# Source paper archive - Source title: Multi-Criteria Evaluation of Hibiscus mutabilis Cultivars for Food Development Using Delphi and Analytic Hierarchy Process - Author(s): Yaoyuwei Fan, Yuqing Ma, Xiaoqing Shi, Yongjie Feng, Yiqiong Liu, Hong Wang, Xi Wang - Source landing URL: https://www.biorxiv.org/content/10.64898/2026.11.05.764492v1 - PDF URL: https://www.biorxiv.org/content/10.64898/2026.11.05.764492v1.full.pdf - PDF SHA-256: ed01637a6782afaa50705609a8d767a2fd595f906c22d9d05dc1fef5e9aeaeb9 - Converter: pymupdf4llm 0.0.27 - Limits: Automated PDF text extraction; reading order, equations, tables and formatting may be incomplete or incorrect. Images are omitted, and no OCR was performed. Markup and links are neutralized for safe downloading. Consult the original PDF for the authoritative version. --- [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.05.764492; this version posted November 6, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.11.05.764492) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) # 1 Multi-Criteria Evaluation of Hibiscus mutabilis Cultivars for 2 # **Food Development Using Delphi and Analytic Hierarchy** 3 Process 4 Yaoyuwei Fan [1], Yuqing Ma [1], Xiaoqing Shi [1], Yongjie Feng [1], Yiqiong Liu [1], Hong Wang [1], Xi 5 Wang [1*] 6 1 Ex Situ Conservation and Resource Utilization of Mountain Plants Key Laboratory of Sichuan 7 Province, Chengdu Botanical Garden / Chengdu Park City Plant Science Research Institute, 8 Chengdu 610036, China 9 - Corresponding author: Xi Wang (wangxi_cdbg@qq.com) # 10 Abstract 11 This study evaluated 11 _Hibiscus mutabilis_ cultivars from Chengdu Botanical Garden to 12 assess their quality for food use and identify promising cultivars. An evaluation framework was 13 developed using the Delphi method to select indicators and the analytic hierarchy process (AHP) 14 to determine their weights. Overall scores ranged from 2.5785 to 3.6043, and the cultivars were 15 classified into three grades. Grade I included _Hibiscus mutabilis_ 'Bairi Huacai', _Hibiscus_ 16 _mutabilis_ 'Chongbanbai' and _Hibiscus mutabilis_ 'Jinqiusong'; Grade II included _Hibiscus_ 17 _mutabilis_ 'Jinqiuhong', _Hibiscus mutabilis_ 'Zuifurong', _Hibiscus mutabilis_ 'Danbanhong', 18 _Hibiscus mutabilis_ 'Jinxiuzi', _Hibiscus mutabilis_ 'Jinbiyu' and _Hibiscus mutabilis_ 'Caixia'; 19 whereas Grade III included _Hibiscus mutabilis_ 'Mudanfen' and _Hibiscus mutabilis_ 'Jinrui'. In 20 summary, _Hibiscus mutabilis_ 'Bairi Huacai', _Hibiscus mutabilis_ 'Chongbanbai' and _Hibiscus_ 21 _mutabilis_ 'Jinqiusong' are the preferred cultivars for food processing and development. The 1 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.05.764492; this version posted November 6, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.11.05.764492) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 22 framework provides a basis for cultivar screening and further evaluation of _H. mutabilis_ flowers 23 for food use. 24 **Key words:** _Hibiscus mutabilis_ ; Food Development Potential; Analytic Hierarchy Process; 25 Delphi Method # 26 Introduction 27 _Hibiscus mutabilis_ L. is a deciduous shrub or small tree belonging to the genus _Hibiscus_ in 28 the family Malvaceae[1]. Widely cultivated as an ornamental plant in China, _H._ _mutabilis_ is 29 valued for its diverse flower colours and forms, long flowering period, and adaptability to 30 different growing conditions[2-4] . Recent efforts in germplasm collection and cultivar 31 development have expanded the resources available for the diversified use of _H._ _mutabilis_ [5-7]. 32 Cultivars differ in flower colour[8], morphology[9], and regional adaptability[10], providing a 33 germplasm foundation for their diversified utilisation . In addition to ornamental uses, the flowers 34 of _Hibiscus_ syriacus and its closely related species, as distinctive plant resources, also have a 35 history of traditional culinary use[11-13] . _Hibiscus_ flowers are used in tea and various culinary 36 preparations[14]. Growing interest in plant-based food ingredients has prompted research into 37 the nutritional and bioactive components of _H._ _mutabilis_ flowers, including amino acids, mineral 38 elements, and flavonoids[15-17] . 39 Currently, relevant research has primarily focused on the ornamental characteristics[17], 40 germplasm resources[5], cultivation adaptability[18] and medicinal value of _Hibiscus_ [16]. 41 However, comparative assessments of cultivars for food use remain limited, and a 42 comprehensive evaluation framework is still needed[19]. Therefore, this study utilised _Hibiscus_ 43 flowers from 11 cultivars as the test material, employing the Delphi method to screen and select 44 evaluation indicators[20-22], and combined this with the Analytic Hierarchy Process (AHP)[23] 2 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.05.764492; this version posted November 6, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.11.05.764492) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 45 to construct a comprehensive evaluation system for the edible quality of _H._ _mutabilis_ . This 46 system was used to assign comprehensive scores, rank and classify the different cultivars, with 47 the aim of providing a reference for the development of superior _Hibiscus_ cultivars and their 48 subsequent utilisation in food processing. # 49 Materials and Methods ## 50 Plant Materials 51 Eleven _H. mutabilis_ L. cultivars were collected from the research nursery of Chengdu 52 Botanical Garden: _Hibiscus mutabilis_ 'Jinxiuzi', _Hibiscus mutabilis_ 'Jinqiusong', _Hibiscus_ 53 _mutabilis_ 'Chongbanbai', _Hibiscus mutabilis_ 'Caixia', _Hibiscus mutabilis_ 'Jinqiuhong', _Hibiscus_ 54 _mutabilis_ 'Zuifurong', _Hibiscus mutabilis_ 'Jinbiyu', _Hibiscus mutabilis_ 'Bairi Huacai', _Hibiscus_ 55 _mutabilis_ 'Jinrui', _Hibiscus mutabilis_ 'Danbanhong' and _Hibiscus mutabilis_ 'Mudanfen' (Fig 56 1)(cultivars are referred to by cultivar name only in subsequent sections). Among these, _Hibiscus_ 57 _mutabilis_ were developed by Chengdu Botanical Garden through its own breeding program. 58 Fully opened flowers, intact and free of mechanical damage or visible pests and diseases, were 59 sampled in mid-October 2025. For each cultivar, three plants were sampled per cultivar. Thirteen 60 flowers from each plant were pooled to form one biological replicate, yielding three biological 61 replicates per cultivar. 62 **Fig 1. 11 cultivars of** _**Hibiscus mutabilis**_ ## 63 Screening of Evaluation Indicators ### 64 Construction of the initial evaluation index system 65 Based on the requirements for the edible quality of _H. mutabilis_ flowers and relevant 66 literature, and taking into account food safety, nutrition and active substances, sensory quality, 3 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.05.764492; this version posted November 6, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.11.05.764492) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 67 and processing adaptability, an initial evaluation index system for the edible quality of _H._ 68 _mutabilis_ flowers was constructed. The initial indicators covered four aspects—safety, nutrition 69 and active substances, sensory quality, and storage and adaptability—comprising 19 indicators in 70 total (Table 1). 71 **Table 1. Initial Evaluation Index System** **Criterion layer** **Specific indicator** 1. Heavy metal contamination risk 2. Microbiological contamination risk Safety 3. Allergenicity 4. gossypol-related toxicity Nutrition and active substances Sensory quality 5. Exogenous pollutant barrier capacity 6. Total flavonoid content 7. Soluble sugar content 8. Total amino acid content 9. Vitamin C content 10. Total mineral element content 11. Fiber coarseness 12. Lingering astringency 13. Herbal aroma recognizability 14. Mouthfeel smoothness 15. Fading rate 16. Appearance and color 17. Storability Storage and adaptability 18. Gel-forming capacity 19. Color stability ### 72 Delphi expert consultation 73 This study employed the Delphi method to screen and optimise the initial evaluation 74 indicators. An expert consultation questionnaire was designed based on the initially constructed 75 19 indicators, and experts in horticulture, plant resources, food safety, food processing, and _H._ 76 _mutabilis_ -related research were invited for evaluation. The questionnaire mainly covered three 77 aspects: (i) A 10-point scale was used to assess the suitability of each indicator for evaluating the 78 edible quality of _Hibiscus_ flowers; (ii) Seek experts’ views on the addition, deletion, merging or 79 amendment of indicators; (iii) Assess the relative importance of the four criteria levels: safety, 4 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.05.764492; this version posted November 6, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.11.05.764492) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 80 nutritional and bioactive compounds, sensory quality, and shelf life and adaptability—using a 81 five-level scale: unimportant, less important, moderately important, important, and very 82 important. 83 In the first round, a total of 32 expert consultation questionnaires were distributed and 32 84 were returned, of which 28 were valid, representing a valid response rate of 87.5%. The 85 participating experts were primarily from the fields of horticulture, botany, food safety, food 86 processing and related research areas, providing good professional representation. Based on the 87 experts’ scores, the mean, standard deviation and coefficient of variation for the importance 88 scores of each indicator were calculated, and the indicators were adjusted in light of the experts’ 89 written comments. The selection of indicators was primarily based on their importance, the 90 degree of consensus among experts, their measurability and their practical application value. 91 Indicators that were difficult to quantify, involved high testing costs or had a weak correlation 92 with the evaluation of edible quality were either deleted or merged. 93 Based on the results of the first round of expert consultation, adjustments have been made to 94 the initial evaluation indicator system. In the safety layer, ‘sensitisation’ and ‘ability to block 95 exogenous contaminants’ were removed from the safety indicators; in the nutritional and active 96 substance layer, ‘soluble sugar content’ and ‘vitamin C content’ were removed from the nutrition 97 and active ingredients indicators; in the sensory quality layer, ‘herbal aroma recognisability’, 98 ‘colour fading rate’ and ‘appearance and colour’ were removed from the sensory quality 99 indicators; and in the shelf-life and adaptability layer, ‘gel-forming ability’ and ‘colour stability’ 100 were removed from the shelf-life and adaptability indicators. To better assess the suitability of _H._ 101 _mutabilis_ flowers as raw materials for food development, yield per unit area and drying yield 102 were incorporated into the revised evaluation framework. After the first round of revisions, the 5 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.05.764492; this version posted November 6, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.11.05.764492) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 103 number of evaluation indicators was reduced from 19 to 12, and the revised indicator system was 104 established for the second round of expert consultation.To enhance the alignment of the 105 evaluation system with the objectives of food development and utilisation, the original criterion 106 ‘storage stability and adaptability’ has been revised to ‘raw material supply characteristics’.To 107 strengthen the link with food development and utilization, the original criterion of 'storage and 108 adaptability' was adjusted 'raw material supply characteristics'. After two rounds of expert 109 consultation and revision, this study established an evaluation framework comprising four 110 criteria—safety, nutritional and active substance, sensory quality, and raw material supply 111 characteristics—and 12 evaluation indicators, as presented in Table 2.In particular, safety 112 assesses the basic eligibility of _H. mutabilis_ as food; nutritional and active substance establish its 113 core nutritional and functional value; sensory quality determines consumer acceptability; and 114 raw material supply characteristics evaluate the species' capacity for a stable supply. 115 **Table 2. Final Evaluation Indexes** **Criterion layer** **Specific indicator** 1. Heavy metal contamination risk Safety 2. Microbiological contamination risk 3. Gossypol-related toxicity risk 4. Total flavonoid content Nutrition and active substances 5. Total amino acid content 6. Total mineral element content 7. Fiber coarseness Sensory quality 8. Lingering astringency 9. Mouthfeel smoothness 10. Yield per unit area Raw material supply characteristics 11. Drying yield 12. Storability 6 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.05.764492; this version posted November 6, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.11.05.764492) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) ## 116 Construction of the AHP-Based Comprehensive Evaluation System 117 and Weight Determination 118 The Analytic Hierarchy Process (AHP) was used to develop a comprehensive evaluation 119 system for the edible quality of _H. mutabilis_ flowers. The system comprises four hierarchical 120 layers: the goal layer (A), the criterion layer (C), the indicator layer (P) and bottom layer (D).The 121 goal layer provides a comprehensive evaluation of the edible quality of _H. mutabilis_ ; the 122 criterion layer comprises safety (C1), nutritional and active substance (C2), sensory quality (C3), 123 and raw material supply characteristics (C4); the indicator layer includes 12 specific evaluation 124 indicators; and the bottom layer covers the 11 _H. mutabilis_ cultivars under evaluation. The 125 detailed indicator framework is presented in Table 3. Specifically, endogenous toxin risk is 126 evaluated via free gossypol content, and Lingering astringency is evaluated via total tannin 127 content. 128 **Table 3. Evaluation model framework** **Goal layer (A)** **Criterion layer (C)** **Indicator layer (P)** **Bottom layer (D)** P1 Heavy metal C1 Safety P2 Microbiological P3 Gossypol-related Comprehensive evaluation of the food development potential of _H. mutabilis_ flowers P4 Total flavonoid C2 Nutritional and P5 Total amino acid active substance P6 Total mineral P7 Fiber coarseness C3 Sensory quality P8 Lingering P9 Mouthfeel D1–D11 _H._ _mutabilis_ cultivars to be evaluated C4 Raw material supply characteristics P10 Yield per unit P11 Drying yield P12 Storability 7 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.05.764492; this version posted November 6, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.11.05.764492) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 129 Based on expert consultation results, a 1–9 scale was used for pairwise comparisons of 130 indicators at the same hierarchical level to construct the decision matrix. SPSSAU was employed 131 to calculate the maximum eigenvalue, weight vector, and consistency index for each matrix. The 132 consistency ratio (CR) was then computed to assess matrix consistency, as expressed in the 133 following formula: _CI_  n -1  __ max - n  n -1   __ max - n 134 _CI_ _CR_  _RI_ 135 where λmax is the maximum eigenvalue of the judgment matrix, n is the matrix order, and 136 RI is the average random consistency index. When CR < 0.10, this indicates that the matrix 137 exhibits satisfactory consistency (Table 4). 138 **Table 4. A-C Consistency Test of Judgment Matrix** **Judgment** **value** **Weight** **W** **Layer** **Matrix** Goal layer A-C Criterion layer Goal layer A-C Criterion layer Goal layer A-C Criterion layer Goal layer A-C Criterion layer C1 sub **Column** **criterion** **order** C1,C2,C3, C4 C1,C2,C3, C4 C1,C2,C3, C4 C1,C2,C3, C4 **Row** **indicator** **λmax** **CI** **CR** **Consi** **stency** C1 1,2,3,2 0.4168 4.071 0.024 0.027 Pass C2 1/2,1,2,2 0.2695 4.071 0.024 0.027 Pass 1/3,1/2,1 C3 C4 ,1/2 1/2,1/2,2 ,1 0.1209 4.071 0.024 0.027 Pass 0.1928 4.071 0.024 0.027 Pass criteria layer C1-P P1,P2,P3 P1 1,5,3 0.6333 3.039 0.019 0.037 Pass 8 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.05.764492; this version posted November 6, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.11.05.764492) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) C1 sub C1-P P1,P2,P3 P2 1/5,1,1/3 0.1062 3.039 0.019 0.037 Pass criteria layer C1 sub C1-P P1,P2,P3 P3 1/3,3,1 0.2605 3.039 0.019 0.037 Pass criteria layer C2 sub C2-P P4,P5,P6 P4 1,1/3,1/5 0.1062 3.039 0.019 0.037 Pass criteria layer C2 sub C2-P P4,P5,P6 P5 3,1,1/3 0.2605 3.039 0.019 0.037 Pass criteria layer C2 sub C2-P P4,P5,P6 P6 5,3,1 0.6333 3.039 0.019 0.037 Pass criteria layer C3 sub C3-P P7,P8,P9 P7 1,3,2 0.5390 3.009 0.005 0.009 Pass criteria layer C3 sub C3-P P7,P8,P9 P8 1/3,1,1/2 0.1638 3.082 0.041 0.079 Pass criteria layer C3 sub C3-P P7,P8,P9 P9 1/2,2,1 0.2973 3.082 0.041 0.079 Pass criteria layer C4 sub criteria layer C4 sub criteria layer C4 sub criteria layer C4-P C4-P C4-P P10,P11,P P10 1,3,1/3 0.2605 3.039 0.019 0.037 Pass 12 P10,P11,P P11 1/3,1,1/5 0.1062 3.039 0.019 0.037 Pass 12 P10,P11,P P12 3,5,1 0.6333 3.039 0.019 0.037 Pass 12 9 139 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.05.764492; this version posted November 6, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.11.05.764492) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) ## 140 Calculation of Comprehensive Evaluation Value and Grade 141 Classification 142 After obtaining the criterion-level weights and indicator-level local weights, the global 143 weight of each indicator relative to the objective level was calculated as follows: 144 _Wi_  _WCj_  _Wij_ 145 In the formula, _Wi_ denotes the global weight of the _i_ -th indicator relative to the objective 146 layer; _WCj_ denotes the weight of the _j_ -th criterion relative to the objective layer; and _Wij_ denotes 147 the local weight of the _i_ -th indicator relative to the corresponding criterion layer. The global 148 weights of the _H. mutabilis_ hierarchy are presented in Table 5. 149 **Table 5. The overall priority of the criterion layer P with respect to the goal layer A** **Criterion C** **C weight to A** **Indicator P** **P weight to C** **P weight to A** P1 0.6333 0.2640 C1 0.4168 P2 0.1062 0.0442 P3 0.2605 0.1086 P4 0.1062 0.0286 C2 0.2695 P5 0.2605 0.0702 P6 0.6333 0.1707 P7 0.5390 0.0652 C3 0.1209 P8 0.1638 0.0198 P9 0.2973 0.0360 P10 0.2605 0.0502 C4 0.1928 P11 0.1062 0.0205 P12 0.6333 0.1221 10 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.05.764492; this version posted November 6, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.11.05.764492) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 150 Based on the measured values and scoring criteria, each indicator was classified into three 151 levels—5, 3, and 1 points—with 2 and 4 points as intermediate scores between adjacent levels 152 (Table 6). The score for each cultivar on each indicator is denoted as Pi; the composite score is 153 calculated using the total weighting Wi for the indicators. The formula for the composite score is 154 _F_    _W_ i  Pi  155 where F is the comprehensive evaluation value of a given cultivar; Wi is the total weight of 156 indicator i relative to the target layer; Pi is the score of the cultivar on indicator i; and n is the 157 number of evaluation indicators (n = 12 in this study). Cultivars with F ≥ 3.5 were classified as 158 Grade I (high edible quality); 3.0 ≤ F < 3.5 as Grade II (moderate); and F < 3.0 as Grade III 159 (limited). 160 **Table 6. Grading Standards for Evaluation Indicators of** _**Hibiscus mutabilis**_ **Food** 161 **Development Potential** **Indicator** **Score 5** **Score 3** **Score 1** P1 Heavy metal ≤ 64.15 μg/kg 96.88–129.61 μg/kg - 162.34 μg/kg contamination risk Total count 5×10 ⁶ –1×10 ⁷ CFU/g; E. coli 100–1×10³ Total count > 5×10 ⁷ CFU/g; E. coli > 1×10 ⁶ P2 Microbiological contamination risk P3 Endogenous toxin risk (free gossypol) P4 Total flavonoid content P5 Total amino acid content P6 Total mineral element content Total count < 1×10 ⁶ CFU/g; E. coli < 10 ≤ 20 mg/kg 22–27 mg/kg - 30 mg/kg - 0.040 g/100 g 0.025–0.030 g/100 g < 0.020 g/100 g ≥ 12.00 mg/g 8.50 ≤ T < 10.00 mg/g < 7.50 mg/g ≥ 34 000 mg/kg 30 000–32 000 mg/kg < 28 000 mg/kg 11 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.05.764492; this version posted November 6, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.11.05.764492) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) P7 Fiber coarseness Not rough Clearly rough Strongly rough P8 Lingering astringency < 3.50 4.00–4.79 ≥ 5.60 (tannin, g/100 g) Honey-smooth, no P9 Mouthfeel smoothness Slight roughness Obvious throat irritation discomfort P10 Yield per unit area ≥ 975 kg/ha 525–749.85 kg/ha < 375 kg/ha P11 Drying yield ≥ 0.20 0.13–0.149 < 0.115 (dry/fresh ratio) P12 Storability (browning < 30% browning 30%–70% browning ≥ 90% browning degree at the same time) 162 **Data Processing and Statistical Analysis** 163 All measurements were performed in triplicate, and the results are expressed as Mean ± SD. 164 Raw data were processed in Excel 2021 to compute the mean, SD, and composite score for each 165 parameter; each parameter was then converted to a 1–5 score according to the predefined scoring 166 criteria. AHP weights were computed and judgement matrix consistency was verified using 167 SPSSAU. Differences among cultivars in measured parameters were assessed by one-way 168 analysis of variance (ANOVA), with significance defined as _P_ < 0.05. Bar charts and heatmaps 169 were generated using Origin 2022 software to visualize the key evaluation indicators across 170 cultivars and the corresponding indicator-level scores. 12 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.05.764492; this version posted November 6, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.11.05.764492) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) # 171 Results ## 172 Measured Values of Comprehensive Evaluation Indicators 173 Notable differences among _H. mutabilis_ cultivars were observed in safety, nutritional value, 174 and bioactive compound profiles (Fig 2).After screening against safety standards, heavy metal 175 analysis revealed distinct accumulation patterns among the 11 cultivars. Specifically, 176 'Jinqiuhong', 'Danbanhong', 'Jinbiyu', and 'Jinxiuzi' accumulated relatively higher levels of 177 certain heavy metals, whereas 'Jinqiusong' and 'Bairi Huacai' showed lower levels, indicating 178 inter-cultivar variation in measured heavy metal concentrations. _Staphylococcus aureus_ was not 179 detected in any of the 11 cultivars. However, other microbial indicators varied among cultivars. 180 Variations in total flavonoid count, mould count, and coliform count were observed among 181 cultivars; specifically, 'Bairi Huacai', 'Chongbanbai', and 'Mudanfen' showed relatively high 182 levels of certain microbial indicators, whereas 'Jinxiuzi' presented a comparatively low overall 183 contamination risk. Free gossypol was detected in only four cultivars, with the highest 184 concentration observed in 'Mudanfen', followed by 'Jinbiyu', 'Caixia', and 'Danbanhong'. 185 Concentrations in the remaining cultivars were below the detection limit, indicating variation in 186 free gossypol concentrations among cultivars. 187 **Fig 2. Comparison of Safety, Nutritional, and Active Substance Indicators among Different** 188 _**Hibiscus mutabilis**_ **Cultivars.** Note:(A) Heatmap of heavy metal elements. (B) Heatmap of 189 microbial contamination indicators. (C) Gossypol-related toxicity. (D) Total flavonoid content. 190 Values are Means ± SD (n = 3). Different lowercase letters denote significant differences among 191 cultivars at _P_ < 0.05 (one-way ANOVA). "ND" indicates not detected (below the detection 192 limit). 13 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.05.764492; this version posted November 6, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.11.05.764492) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 193 Among nutritional and active substance, notable variations were observed in total flavonoid, 194 amino acid, and mineral element concentrations across cultivars. 'Jinxiuzi' exhibited the highest 195 total flavonoid content, followed by 'Zuifurong'. For amino acids, 'Jinxiuzi', 'Jinqiusong', and 196 'Chongbanbai' ranked highest. Regarding mineral elements, 'Jinqiuhong', 'Danbanhong', 'Caixia', 197 and 'Mudanfen' showed comparatively higher total mineral levels, whereas 'Bairi Huacai' and 198 'Jinqiusong' were relatively lower.Among sensory parameters, total tannin content varied 199 considerably across cultivars. 'Bairi Huacai' exhibited the highest total tannin content, followed 200 by 'Caixia' and 'Chongbanbai', indicating a relatively high risk of lingering astringency. 201 Conversely, 'Danbanhong', 'Jinqiusong', and 'Jinrui' showed lower total tannin content, 202 suggesting relatively better sensory palatability. Yield per unit area results showed that 'Bairi 203 Huacai' achieved the highest fresh and dried flower yields (kg/ha). 'Chongbanbai', 'Jinxiuzi', 204 'Zuifurong', and 'Jinqiuhong' ranked second, whereas 'Danbanhong', 'Mudanfen', and 'Jinrui' 205 showed relatively lower yields (Fig 3). 206 **Fig 3. Comparison of Nutritional, Active Substances, and Yields among Different** _**Hibiscus**_ 207 _**mutabilis**_ **Cultivars.** Note: (A) Heatmap of amino acid profiles. (B) Heatmap of mineral element 208 concentrations. (C) Total tannin content. (D) Yield per unit area of dried flowers and fresh 209 flowers. Values are means ± SD (n = 3). Different lowercase letters denote significant 210 differences among cultivars at _P_ < 0.05. ## 211 Indicator-Layer Scores of Different Cultivars 212 Based on the measured parameters and sensory evaluation results, the performance of the 11 213 cultivars across all 12 indicators was converted to 1–5 scores according to the predefined scoring 214 criteria; the results are presented in Fig 4.Based on the scoring results, 'Jinqiusong', 'Bairi 215 Huacai', and 'Jinbiyu' performed best on the heavy metal contamination risk, whereas 'Jinxiuzi' 14 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.05.764492; this version posted November 6, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.11.05.764492) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 216 and 'Jinqiuhong' scored highest on the microbial contamination risk index. For gossypol-related 217 toxicity risk, 'Zuifurong', 'Jinqiusong', 'Jinqiuhong', 'Chongbanbai', 'Bairi Huacai', and 'Jinrui' 218 ranked highest. Among nutritional and active substance, 'Jinxiuzi' led in total flavonoid and total 219 amino acid content, while 'Danbanhong', 'Jinqiuhong', and 'Mudanfen' ranked highest in mineral 220 element concentrations. For sensory quality, most cultivars scored well on fibre coarseness, 221 whereas marked differences were observed in lingering astringency and mouthfeel smoothness. 222 Regarding raw material supply characteristics, 'Bairi Huacai' ranked highest in yield per unit 223 area, 'Chongbanbai' in storability, and 'Danbanhong' in drying yield. 224 **Fig 4. Scores of Criterion Layer Indicators among Different** _**Hibiscus mutabilis**_ **Cultivars.** 225 Note: Comprehensive quality evaluation of 11 _Hibiscus mutabilis_ cultivars based on raw material 226 supply characteristics, sensory quality, nutritional and active substance, and safety-related 227 indicators. Each indicator was evaluated using a five-point scoring system, with higher scores 228 indicating better quality performance. Undesirable attributes, including lingering astringency, 229 fiber coarseness, endogenous toxin risk, microbial contamination risk, and heavy metal 230 contamination risk, were scored so that higher scores indicate less pronounced undesirable 231 characteristics or lower risks. Colors range from blue (1, poorest performance) to red (5, best 232 performance) ## 233 Comprehensive Evaluation Values and Grade Classification 234 Based on the indicator scores and their global weights relative to the objective layer, 235 comprehensive evaluation values were calculated for the 11 cultivars, ranked, and assigned to 236 grade levels, as presented in Table 7. For example, 'Bairi Huacai' achieved a weighted 237 comprehensive score of 3.6043 across the 12 indicators, corresponding to Grade I. 15 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.05.764492; this version posted November 6, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.11.05.764492) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 238 As presented in Table 7, the comprehensive evaluation scores of the 11 cultivars ranged 239 from 2.5785 to 3.6043. 'Bairi Huacai' achieved the highest score (3.6043), followed by 240 'Chongbanbai' (3.5533) and 'Jinqiusong' (3.5055). Based on the grading criteria, three cultivars 241 were classified as Grade I, representing 27.27% of the samples: 'Bairi Huacai', 'Chongbanbai', 242 and 'Jinqiusong', indicating superior edible quality. Six cultivars were classified as Grade II 243 (54.55%): 'Jinqiuhong', 'Zuifurong', 'Danbanhong', 'Jinxiuzi', 'Jinbiyu', and 'Caixia', representing 244 an intermediate overall performance. The remaining two cultivars, 'Mudanfen' and 'Jinrui', were 245 classified as Grade III (18.18%), with relatively lower overall scores. 246 **Table 7. Comprehensive evaluation scores and grading of 11** _**Hibiscus mutabilis**_ **Cultivars** **Rank** **Cultivar** **Final score** **Grade** 1 ' Bairi Huacai ' 3.6043 I 2 ' Chongbanbai ' 3.5533 I 3 ' Jinqiusong ' 3.5055 I 4 ' Jinqiuhong ' 3.4930 II 5 ' Zuifurong ' 3.4069 II 6 ' Danbanhong ' 3.3178 II 7 ' Jinxiuzi ' 3.2380 II 8 ' Jinbiyu ' 3.1325 II 9 ' Caixia ' 3.1269 II 10 ' Mudanfen ' 2.7113 III 11 ' Jinrui ' 2.5785 III 16 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.05.764492; this version posted November 6, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.11.05.764492) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) # 247 Discussion ## 248 Analysis of Evaluation Indicator Weights 249 As shown by the criterion layer weights relative to the goal layer, safety received the 250 highest weight, followed by nutritional and active substance, raw material supply characteristics, 251 and sensory quality. This confirms that safety is the primary consideration when transitioning 252 _Hibiscus mutabilis_ L. from ornamental plants to food resources[24-25]. 253 At the indicator layer, the global weights of measured heavy metal concentrations, mineral 254 element concentrations, and gossypol-related toxicity risk have the greatest influence on the 255 comprehensive evaluation results, indicating that safety, together with nutritional and active 256 substance, jointly determines the priority of _H. mutabilis_ flowers for food development[26-27]. 257 By comparison, sensory indicators such as fibre coarseness and mouthfeel smoothness received 258 relatively lower weights; nevertheless, given their influence on consumer acceptance and product 259 palatability, they should not be neglected in future product development[28-29]. 260 The AHP-derived weights primarily reflect the relative importance assigned by experts and 261 are inherently influenced by the composition of the expert panel, the construction of the 262 judgement matrix, and the definition of evaluation objectives[30-31]. Accordingly, the 263 evaluation model established in this study is best suited for preliminary screening of _H. mutabilis_ 264 cultivars; further validation should incorporate empirical measurement data, food safety 265 standards, and product processing trials. ## 266 Analysis of Comprehensive Evaluation Results 267 Based on the comprehensive evaluation scores, the 11 cultivars were classified into three 268 grades. Three cultivars (27.27%) were assigned to Grade I: 'Bairi Huacai', 'Chongbanbai', and 17 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.05.764492; this version posted November 6, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.11.05.764492) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 269 'Jinqiusong'. Six cultivars (54.55%) were assigned to Grade II: 'Jinqiuhong', 'Zuifurong', 270 'Danbanhong', 'Jinxiuzi', 'Jinbiyu', and 'Caixia'. The remaining two cultivars (18.18%) were 271 assigned to Grade III: 'Mudanfen' and 'Jinrui'. 272 Of the 11 cultivars, 'Bairi Huacai' (3.6043)achieved the highest overall score under the 273 present weighting scheme. However, its high total tannin content suggests a potential limitation 274 related to astringency that warrants further sensory evaluation. 'Chongbanbai' ranked second 275 (3.5533), with excellent safety and several nutritional and active substance indicators; however, 276 it was relatively weak in storability. If used for food processing or fresh produce development, 277 complementary post-harvest preservation and short-term storage technologies should be 278 prioritized. 'Jinqiusong' scored 3.5055, meeting Grade I criteria but with a comparatively lower 279 overall score. 280 Grade II cultivars achieved composite scores between 3.0 and 3.5, indicating intermediate 281 performance under the present evaluation framework. These cultivars exhibited strengths in 282 certain indicators but limitations in others, suggesting that their suitability for food development 283 may depend on specific processing requirements and intended applications. # 284 References 285 [1] Li YP, Zhang XL, Wu WT, Miao SX, Chang JL. Chromosome and karyotype analysis of 286 _Hibiscus mutabilis_ f. _mutabilis_ . Frontiers in Life Science. 2015;8(3): 300-304. doi: 287 10.1080/21553769.2015.1041166 288 [2] Yang YZ, Liu XD, Shi XQ, Ma J, Zeng XM, Zhu ZS, et al. A high-quality, chromosome 289 level genome provides insights into determinate flowering time and color of cotton rose 290 ( _Hibiscus mutabilis_ ). Front Plant Sci. 2022;13: 818206. doi: 10.3389/fpls.2022.818206 18 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.05.764492; this version posted November 6, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.11.05.764492) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 291 [3] Shi X, Wang T, Ai S, Li J. Integrated transcriptomics and metabolomics analysis for the 292 mechanism underlying white-to-pink petal color transition in _Hibiscus mutabilis_ flowers. 293 Phyton (B Aires). 2024;93(10): 2571-2581. doi: 10.32604/phyton.2024.056606 294 [4] Blythe EK. Hardwood cutting propagation of Confederate rose using auxin and wounding 295 treatments. HortTechnology. 2012;22(4): 476-478. doi: 10.21273/HORTTECH.22.4.476 296 [5] Wei J, Liu X, Li C, Yang Y, Song C, Chen Y, et al. Identification and characterization of 297 _Hibiscus mutabilis_ varieties resistant to Bemisia tabaci and their resistance mechanisms. 298 Insects. 2024;15(6): 454. doi: 10.3390/insects15060454 299 [6] Sari MF, Purwantoro A. Diversity analysis of 15 _Hibiscus_ accession based on RAPD 300 marker. Ilmu Pertanian (Agricultural Science). 2019;3(2): 89-94. doi: 10.22146/ipas.33211 301 [7] Sang SY, Liu YQ, Li X, Ma J, Liu XL, Yang YZ. A novel gene silencing strategy based on 302 tobacco rattle virus in _Hibiscus mutabilis_ . PeerJ. 2024;12: e18211. doi: 10.7717/peerj.18211 303 [8] Jia ZS. Cloning and functional study of HmF3H and HmANS, key genes of color change in 304 _Hibiscus mutabilis_ 'Drunk Girl' [in Chinese]. M.Sc. Thesis, Sichuan Agricultural University. 305 2024. 306 [9] Xue YT, Liu J, Liu XL, Li FW, Shi XQ, Ma J, et al. Quantitative classification and principal 307 component analysis of 18 _Hibiscus mutabilis_ cultivars and 2 related species [in Chinese]. 308 Journal of Central South University of Forestry & Technology. 2022;42(1): 70-77. 309 [10] Zhang L, Xu Q, Yong X, Wu MX, Jiang BB, Jia Y, et al. Effects of water deficit on two 310 cultivars of _Hibiscus mutabilis_ : a comprehensive study on morphological, physiological, and 311 metabolic responses. Plant Physiol Biochem. 2024;217: 109269. doi: 312 10.1016/j.plaphy.2024.109269 19 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.05.764492; this version posted November 6, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.11.05.764492) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 313 [11] Puckhaber LS, Stipanovic RD, Bost GA. Analyses for flavonoid aglycones in fresh and 314 preserved _Hibiscus_ flowers. In: Janick J, Whipkey A, editors. Trends in New Crops and 315 New Uses. Alexandria (VA): ASHS Press; 2002. pp. 556-563. 316 [12] Nie GH, Xiang QB. Study on characteristics and landscape applications of _Hibiscus_ 317 _mutabilis_ L. [in Chinese]. Horticulture & Seed. 2012;32(6): 81-83. 318 [13] Ren QJ, Shan WT, Liu YL. Exploitation and utilization of woody economic plant H _ibiscus_ 319 _syriacus_ L. [in Chinese]. Resources Development & Market. 2000;16(1): 29-30. 320 [14] Pan SL. Hundred flowers recipe (X): flower of _Hibiscus mutabilis_ [in Chinese]. Garden. 321 2006;(10): 52-53. 322 [15] Hu P, Wang Y, Ye L, Yan X, Zeng YJ, Jiang ZR, et al. A novel anti-inflammatory flavonoid 323 from flowers of _Hibiscus mutabilis_ L. Nat Prod Res. 2025;39(21): 6279-6286. doi: 324 10.1080/14786419.2024.2378993 325 [16] Xiao M, Zhang T, Cao F, Liang W, Yang Y, Huang T, et al. Anti-influenza properties of 326 tiliroside isolated from _Hibiscus mutabilis_ L. J Ethnopharmacol. 2023;303: 115918. doi: 327 10.1016/j.jep.2022.115918 328 [17] Zhu Z, Zeng X, Shi X, Ma J, Liu X, Li Q. Transcription and metabolic profiling analysis of 329 three discolorations in a day of _Hibiscus mutabilis_ . Biology. 2023;12(8): 1115. doi: 330 10.3390/biology12081115 331 [18] Pounders CT, Sakhanokho HF. 'Hapa White', 'Hapa Pink', and 'Hapa Red' interspecific 332 hybrid _Hibiscus_ cultivars. HortScience. 2016;51(12): 1616-1617. doi: 333 10.21273/HORTSCI11291-16 20 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.05.764492; this version posted November 6, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.11.05.764492) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 334 [19] Pires TCS, Barros L, Santos-Buelga C, Ferreira ICFR. Edible flowers: emerging 335 components in the diet. Trends Food Sci Technol. 2019;93: 244-258. doi: 336 10.1016/j.tifs.2019.09.020 337 [20] Cheng X, Feng Y, Chen D, Luo C, Yu X, Huang C. Evaluation of Rosa germplasm 338 resources and analysis of floral fragrance components in R. rugosa. Front Plant Sci. 339 2022;13: 1026763. doi: 10.3389/fpls.2022.1026763 340 [21] Zartha Sossa JW, Halal W, Hernandez Zarta R. Delphi method: analysis of rounds, 341 stakeholder and statistical indicators. Foresight. 2019;21(5): 525-544. doi: 10.1108/FS-11 342 2018-0095 343 [22] Staykova MP. Rediscovering the Delphi technique: a review of the literature. Adv Soc Sci 344 Res J. 2019;6(1): 218-229. doi: 10.14738/assrj.61.5959 345 [23] Zhou R, Lin L, Li Q, Zeng S, Ye J, Luo C. Application of a SODOSM-based MCDM 346 method for evaluating comprehensive fruit quality: a case study of pineapple. PLoS One. 347 2025;20(9): e0330496. doi: 10.1371/journal.pone.0330496 348 [24] Benvenuti S, Mazzoncini M. The biodiversity of edible flowers: discovering new tastes and 349 new health benefits. Front Plant Sci. 2021;11: 569499. doi: 10.3389/fpls.2020.569499 350 [25] Matyjaszczyk E, Śmiechowska M. Edible flowers: benefits and risks pertaining to their 351 consumption. Trends Food Sci Technol. 2019;91: 670-674. doi: 10.1016/j.tifs.2019.07.017 352 [26] Lu B, Li M, Yin R. Phytochemical content, health benefits, and toxicology of common 353 edible flowers: a review (2000-2015). Crit Rev Food Sci Nutr. 2016;56(sup1): S130-S148. 354 doi: 10.1080/10408398.2015.1078276 21 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.05.764492; this version posted November 6, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.11.05.764492) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 355 [27] Li X, Chen X, Li MX, Zeng XM. Response of different _Hibiscus mutabilis_ L. varieties to 356 cadmium stress [in Chinese]. Sichuan Lin Ye Ke Ji. 2024;45(1): 58-65. doi: 357 10.12172/202303300001 358 [28] Takahashi JA, Rezende FAGG, Moura MAF, Dominguete LCB, Sande D. Edible flowers: 359 bioactive profile and its potential to be used in food development. Food Res Int. 2020;129: 360 108868. doi: 10.1016/j.foodres.2019.108868 361 [29] Zhejiang Chinese Medical University. _Hibiscus mutabilis_ meat tenderizer powder, its 362 preparation process and detection method for application effect [in Chinese]. China patent 363 201410639714.7. 2015 Feb 18. 364 [30] Ishizaka A, Labib A. Review of the main developments in the Analytic Hierarchy Process. 365 Expert Syst Appl. 2011;38(11): 14336-14345. doi: 10.1016/j.eswa.2011.04.143 366 [31] Liu W. Research on the construction and application of food safety index [in Chinese]. 367 Ph.D. Thesis, Huazhong Agricultural University. 2013. 22 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.05.764492; this version posted November 6, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.11.05.764492) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY 4.0 International license.](http://creativecommons.org/licenses/by/4.0/)