supplementary Material S3, note the ones sourced from the same species are highlighted). Figure 4 shows the distribution of pairwise JI values of the drugs from the same species.

Accordingly, the average value of all the 215 JI values is 0.54617, suggesting that drugs of the same species differ in their therapeutic uses. However, these values fit for beta distribution instead of Normal distribution. There are 30 JI values at the region (0.9–1.0], indicating that many of these drugs from the same species are still used highly similarly. For example, leaf, stem, aerial part and flower of Dendrophthoe falcata (L.f.) Ettingsh are recorded as four drugs separately, but they have very similar therapeutic uses; however, the therapeutic uses of its fruit are distinguishable. In other cases, the low JI values indicate the different therapeutic uses of drugs from the same species. For example, the seed and whole plant of Datura metel L. only have one common therapeutic use for PAR, and their JI value is 0.1. As is shown in Figure 4A, only four of these drugs are with distinct therapeutic uses, viz. the seed and whole plant of D. metel L., and the root and the stem of Saccharum officinarum L. As a result, the therapeutic uses of the drugs sourced from the same species may differ to different extent.

Indices for phylogenetic diversity and structure of the source species of the single botanical drugs in API (Table 2) indicate the Ayurvedic medicinal plants for different therapeutic uses distribute evenly in the perspective of phylogeny. Faith’s phylogenetic diversity (PD) correlates with the numbers of species (Nsp) positively, fitting the formula PD = 44.836*Nsp +2,197.3 (R ² = 0.9833) and this linear relationship hints that the phylogenetic distances between the species are even. This evenness is also supported by the similar Dp values which range from 110 to 142. Similarly, values of mean pairwise trait distance among taxa (MPD) also fluctuate slightly, while values of the mean distance to nearest neighbor trait distance (MNTD) are with relative larger fluctuation. Net Relatedness Index (NRI) and Nearest Taxon Index (NTI) are indices for indicating the significant clustering. Specifically, values below −1.96 indicate significant “over-dispersal” while values above 1.96 indicate significant “clustering.” As are shown in Table 2, there is no significant “over-dispersal” or “clustering” except for those of GAS and MET, which are clustering significantly.

The algorithm “nodesig” in phylocom allows for finding the taxa of “hot nodes,” which were hypothesized to have higher potential for specific therapeutics. Accordingly, 172 species were found out in “hot nodes,” contributing 276 “hot nodes” hits for 20 of the therapeutic categories. Table 3 shows the number of these species by family and their hot nodes hits of each therapeutic category.

The results show many useful clues for the relationship between families and therapeutic uses. Eight species of Moraceae contributing the most hot nodes hits (32) to CAR, DER, GYN and MET, suggesting Moraceae’s potential functions in the above therapeutic uses; 13 species of Apocynaceae are found to be hot nodes of DER, GNR and ORA; almost all the hot nodes species of Poaceae are found to linked with MET and URO; 18 species of Fabaceae are with high potential for the therapeutic use of AND, which 2 of them are also for EYE. Several species are found to be related to more than one therapeutic use. For example, Artocarpus heterophyllus Lam. (Moraceae) is “hot” for CAR, DER, GYN and MET, Coleus forskohlii (Willd.) Briq. (Lamiaceae) for ANT, PAR, PSY and URO, Ocimum tenuiflorum L. (Lamiaceae) for ANT, PAR, PSY and URO, highlighting their multiple functions. With regards to the therapeutic categories, GNR and MET have the most hot nodes species (45 and 42 separately), this might because of the frequent occurrence rate of related diseases such as fever. Hot nodes species of GNR are commonly found in the Apocynaceae (13), Asteraceae (13), Apiaceae (11) and Piperaceae (5), while those for MET are from the Moraceae (8), Poaceae (15), Cucurbitaceae (11), Cyperaceae (2) and Typhaceae (2). Additionally, there are 13 families that have only one hot nodes species for one therapeutic category. It is worth noting that no hot taxon was found for GAS or PED, which can be accounted for based on different reasons: (a) There are 293 species (accounting for 74.5% of all the 393 species) with the therapeutic use for GAS, while the algorithm “nodesig” does not detect any “hot nodes,” hinting that no taxonomic preference for these species; (b) differently, only 12 species are used for PED, which is not sufficient for finding out the phylogenetic distribution pattern.