Unlocking the physicochemical, bioactive, and antioxidant potential of Indonesian Liberica coffee: roles of geographical origin and processing

Abstract

Commercial interest in Liberica coffee (Coffea liberica) in Southeast Asia is growing, but evaluations of the impact of postharvest treatments on its physicochemical and bioactive characteristics remain limited. This study investigates the effects of geographic origin (Jambi, Kayong, Meranti) and postharvest methods (natural, honey) on the properties of Indonesian Liberica green coffee beans. The results indicate that geographic origin has a significant effect (p < 0.05) on most parameters, reflecting strong environmental influences. Postharvest processing primarily affects moisture content, lipid fractions, and antioxidant capacity. The Jambi honey-processed sample exhibited the highest alkaloid content (1.53 g/100 g db), while the highest diterpene content was recorded in Kayong natural-processed beans (304.70 mg/100 g db). The Meranti sample processed naturally exhibited the highest caffeoylquinic acids (CQAs) content (5.38 g/100 g db) and the strongest antioxidant activity (IC₅₀: 101–124 μg/mL). Multivariate analysis revealed primary clustering based on location, while processing methods served as a secondary distinguishing factor. These findings provide preliminary insights into the associations of geographical origin and postharvest processing with the physicochemical and antioxidant properties of Liberica green bean.

Highlights

• •Geographical origin shapes Liberica bioactive profiles.
• •Natural and honey processing affect coffee composition.
• •Meranti shows the highest chlorogenic acids and antioxidant activity.
• •Chlorogenic acids correlate with antioxidant capacity.
• •Origin has a greater impact than processing.

1. Introduction

Coffee (Coffea sp.) is a global agricultural commodity valued not only for its stimulant properties but also as a rich source of bioactive compounds with potential health benefits (Ismail et al., 2024). Global coffee production and trade are largely dominated by Coffea arabica and Coffea canephora. This domination has led to a biased, limited understanding of the global chemical profile of coffee, focusing solely on these two species alone. Among cultivated coffee species, Coffea liberica has attracted increasing attention due to its distinctive morphological traits, chemical composition, and functional potential. C. liberica exhibits remarkable ecological adaptability, allowing it to grow well in marginal environments such as peatlands and low-fertility soils, where Arabica and Robusta are less suitable (Davis et al., 2022). In addition, this species shows relatively high resistance to major coffee diseases, including coffee leaf rust and berry borer infestation, making it a promising crop for sustainable coffee production systems (Martono et al., 2020; Wahyuni et al., 2022). Moreover, Liberica coffee shows superior resistance to high temperatures and strong adaptability to climate change, highlighting its potential as a climate-resilient crop under ongoing global warming conditions (Davis et al., 2022; Hafif et al., 2024). Beyond its agronomic value, in Indonesia Liberica coffee is also used as a conservation crop, commonly planted as a secondary vegetation layer following mangrove restoration, and is capable of growing at very low elevations, including areas approximately 2 m above sea level.

Liberica coffee has lower caffeine (1.10%) than Arabica (1.46%) and Robusta (2.70%), which may benefit caffeine-sensitive consumers (Sunarharum et al., 2023; Vanathi et al., 2025). It also possesses a distinctive sensory profile, often described by a jackfruit-like aroma, along with chocolatey notes, lower bitterness than Robusta, and acidity comparable to Arabica (Insanu et al., 2021; Wibowo et al., 2021). Beyond sensory attributes, Liberica coffee has gained interest as a potential functional beverage due to its bioactive compounds, including caffeoylquinic acids and alkaloids, which contribute to antioxidant activity. Previous studies have highlighted the potential of Liberica coffee for producing lower caffeine products with enhanced antioxidant properties, as well as the role of fermentation in improving sensory quality and phenolic content (Tarigan et al., 2024; Vanathi et al., 2025).

Several studies have reported substantial regional variation in the antioxidant capacity and phenolic composition of Liberica coffee from different Indonesian regions, including Riau, Aceh, and Jambi (Insanu et al., 2021). The differences in physicochemical properties and antioxidant activity in Liberica coffee from various districts in Jambi highlight the strong influence of geographical origin (Hanifah et al., 2022), which, together with factors such as soil characteristics, climate, altitude, and local cultivation practices, shapes the chemical composition and functional properties of green coffee beans (Abebe et al., 2020; Ahmed et al., 2021).

Postharvest methods, such as natural and honey significantly affect the physicochemical and sensory characteristics of coffee beans by altering chemical composition and fermentation dynamics during processing. Washed processing was not included in this study as it is not commonly practiced by GI Liberika coffee farmers in the study regions, partly due to limited access to clean water in these areas. These differences are largely associated with variations in mucilage removal and fermentation intensity, which can substantially modify the chemical profile of the beans. In particular, natural and pulped natural (honey) processing differ in fermentation intensity and drying conditions, leading to measurable changes in caffeoylquinic acids, lipids, caffeine, and antioxidant activity (Aswathi et al., 2024; Sunarharum et al., 2023; Wulandari et al., 2021). These factors indicate that the final physicochemical and functional properties of coffee are determined by the interaction between geographical origin and postharvest processing rather than by either factor alone. Differences in composition and bioactive profiles among coffees from different regions highlight the importance of the Geographical Indication (GI) concept as a quality differentiation tool that reflects both environmental and human factors (Teuber, 2010). In Indonesia, several Liberica coffees have obtained GI certification; however, systematic comparative evaluations of their physicochemical characteristics, bioactive compounds, and antioxidant capacity across different postharvest processing methods remain limited.

This study aimed to reveal the bioactive and antioxidant potential of Indonesian Liberica coffee by examining the role of geographical origin and postharvest processing methods on physicochemical characteristics, main bioactive compound content (caffeoylquinic acid, alkaloids, and diterpenes), and in vitro antioxidant capacity. This study was conducted on Liberica coffee labeled with Geographical Indication (GI) originating from Tanjung Jabung Barat (Jambi), Kepulauan Meranti (Riau), and Kayong Utara (West Kalimantan), processed using natural and honey methods. The results of this study are expected to provide a scientific basis for understanding how geographical factors and processing shape the bioactive and antioxidant potential of Liberica coffee, and to provide a preliminary basis for future research on postharvest practices and the geographical characterization of Liberica coffee from Indonesian Geographical Indication (GI) regions. To the best of our knowledge, this is the first study to systematically compare the physicochemical characteristics, bioactive compound profiles (caffeoylquinic acids, alkaloids, and diterpenes), and in vitro antioxidant capacity of Geographical Indication-certified Indonesian Liberica coffees from three distinct origins (Jambi, Riau, and West Kalimantan) under two postharvest processing methods (natural and honey), providing a comprehensive biochemical basis for geographical differentiation and postharvest optimization of this underexplored species.

2. Materials and methods

2.1. Sampling from geographical indication regions and postharvest processing of Liberica coffee

Liberica coffee cherries were harvested between May and June 2024 from three Geographical Indication (GI) production regions in Indonesia: Tanjung Jabung Barat Jambi, North Kayong West Borneo, and the Meranti Islands Riau to compare coffee from various geographical origins. Currently, only three Liberica coffees in Indonesia have obtained Geographical Indication (GI) certification. Samples were obtained from one certified farmer group per origin, selected through purposive sampling based on active compliance with established GI standard practices, to ensure that the samples authentically represent the profile of certified Liberica coffee from each geographic origin. The detailed sampling locations are presented in Fig. 1.

Fig. 1.

Regions of Liberica coffee geographical indication.

Fresh coffee cherries harvested from each location are processed separately using the natural and honey post-harvest methods, in accordance with the Geographical Indication (GI) standards for each origin. Fresh coffee cherries harvested from each location are processed separately using the natural and honey post-harvest methods, in accordance with the Geographical Indication (GI) standards for each origin. Natural processing begins with the selection and sorting of fresh, ripe, and good-quality coffee cherries, followed by drying on elevated drying racks inside a greenhouse solar dryer for 20–30 days. Honey processing also begins with the selection and sorting of ripe coffee cherries, followed by pulping to remove the outer skin while retaining the entire mucilage layer (100% mucilage retention; no washing is performed after pulping). The mucilage-coated beans are then dried on elevated drying racks inside a greenhouse solar dryer for 7–15 days, with the beans turned 2–3 times a day to ensure even drying. For both methods, the drying process is stopped when the moisture content reaches approximately 12%, which is initially assessed by the rattling sound produced when the beans are stirred, indicating sufficient moisture reduction, and then verified through an oven-drying method in the laboratory. The drying temperature and duration vary by location depending on local conditions, as processing is carried out by individual farmer groups in accordance with the standard practices of their respective Geographical Indications (GI). Next, the seed coats are removed to obtain green seeds for all subsequent analyses.

Detailed postharvest processing stages for each geographical indication (Jambi, Kayong, and Meranti) are provided in Supplementary 2 Table S5.

2.2. Chemicals and reagents

The chemicals were sourced from Sigma -Aldrich (St. Louis, USA): the standards CQA, trigonelline, diterpene (kahweol and cafestol), 2,2-diphenyl-1-picrylhydrazyl (DPPH), 2,4,6-tri(2-pyridyl)-1,3,5-triazine (TPTZ), and 6-hydroxy-2,5,7,8-tetramethylchroman-2-carboxylic acid (Trolox). The standard caffeine, iron (III) chloride hexahydrate (FeCl₃·6H₂O), ascorbic acid, hydrochloric acid (HCl), formic acid, acetic acid, sodium acetate trihydrate, ethanol, and methanol (LC–MS grade) were supplied by Merck (Darmstadt, Germany). Ultra-high purity water was prepared by filtration using a Milli-Q system.

2.3. Physical characteristics of coffee cherries and green beans

The physical properties of coffee cherries and green beans were evaluated, including dimensions, volume, color, bulk density, and bean mass. Cherry and bean dimensions (length, width, and thickness) were measured using a caliper. Volumes were calculated using ellipsoidal equations, as previously described by Ismail et al. (2014). Dimension measurements were performed on 50 cherries and 100 green beans per sample. To validate the calculated volume, cherry volume was also determined using a water-displacement method by immersing 10 cherries in a graduated cylinder containing 50 mL of water. Cherry color was measured on-site using a smartphone-based colorimeter application (Colorimeter Lab Tools v1.6.6.6 APL) installed on a Realme C53 smartphone (Android 13). Green bean color was analyzed using a chromameter (Minolta CR-400, Japan) under D65 illumination following the CIE Lab* system.

2.4. Sample preparation and extraction

Green coffee beans were ground in a household grinder using liquid nitrogen to produce a homogeneous powder, which was stored at −20 °C until analysis. Coffee extracts for pH, total dissolved solids, bioactive compounds, and antioxidant assays were prepared by extracting 5 g of ground coffee with boiling distilled water, followed by cooling, dilution to 250 mL, and filtration (Whatman No. 1) (Herawati et al., 2022).

Diterpene extraction was performed using the Direct Hot Saponification (DHS) method following Dias et al. (2010). Briefly, coffee bean samples were saponified in ethanolic KOH under heating, and the unsaponifiable fraction was extracted with tert-butyl methyl ether. The organic phase was collected, evaporated to dryness, and resuspended in a methanol–water mobile phase prior to analysis.

2.5. Proximate analysis

Moisture, ash, crude fat, and crude protein contents were determined according to the AOAC Official Methods 925.10, 923.03, 920.39, and 984.13, respectively (AOAC, 2019). Carbohydrate content was calculated by difference, as the remaining percentage after subtracting the percentages of moisture, ash, crude fat, and crude protein from 100%. All analyses were performed in triplicate, and each replicate was analyzed in duplicate (duplo), with the duplicate measurements averaged before statistical analysis.

2.6. Mineral and fatty acid composition

Mineral analysis was performed after nitric acid digestion and quantified using Inductively Coupled Plasma Mass Spectrometry (ICP-MS) (iCAP RQ Plus, Thermo Scientific) equipped with an autosampler (ASX-560, Teledyne CETAC Technologies) following standard procedures. Approximately 1 g of sample was digested with 15 mL of concentrated HNO₃ for 30 min until the brown fumes disappeared. The digest was then diluted with distilled water to a final volume of 100 mL and filtered through a 0.25 μm membrane filter. Quantification was performed using an external calibration curve prepared from multi-element standard solutions in the range of 0–1000 ppb (Albals et al., 2021).

Fatty acid composition was determined by gas chromatography after conversion to fatty acid methyl esters (FAMEs) following (Martín et al., 2001). Coffee oil was converted to fatty acid methyl esters (FAMEs) using a BF₃–methanol solution. The resulting FAMEs were analyzed using a gas chromatograph equipped with a flame ionization detector (GC-FID; Hitachi GC-263-50) and a DEGS capillary column. Nitrogen and hydrogen were used as carrier and fuel gases at flow rates of 1.0 and 0.5 kgf cm⁻², respectively. The oven temperature was programmed from 150 to 180 °C, while injector and detector temperatures were set at 200 and 250 °C, respectively. A sample volume of 2 μL was injected. Detailed methods for diterpene extraction, proximate analysis, mineral analysis, and fatty acid analysis are provided in Supplementary 1.

2.7. Determination of bioactive compounds

Filtered extracts (0.45 μm PTFE) were analyzed using an HPLC equipped with a UV–Vis detector (LC-20 CE system; SHIMADZU Corp., Kyoto, Japan), separated using a Zorbax C18 column (id. 4.6 × 150 mm, 5 μm) (Agilent Technologies, Santa Clara, USA). The coffee extract (20 μL) was injected into the HPLC system for all analyses (CQA, alkaloids, and diterpenes).

2.7.1. Caffeoylquinic acids (CQAs)

CQAs content analysis was conducted using an isocratic elution with a mobile phase consisting of a mixture of HPLC-grade methanol (A) and 0.05% formic acid in milli-Q water (B), at a flow rate of 1 mL/min. Caffeoylquinic acid compounds were detected at 320 nm. CQA analysis was performed at a wavelength of 320 nm (Herawati et al., 2022). Three standards (3CQA, 4CQA, and 5CQA) were used to measure caffeoylquinic acid with a concentration range (3CQA 1.95–62.5 μg/mL, 4CQA 3.9–125 μg/mL, 5CQA 7.8–250 μg/mL) on a 6-point calibration curve at three replicates (Limit of Detection [LoD] 3CQA = 0.74 μg/mL and R² = 0.999; LoD 4CQA = 0.98 μg/mL and R² = 1; and LoD 5CQA = 1.78 μg/mL and R² = 1).

2.7.2. Alkaloids

Alkaloid content analysis was conducted using an isocratic elution with a mobile phase consisting of a mixture of HPLC-grade methanol (A) and 0.3% formic acid in milli-Q water (B), at a flow rate of 0.4 mL/min. Detection of alkaloid compounds was performed at a wavelength of 265 nm. Alkaloid content analysis included trigonelline and caffeine compounds (Herawati et al., 2022). A calibration curve was created with 3 replicates at 6 points, covering concentrations (trigonelline 0.7–25 μg/mL, theobromine 0.04–2.5 μg/mL, and caffeine 3.1–100 μg/mL). (Limit of Detection [LoD] for trigonelline = 0.78 μg/mL and R² = 0.999; LoD for theobromine = 0.06 μg/mL and R² = 0.999; and LoD for caffeine = 1.51 μg/mL and R² = 0.999).

2.7.3. Diterpenes

Elution was performed using an isocratic methanol: water (85:15) mobile phase for 20 min at a flow rate of 0.7 mL/min. Analysis was performed at 290 nm for kahweol and 230 nm for cafestol (Liguori et al., 2024), with standard curves constructed using 3 replicates at 6 points over the concentration range (kahweol 1.6–50 μg/mL and cafestol 7.8–250 μg/mL). (Limit of Detection [LoD] for kahweol = 10.2 μg/mL and R² = 0.999; LoD for cafestol = 5.3 μg/mL and R² = 0.999). The analysis was performed three times. The HPLC method validation data are presented in Supplementary 2 (Table S1).

2.8. In vitro antioxidant capacity

Antioxidant capacity was evaluated using DPPH radical scavenging and ferric reducing antioxidant power (FRAP) assays. DPPH Antioxidant analysis was determined using a method developed by Tunnisa et al. (2022), with several modifications. A total of 100 μL of each coffee bean filtrate sample was prepared at 7.8–250 μg/mL in a 96-well microplate and mixed with 100 μL of 125 μM DPPH solution (in ethanol). The samples were incubated for 30 min at room temperature, after which absorbance was measured at 517 nm using a microplate reader. A reagent blank (DPPH solution mixed with ethanol without sample) and a sample blank (coffee sample solution mixed with ethanol without DPPH solution, to account for the intrinsic color of the coffee extract) were included for each sample concentration. The corrected absorbance values were used to calculate percentage inhibition. IC₅₀ values were determined from concentration–response curves using linear regression analysis in Microsoft Excel, calculated as the concentration corresponding to 50% inhibition based on the resulting regression equation, and are reported as the mean ± standard deviation of three independent composite replicates (n = 3) per origin × processing combination, with each replicate analyzed in analytical duplicate; duplicate measurements were averaged and were not treated as independent replicates in statistical analysis.

The FRAP antioxidant analysis was conducted according to the method described by Hanifah et al. (2022), with modifications. FRAP reagent was prepared by mixing 2.5 mL of 10 mM tripyridyltriazine (TPTZ) with 40 mM HCl; 2.5 mL of 20 mM FeCl₃.6H₂O (Sigma-Aldrich, USA); and 25 mL of 0.3 mM acetate buffer (pH 3.6), then incubating at 37 °C for 30 min. The analysis was performed by mixing 180 μL of FRAP reagent with 20 μL of coffee filtrate, then incubating for 30 min at 37 °C. The absorbance of the sample was measured with a microplate reader at 595 nm. Absorbance values were corrected against a reagent blank (FRAP reagent mixed with extraction solvent instead of sample) to account for background absorbance, including the intrinsic color of the coffee extract. FRAP values were calculated from a Trolox standard calibration curve and expressed as g TEAC/100 g db, based on three independent composite replicates (n = 3) per origin × processing combination, each analyzed in analytical duplicate. Detailed procedures for the antioxidant assays using the DPPH and FRAP methods are provided in Supplementary 3. The standard calibration curves and representative concentration–response curves are presented in Supplementary 2 (Figs. S5–S8).

2.9. Statistical analysis

Prior to statistical analysis, data normality was assessed using the Shapiro–Wilk test, and homogeneity of variance was evaluated using Levene's test. Parametric analyses were performed only when both assumptions were satisfied (p > 0.05). For cherry coffee data, differences among geographical origins were analyzed using one-way analysis of variance (ANOVA), followed by Duncan's multiple range test for pairwise comparisons. The chemical composition, bioactive compounds, and antioxidant activity of green coffee beans were evaluated for the effects of geographical origin, postharvest processing method, and their interaction using two-way ANOVA, followed by Tukey's honest significant difference (HSD) test for multiple comparisons. When either the normality or homogeneity assumption was violated, nonparametric analyses were applied. Differences among geographical origins were assessed using the Kruskal–Wallis test followed by Dunn's post hoc test, whereas differences between postharvest processing methods were evaluated using the Mann–Whitney U test. Statistical significance was established at p < 0.05. Statistical analysis and multivariate analysis, including Principal Component Analysis (PCA), were performed using XLSTAT 2019 (Addinsoft, Paris, France). Prior to PCA, all variables were autoscaled (mean-centered and divided by standard deviation) using a correlation matrix. Heatmap visualization was performed using MetaboAnalyst 6.0 (www.metaboanalyst.ca). Features were autoscaled prior to hierarchical clustering using Euclidean distance with Ward's linkage.

3. Results

3.1. Physical properties of coffee cherries from different origins

Shape and size are key attributes for product classification, grading, and quality assessment, and are often among the first physical cues consumers evaluate before purchase (Quiñones-Ruiz, 2020; Saparita et al., 2019). In this study, the physical properties of Liberica coffee cherries were evaluated based on major dimensional parameters (length, width, and thickness), geometric volume, fruit mass, and skin percentage. Table 1 summarizes the physical properties of Liberica coffee cherries from Indonesian Geographical Indication regions. Cherries from Meranti were the largest and heaviest, whereas those from Jambi and Kayong were smaller, though still exceeding the typical size reported for Arabica and Robusta. Liberica coffee cherries are four times heavier than Arabica and Robusta cherries (Ismail et al., 2014).

Table 1.

Dimensions and physical properties of Liberica coffee cherries.

Properties	Jambi		Kayong		Meranti
	min-max	average ± SD	min-max	average ± SD	min-max	average ± SD
Length (mm)	16.25–29.95	22.43 ± 2.13a	17.05–26.60	22.67 ± 2.28a	16.05–33.95	24.26 ± 3.31b
Width (mm)	15.10–23.50	20.25 ± 1.91b	13.6–21.95	19.13 ± 1.87a	14.85–25.5	20.84 ± 2.29b
Thick (mm)	14.60–28.60	18.51 ± 2.08b	11.95–19.15	16.32 ± 1.69a	14.4–25	19.58 ± 2.17c
Mass (g)	2.11–7.21	4.99 ± 1.16b	2.14–6.15	4.11 ± 1.02a	2.18–11.94	5.97 ± 1.87c
Sphericity	0.77–1.07	0.91 ± 0.05b	0.73–0.94	0.85 ± 0.05a	0.77–1.08	0.89 ± 0.06b
Aspect ratio	0.71–1.00	0.91 ± 0.07b	0.74–0.98	0.85 ± 0.07a	0.69–1.13	0.87 ± 0.09a
Volume geometri (x10−9 m3)	1939.04–6927.93	4472.31 ± 1064.09b	1617.66–5319.75	3766.92 ± 909.92a	1796.15–10,747.54	5335.74 ± 1683.61c
True volume (mm3)	4250–5000	4705 ± 295b	2900–4190	3339 ± 414a	4750–7000	5325 ± 646c
Density* (g/mL)	0.98–1.08	1.05 ± 0.03a	0.93–1.1	1.04 ± 0.07a	0.91–1.19	1.05 ± 0.07a
Pericarp* (%)	36.59–66.27	54.8 ± 7.68b	40.45–57.23	46.39 ± 4.69a	48.14–59.69	55.58 ± 3.24b

Values are expressed as minimum–maximum ranges and means ± standard deviations. n = 100 per parameter, except (*) n = 30. Means followed by different lowercase superscript letters in the same row indicate significant differences in geographical origins based on a One-Way ANOVA and Duncan's test (P < 0.05).

Liberica coffee cherries displayed a generally round-to- oval morphology. During ripening, the cherry color transitioned from green to yellow and orange, and finally to red. Color evaluation shows that the dominant cherry color in all regions is brick red, as shown in Fig. 2, although additional color variations were also observed, including strawberry red, dark red, dark golden red, potter's clay, golden orange, and sepia brown. These variations occur across every region and are not exclusive to any particular region.

Fig. 2.

Liberica coffee from different origins.

3.2. Physical and chemical properties of green beans from different origins

The physical properties of Liberica green coffee beans showed clear, consistent variation primarily driven by geographical origin, with postharvest processing as a secondary modifying factor (Table 2). Across all samples, beans from Meranti consistently exhibited the largest dimensions, volume, mass, and density, followed by Jambi and Kayong. These size-related attributes were significantly influenced by origin (p < 0.05), while the processing effects are less pronounced and mainly visible in specific parameters such as density and color, as shown by the visual differences between Liberica green coffee beans from various origins in Fig. 2.

Table 2.

Physical and chemical properties of green beans from different origins.

Properties	Jambi		Kayong		Meranti		p value of Interaction Origin x Postharvest
	Natural	Honey	Natural	Honey	Natural	Honey
A. Physical properties
Length (mm)	11.55 ± 1.37bA	11.44 ± 1.27bA	9.98 ± 0.93aA	10.4 ± 1.04aB	11.26 ± 1.14bA	11.41 ± 1.30bA	–
Width (mm)	7.69 ± 0.62bA	7.88 ± 0.54bA	7.54 ± 0.55aA	7.75 ± 0.61aA	7.82 ± 0.66bA	7.95b ± 0.84bA	–
Thick (mm)	4.39 ± 0.48aA	4.57 ± 0.64aA	4.53 ± 0.43bA	4.48 ± 0.43aA	4.86 ± 0.65cA	4.76b ± 0.43bA	–
Volume (mm3)	205.59 ± 44.41bA	215.02 ± 40.84bA	179.45 ± 32.67aA	187.02 ± 39.2aA	226.54 ± 55.76bA	227.68 ± 48.85bA	–
Mass of 100 bean (g)	23.71 ± 0.67b	24.47 ± 0.81c	19.43 ± 0.51a	19.51 ± 0.52a	26.36 ± 0.73e	25.02 ± 0.78d	p < 0.0001
Bulk density (kg/m3)	637.8 ± 7.7c	609.1 ± 4.9b	606.2 ± 5.1b	576.1 ± 6.8a	698.4 ± 2.6e	690.8 ± 7.4d	p < 0.0001
Color L*	68.02 ± 1.08a	70.87 ± 0.79c	73.34 ± 1.66d	74.63 ± 0.19e	69.47 ± 0.31b	71.79 ± 0.24c	p = 0.016
a*	2.11 ± 0.09d	2.22 ± 0.11d	0.62 ± 0.14b	0.12 ± 0.05a	1.12 ± 0.08c	1.24 ± 0.20c	p < 0.0001
b*	23.49 ± 0.31b	23.60 ± 0.37b	24.27 ± 0.49c	22.95 ± 0.13a	22.65 ± 0.38a	23.00 ± 0.24a	p < 0.0001
B. Chemical properties
pH	5.84 ± 0.00c	5.61 ± 0.01a	5.95 ± 0.01f	5.75 ± 0.00b	5.89 ± 0.00d	5.93 ± 0.01e	p < 0.0001
TDS (%)	1,91 ± 0.02aA	1,83 ± 0.04aA	1,87 ± 0.05aA	1,83 ± 0.03aA	1,93 ± 0.02aA	1.88 ± 0.01aA	p = 0.750
Proximate
Moisture (%)	9.67 ± 0.28bA	10.62 ± 0.04bB	10.9 ± 0.2cA	11.17 ± 0.16cB	8.42 ± 0.16aA	9.41 ± 0.62aB	ns
Ash (% db)	4.29 ± 0.05b	4.3 ± 0.01b	4.03 ± 0.06a	3.93 ± 0.01a	4.47 ± 0.02c	4.67 ± 0.01d	p = 0.001
Protein (% db)	17.46 ± 0.01bB	16.31 ± 0.1bA	14.85 ± 0.7aB	14.75 ± 0.22aA	17.70 ± 0.17cB	17.27 ± 0.24cA	ns
Carbohydrate (%db)	63.56 ± 0.42aA	64.77 ± 0.11aB	64.27 ± 1.31bA	65.23 ± 0.29bB	62.19 ± 0.65aA	63.95 ± 0.47aB	ns
Fiber (%db)	9.23 ± 0.27bA	9.11 ± 0.24bA	8.71 ± 0.66aA	7.95 ± 0.37aA	8.82 ± 0.93abA	8.31 ± 1.23abA	ns
Lipid (%db)	5.38 ± 0.08aB	5.22 ± 0.06aA	6.23 ± 0.61bB	5.51 ± 0.02bA	6.06 ± 0.08bB	5.72 ± 0.03bA	ns
Mineral (mg/kg)
K	15,377.88 ± 93.82abA	16,583.99 ± 1444.20abA	16,429.59 ± 309.43aA	15,268.19 ± 1927.56aA	17,951.21 ± 725.30bA	17,218.35 ± 552.19bA	ns
Mg	1851.73 ± 37.77b	1937.73 ± 80.02b	1815.63 ± 35.39ab	1631.74 ± 113.17a	1773.42 ± 94.87ab	1792.14 ± 29.99ab	p = 0.023
Ca	1845.79 ± 45.77bA	1406.70 ± 42.54bA	1052.49 ± 25.03aA	1062.57 ± 267.68aA	1757.08 ± 294.22bA	1565.12 ± 351.14bA	ns
Na	97.60 ± 10.92a	252.07 ± 66.03b	131.81 ± 10.44a	173.88 ± 44.80ab	138.92 ± 52.06a	124.68 ± 18.62a	p = 0.010
Mn	9.68 ± 0.18c	7.09 ± 0.64ab	5.76 ± 0.38a	6.84 ± 0.39ab	7.24 ± 0.71ab	7.10 ± 0.64b	p < 0.0001
Fatty acid (% of total fatty acid)
Linoleic acid	39.00 ± 0.24aB	39.23 ± 0.3aA	40.04 ± 0.11bA	40.45 ± 0.11bA	39.15 ± 0.21aB	39.62 ± 0.43aA	ns
Palmitic acid	35.16 ± 0.20bc	34.70 ± 0.36ab	35.06 ± 0.09abc	35.31 ± 0.09c	34.48 ± 0.23a	34.64 ± 0.12b	p = 0.027
Oleic acid	10.00 ± 0.14aB	10.21 ± 0.03aA	10.70 ± 0.20bB	11.01 ± 0.20bA	10.00 ± 0.14aB	10.21 ± 0.03aA	ns
Stearic acid	8.07 ± 0.27bA	8.16 ± 0.03bA	7.88 ± 0.29bA	8.38 ± 0.29bA	7.68 ± 0.14aA	7.78 ± 0.29aA	ns
Linolenic acid	0.85 ± 0.06aB	1.08 ± 0.03aA	1.12 ± 0.06bB	1.22 ± 0.06bA	0.95 ± 0.10aB	1.06 ± 0.08aA	ns
Miristic acid	0.29 ± 0.02bA	0.33 ± 0.03bA	0.38 ± 0.03cA	0.39 ± 0.03cA	0.22 ± 0.03aA	0.23 ± 0.06aA	ns
Lauric acid	0.04 ± 0.03aA	0.03 ± 0.01aA	0.09 ± 0.03bA	0.06 ± 0.04bA	0.03 ± 0.02aA	0.03 ± 0.01aA	ns

Values are presented as mean ± standard deviation (n = 100 for dimensions; n = 30 for mass and bulk density; n = 12 for color; n = 3 for all other parameters). Physical dimension parameters (length, width, thickness, and volume) were analyzed using the Kruskal–Wallis test followed by Dunn's post hoc test (among origins) and the Mann–Whitney U test (among postharvest methods) due to non-normal distribution. All other parameters were analyzed using two-way ANOVA followed by Tukey's HSD test (p < 0.05). Different lowercase and uppercase superscript letters indicate significant differences among origins and postharvest methods, respectively. For parameters showing a significant interaction, superscript letters denote differences among combined origin × postharvest treatment means (Tukey's HSD, p < 0.05). Interaction analysis was not applicable (−). ns: not significant.

Natural and honey processing produced similar dimensional characteristics; however, honey-processed beans tended to exhibit slightly higher uniformity and brightness. Density values varied significantly by origin and processing (p < 0.05), with Meranti natural-processed beans showing the highest density and Kayong honey-processed beans the lowest. Color parameters (L*, a*, b*) were significantly affected by both factors and their interaction (p < 0.05), indicating that the impact of origin on bean appearance depended on the processing method applied.

The chemical composition of Liberica green coffee beans showed clear regional characteristics, while several components were further shaped by postharvest processing (Table 2). Geographical origin played a major role in determining pH, ash content, mineral profile, macronutrient composition (protein, carbohydrates, and lipids), and fatty acid patterns (p < 0.05). Postharvest processing, on the other hand, mainly influenced moisture content, lipid levels, linoleic acid, oleic acid, and linolenic acid (p < 0.05). Although the pH values of Liberica coffee beans fell within a relatively narrow range (5.61–5.95), significant differences among geographical origins were observed, potentially reflecting variations in environmental growing conditions. Natural processing generally resulted in lower moisture content and higher lipid accumulation than honey processing. In contrast, ash content and most mineral elements were primarily influenced by geographical origin, indicating that their composition is more closely associated with soil characteristics and environmental conditions than with postharvest processing. A summary of the p-values from the two-way ANOVA is presented in Supplementary 2 (Table S4).

3.3. Bioactive compounds in coffee beans

3.3.1. Caffeoylquinic acids (CQAs)

The Caffeoylquinic acids (CQAs) profile of Liberica green coffee beans shows significant regional differences, while postharvest processing mainly affects the relative distribution of individual CQA isomers (Fig. 3A). Total CQA content ranged from 3.59 to 5.38 g/100 g db, with Meranti beans showing the highest level (5.38 g/100 g db), significantly higher than Jambi and Kayong (P < 0.05), consistent with previous reports for Indonesian Liberica (Cheng et al., 2016; Hanifah et al., 2025). Caffeoylquinic acid (CQA) is the dominant fraction, with 5-CQA consistently representing the largest proportion of total CQA, while 3-CQA and 4-CQA are present at lower levels and are more sensitive to postharvest processing. Honey processing significantly reduced 3-CQA and 4-CQA compared to natural processing (P < 0.05), while 5-CQA remained stable (Wang et al., 2025). The observed CQA distribution is consistent with previous findings showing 5-CQA as the most abundant isomer, typically accounting for 60–70% of total mono-CQA, followed by 3-CQA and 4-CQA (Munyendo et al., 2021; Nguyen et al., 2024; Zhu et al., 2021). A two-way ANOVA analysis confirmed that geographical origin is the primary factor determining total CQA accumulation, while selective postharvest processing modulates the less stable isomers without significantly altering total CQA content.

Fig. 3.

Bioactive compounds of Liberica coffee from different origins: (A) Caffeoylquinic acids (CQAs), (B) Alkaloids, (C) Diterpenes. Values are presented as mean ± standard deviation (n = 3). Different lowercase and uppercase superscript letters indicate significant differences among geographical origins and processing methods, respectively, based on Tukey's HSD test (p < 0.05). No significant interaction was observed for any CQA parameter (two-way ANOVA, p > 0.05); therefore, only main effects were considered.

3.3.2. Alkaloids

The alkaloid profile of Liberica green coffee beans varies depending on geographical origin and postharvest processing (Fig. 3B).Total alkaloid content ranged from 1.25 to 1.53 g/100 g db, with significant differences between samples from Jambi, Meranti, and Kayong (p < 0.05), highlighting the strong influence of origin (Mehari et al., 2016). Trigonelline, the primary non-caffeine alkaloid, ranged from 0.27 to 0.34 g/100 g db, lower than previous reports from Jambi (0.69–0.89 g/100 g db) (Hanifah et al., 2025) and consistent with the generally lower levels in C. liberica compared to C. arabica and C. canephora (Campa et al., 2004). Regional variations likely reflect differences in growing conditions and fruit ripeness (Hu et al., 2020; Koshiro et al., 2007). Postharvest processing significantly affected trigonelline retention (p < 0.05), with naturally processed beans retaining more trigonelline than honey-processed beans. Trigonelline is a pyridine alkaloid with high hydrophilic properties (Konstantinidis et al., 2023). The lower trigonelline content observed in the honey process may also be attributed to direct heat exposure of depulped beans during drying, which may accelerate its degradation into nicotinic acid and nicotinamide (Yulianti et al., 2022), although this mechanism was not directly measured in the present study. Caffeine content also varied significantly depending on origin and processing method (0.91–1.26 g/100 g db; p < 0.05), with honey-processed beans generally exhibiting higher caffeine levels than naturally processed beans across all producing regions. Theobromine, another methylxanthine occasionally found in coffee, was not detected in any of the samples, as its content was below the limit of detection (LoD) in all cases. This finding is consistent with Mehari et al. (2016), who similarly reported theobromine content below the LoD in Arabica green coffee beans sourced from various growing regions in Ethiopia.

3.3.3. Diterpenes

The diterpenes composition of Liberica green coffee beans is dominated by cafestol, while kahweol is present at much lower levels (Fig. 3C). Kahweol concentrations ranged from 56.91 to 85.93 mg/100 g db, while cafestol ranged from 144.07 to 218.77 mg/100 g db, resulting in a low kahweol to cafestol (K/C) ratio (0.4–0.44). This profile is consistent with previous reports for C. liberica var. liberica and var. dewevrei, where cafestol dominates over kahweol.

Geographical origin significantly influenced kahweol and cafestol content (p < 0.05), with natural Kayong coffee showing the highest diterpene levels, while Jambi coffee processed using the honey method had the lowest kahweol content. In contrast, postharvest processing had a limited effect and was only significant for kahweol (p < 0,05). The significant interaction between region and processing method indicates that regional agroecological conditions play a dominant role in determining diterpene levels, with processing acting as a secondary modifying factor, consistent with previous studies on the influence of genotype and environment on coffee diterpenes (Francisco et al., 2021; Kitzberger et al., 2013). Representative HPLC chromatograms of the bioactive compounds are presented in Supplementary 2 (Figs. S1–S4).

3.4. In vitro antioxidant capacity

The antioxidant capacity of Liberica green coffee beans varies depending on geographical origin and postharvest processing, depending on the analytical method used (Fig. 4). Based on the DPPH test, the IC₅₀ values ranged from 101 to 124 μg/mL, indicating moderate antioxidant activity compared to pure reference antioxidants such as Trolox and vitamin C. Lower IC₅₀ values are associated with stronger radical scavenging capacity. Significant differences in antioxidant activity based on DPPH were observed among samples (p < 0.05). Coffees with higher caffeoylquinic acids content, particularly 5-CQA, exhibited lower IC₅₀ values and therefore more potent antioxidant activity. This pattern supports previous findings that identified caffeoylquinic acid as a major contributor to free radical scavenging in coffee and highlights the influence of environmental conditions and postharvest processing on antioxidant performance (Fitri et al., 2025; Liang & Kitts, 2014; Yulianti et al., 2024).

Fig. 4.

Antioxidant Capacity Using the DPPH and FRAP Methods. J = Jambi, K = Kayong, M = Meranti; N = natural, H = honey. The different lowercase letters a and b represent significant differences in origins. The different upper-case letters A and B represent significant differences in postharvest. Results were calculated using the Tukey test (p < 0.05) after two-way ANOVA.

In contrast, antioxidant capacity measured by the FRAP method showed a relatively uniform profile across all samples, with values ranging from 8.0 to 8.5 g TEAC/100 g db. and no significant differences between regions or processing methods (p > 0.05). This consistency indicates that the overall reducing power of Liberica coffee remains stable despite the variations detected by the DPPH assay. Similar differences between DPPH and FRAP responses have been reported in previous studies on Liberica and other coffee species (Hanifah et al., 2022), highlighting the complexity of coffee's antioxidant behavior.

3.5. Multivariate analysis

Principal component analysis (PCA) was applied to evaluate patterns in the chemical properties and composition of bioactive compounds in Liberica green coffee beans across different regions and postharvest methods (Fig. 5). The PCA model (Fig. 5A), constructed using data from Table 2 and the concentrations of CQA, alkaloids, and diterpenes, explained 60.50% of the total variance, with PC1 (F1) contributing 45.38% and PC2 (F2) 15.12%. The PCA score plot shows a clear separation of samples into three distinct clusters corresponding to their geographical origin. Samples from Meranti (MH and MN) are located in the positive F1 and high-positive F2 regions, samples from Kayong (KH and KN) are in the negative F1 and positive F2 regions. In contrast, samples from Jambi (JH and JN) are grouped in the low-to-moderate positive F1 and negative F2 regions. This grouping indicates different chemical profiles among the three production regions. The factor loadings for F1 and F2 are presented in Supplementary 2 (Table S2).

Fig. 5.

(A) Scatter plot PCA separation of Jambi (J), Kayong (K), and Meranti (M) coffee beans, N = natural, H = honey. (B) The biplot shows the relationships between bioactive compound content (CQA, alkaloids, diterpenes) and antioxidant capacity (DPPH and FRAP).

The Jambi samples were closer to the Meranti samples, reflecting similarities in their caffeoylquinic acid profiles. In contrast, the Kayong samples were clearly separated along PC1 and associated with higher diterpene content, particularly cafestol and kahweol. Differences between honey-processed and natural coffees can be observed within each regional cluster, but are less pronounced than the separation driven by geographic origin.

The PCA biplot (Fig. 5B) depicts the relationship between major bioactive compounds and antioxidant capacity as measured by DPPH IC₅₀ and FRAP. Total CQA, individual CQA isomers (3-CQA, 4-CQA, and 5-CQA), and trigonelline are located opposite the IC₅₀ vector, indicating an inverse relationship between these compounds and IC₅₀ values. Samples located close to the CQA and trigonelline vectors, especially Meranti samples, showed lower IC₅₀ values. The FRAP vector showed a strong positive alignment with total CQA, individual CQA isomers, and trigonelline, indicating a strong relationship among these compounds and total reduction capacity. In contrast, the cafestol and kahweol vectors were located further from the IC₅₀ and FRAP vectors, suggesting a weaker relationship with the antioxidant capacity measured by this assay. Alkaloids such as caffeine show a more orthogonal orientation relative to IC₅₀ and FRAP, indicating a different pattern of contribution compared to caffeoylquinic acid. PCA results show that the chemical composition profile effectively distinguishes Liberica coffee samples by geographical origin and highlights the dominant contributions of caffeoylquinic acids and trigonelline to antioxidant related variation. The factor loadings for F1 and F2 are presented in Supplementary 2 (Table S3).

Heat map analysis (Fig. 6) showed that Geographical Indication (GI) coffee samples form clear groupings based on similarities in their chemical profiles, indicating that geographical origin and postharvest methods play important roles in shaping each sample's chemical characteristics. These variations are mainly influenced by differences in diterpene content (cafestol and kahweol), caffeoylquinic acids (CQAs), fatty acid composition, and mineral content, which have been identified as the main markers for distinguishing between Liberica coffee samples.

Fig. 6.

Heatmap analysis of Liberica coffee from different origins.

4. Discussion

4.1. The influence of geographical origin on the physical properties of Liberica coffee cherries

The differences in physical characteristics observed in Liberica coffee cherries from Jambi, Meranti, and Kayong, as shown in Table 1, highlight the important role of geographical origin in shaping fruit morphology. Variations in the size, mass, volume, and shape of coffee beans are generally related to agroecological factors, including soil fertility, water availability, temperature, and radiation intensity, which collectively influence cell expansion and dry matter accumulation during fruit development. The consistently larger size and mass of Meranti fruits indicate more favorable growing conditions, which support longer fruit filling and higher biomass accumulation, a phenomenon previously reported in Liberica grown in optimal edaphic climatic environments (Ismail et al., 2014; Martono et al., 2020).

These three GI regions have distinct key environmental characteristics, as outlined in their respective GI Specification Manuals. Soil types also vary significantly. In Tungkal Jambi and the Meranti Islands, the soils are predominantly peat (Organosol/Gleihumus, pH < 4), while in North Kayong, they are predominantly clay-textured (pH 3.195). All three regions cultivate Liberica on very low-lying coastal plains, specifically at 2–5 m above sea level in the Meranti Islands, 1–10 m above sea level in Tungkal Jambi, and 5–12 m above sea level in Kayong Utara. The respective annual average temperatures are 29.9 °C, 32.3 °C, and 29.21 °C; the average relative humidity is 83.3%, 91.8%, and 66%; and annual rainfall ranges from 2036–2,523 mm, 2098–3778 mm, and 2330.5 mm for Tungkal Jambi, North Kayong, and the Meranti Islands, respectively (MPIG, 2014; MPIGKLKU, 2022; MPKLRM, 2015). These conditions indicate that Liberica coffee thrives in lowland peatland environments with warm climates, which may explain the physicochemical and bioactive variations observed across regions.

The size and shape of coffee cherries affect the postharvest process. The large size of Meranti coffee cherries requires adjustments to processing equipment, particularly pulpers, to ensure the pulping unit geometry and pulping gap are appropriate to prevent bean damage (Bizimungu et al., 2024). In addition to affecting the design of the pulping machine, the larger cherry size of Meranti Liberica coffee is also associated with a higher pericarp percentage. Liberica fruit also has a thicker outer skin (epicarp) and pulp (mesocarp), with a lower percentage of endocarp (parchment), integument (silver skin), and endosperm (beans) compared to other coffee species (Ismail et al., 2014). In this study, the highest skin percentage was found in Jambi Liberica coffee at 55.58%. The presence of this coffee cherry skin has a significant effect on the drying time of coffee beans. The pericarp of coffee beans shows a high level of resistance to dehydration (De Almeida Dias et al., 2020). The layers of the pericarp slow water evaporation, so drying coffee using the natural method takes longer than with the honey and full-washed methods (Yulianti et al., 2022). The drying process for Liberica coffee using the natural method takes between 30 and 40 days, which is longer than Robusta and Arabica coffee, which only takes about 15 days (Aswathi et al., 2023; Hanifah et al., 2022; Yulianti et al., 2022). The physical properties of Liberica coffee cherries are a complex interaction between their genetic background and the environment in which they are grown. These factors determine properties such as size, shape, mass, and density, which in turn influence subsequent coffee processing.

4.2. The influence of geographical origin on the physicochemical properties of Liberica coffee beans

Physical Properties are among the main factors in assessing the quality of coffee beans, as they affect their sensory complexity. Physical analysis of coffee beans is also important for geographical traceability and classification in the international coffee trade (Urugo et al., 2024). The measurement of Physical Properties of coffee beans in the coffee industry aims to effectively distinguish defective coffee beans from non-defective ones, which is usually done through screening based on bean size, volume, density, and color (Mendonça et al., 2009). The physical properties of Liberica coffee beans are detailed in Table 2 and show significant variation influenced by regions and postharvest processing methods (p < 0.05), with Meranti Liberica coffee beans being the largest. This means that Liberica coffee requires different sieving and grading equipment from that used for Arabica and Robusta (Ismail et al., 2014). Soil characteristics, particularly pH, Mn, organic matter concentration, and higher sand content, affect coffee bean size (Abebe et al., 2020). Furthermore, Oliveira et al. (2015) reported Mn as a key elemental marker for discriminating the geographical origin of coffee, highlighting its importance not only in coffee growth but also in origin authentication.

Green coffee beans have a complex chemical composition, that significantly affects their quality (Cardoso et al., 2023). The chemical composition of Liberica green coffee beans presented in Table 2 reflects a combination of inherent regional characteristics and postharvest modifications. Geographical origin dominates mineral composition, ash content, and macronutrient distribution, in line with the strong influence of soil composition and nutrient availability on element uptake and metabolic allocation in coffee plants (Francisco et al., 2021). The relatively small impact of processing on these parameters suggests that they are largely determined in the pre-harvest phase. In contrast, postharvest processing plays a more prominent role in modifying parameters related to solubility, lipid retention, and moisture dynamics. The differences between natural and honey processing are likely influenced by variations in fermentation intensity, mucilage presence, and drying rate, which affect enzyme activity and compound diffusion. Similar trends have been reported in other coffee species, where semi-dry or honey processing alters physicochemical properties without compromising regional characteristics (Aswathi et al., 2023; Wulandari et al., 2021).

4.3. Influence of geographic origin and selective processing on bioactive compounds

Caffeoylquinic acids (CQAs) is one of the bioactive compounds most responsive to environmental and postharvest factors as can be observed in Fig. 3B. The significant regional differences in total CQA content observed in this study are consistent with previous findings linking CGA biosynthesis to plant stress responses, including UV exposure, temperature fluctuations, and water availability (Cheng et al., 2016; Nguyen et al., 2024). Genetic diversity within Coffea liberica populations further reinforces this variability, as different local varieties exhibit different metabolic capacities. Selective postharvest processing affects individual CQA isomers rather than total CQA content, indicating different stability during fermentation and drying.

Changes in CQA during postharvest and drying are driven by a combination of physical, biochemical, and physiological processes (Bastian et al., 2021). Differences may also be influenced by the dynamics of microbial fermentation across processing methods. In the honey process, sugar-rich mucilage allows fermentation to occur simultaneously with drying, which has been reported to enable microorganisms and enzymes to selectively metabolize and degrade CQA (Aswathi et al., 2024; Wulandari et al., 2021), however, microbial activity and enzymatic degradation were not directly assessed in the present study, and this explanation remains a plausible hypothesis based on comparison with the cited literature rather than an established mechanism for the samples analyzed here. The 3-CQA and 4-CQA isomers tend to be hydrolyzed more easily into caffeic acid or transformed to their isomer than 5-CQA, which is relatively more stable (Aguirre Santos et al., 2018; Bel-Rhlid et al., 2013; Wang et al., 2025).

The alkaloid profile (Fig. 3B), particularly caffeine and trigonelline, demonstrated sensitivity to both regional and processing factors. Lower trigonelline levels relative to some previous Liberica studies may reflect varietal or harvest maturity differences, as trigonelline accumulation peaks a more advanced stage of fruit maturity (Hu et al., 2020; Koshiro et al., 2007). For caffeine, the range of 0.90–1.26 g/100 g db, with the lowest value recorded in Kayong Natural and the highest in Jambi Honey, is likely influenced by soil characteristics, as lower caffeine levels have been associated with high organic matter, total nitrogen, and sand content, while higher levels correlate with clay content (Abebe et al., 2020). The consistently higher caffeine levels in honey-processed beans across all origins may be linked to microbial activity during fermentation, which can convert caffeine into other methylxanthines such as 1-methylxanthine and 7-methylxanthine through the metabolism of microorganisms, including Lactobacillus casei, Leuconostoc mesenteroides, Rhizopus oryzae, and Saccharomyces cerevisiae (Purwoko et al., 2023). The higher caffeine levels observed in coffee beans processed using the honey method in this study are consistent with reports indicating that the shorter drying period in honey processing, compared to natural processing, may limit caffeine degradation (Yulianti et al., 2022). However, an opposite pattern has been reported for green Robusta beans, in which honey processing resulted in lower caffeine content (1.54% db) than natural processing (1.81% db) (Wulandari et al., 2021). These conflicting findings indicate that caffeine response to postharvest processing is not uniform and is likely influenced by species, variety, geographical origin, fermentation duration, and drying conditions.

The diterpene composition, as observed in Fig. 3C, characterized by a higher cafestol content than kahweol, confirms the distinctive biochemical characteristics of Liberica coffee and supports its differentiation from Arabica and Robusta. The observed kahweol to cafestol ratio (0.4) is still in line with previous reports on C. liberica var. liberica (0.5) (De Roos et al., 1997), while the kahweol to cafestol ratio in arabica was 1.7 (Juwita et al., 2025) and in robusta was 0.02 (De Roos et al., 1997). This reinforces the use of diterpenes as chemical taxonomic markers. Variations in diterpene levels in specific regions likely reflect differences in lipid biosynthesis pathways influenced by genotype and environmental stress. At the same time, the relatively small effect of processing indicates limited diterpene transformation during the postharvest stage (Moeenfard et al., 2015).

The relatively high diterpene content observed in Kayong Liberica coffee is noteworthy, as a recent review by Ren et al. (2019) summarized evidence that coffee diterpenes, particularly cafestol and kahweol, exhibit various biological activities, including antidiabetic, anti-obesity, anticancer, and anti-inflammatory effects. Given the limited information available on the diterpene composition of Liberica coffee compared with Arabica and Robusta, the present findings suggest that Indonesian Liberica green coffee beans, particularly those from Kayong, warrant further investigation as a potential natural source of coffee diterpenes, especially cafestol. Future studies should focus on targeted quantification, purification, and biological activity evaluation to better assess their potential applications in pharmaceutical, nutraceutical, and functional food products. However, these potential applications require further validation before any application- or health-related claims can be made.

4.4. Factors influencing antioxidant potential

Fig. 4 confirms that the antioxidant behavior of Liberica coffee is influenced by the concentration of compounds and antioxidant mechanisms. Stronger discrimination in DPPH than in FRAP highlights the sensitivity of radical scavenging assays to variations in hydrogen-donating compounds, particularly caffeoylquinic acid. Caffeoylquinic acid is known to be a major contributor to antioxidant activity (Fitri et al., 2025). Environmental factors and postharvest practices that facilitate the accumulation of CQAs and other antioxidant compounds also determine free radical scavenging capacity (Yulianti et al., 2024).

Conversely, the relative stability of FRAP values between samples indicates that total electron-donating capacity is maintained despite compositional differences, suggesting the functional resilience of Liberica coffee as an antioxidant source (Liang & Kitts, 2014). These contrasting test responses emphasize that antioxidant capacity cannot be represented by a single method and that shifts in composition may affect specific mechanisms rather than overall redox potential. The difference in response between DPPH and FRAP confirms that the two methods evaluate different antioxidant mechanisms. FRAP measures antioxidant capacity through the electron transfer mechanism (reduction of Fe³⁺-TPTZ to Fe²⁺-TPTZ), while DPPH assesses the ability of compounds to donate hydrogen or electrons in stabilizing DPPH radicals. This difference in mechanisms reflects the diversity of antioxidant compounds that predominantly play a role in each method (Hanifah et al., 2022).

4.5. Multivariate analyses of physicochemical and bioactive compounds

The PCA scatter plot (Fig. 5A) and biplot (Fig. 5B) provide an integrated view of how physical, chemical, and functional attributes collectively differentiate Liberica coffee based on its origin. Apparent clustering based on region confirms that geographical factors have a dominant influence on the overall quality profile, while processing acts as a secondary modifier. CQA and trigonelline show a strong correlation with DPPH but only a weak correlation with FRAP, suggesting a greater influence on radical scavenging capacity than on ferric reducing power. At the same time, diterpenes and caffeine appear to contribute through different or indirect mechanisms. This multivariate pattern supports the use of chemical fingerprinting as an indicator of geographical origin and functional quality in Liberica coffee, in line with previous chemometric studies on coffee tracking (Babova et al., 2016; Tieghi et al., 2024). The results of the heat map analysis (Fig. 6) reinforce the PCA findings by highlighting the biochemical basis of geographic differentiation. Taken together, these compounds support the potential development of a unique chemical fingerprint for the geographical origin differentiation and development of Liberica coffee products.

This study has several limitations. Coffee cherries from each geographical origin were represented by a composite sample collected from one GI-certified farmer group selected through purposive sampling, while the numbers of individual farms and trees contributing to each composite sample were not recorded. In addition, the natural and honey processing treatments were each represented by coffee cherries collected from a single harvest season. Consequently, the present study does not capture seasonal variation (e.g., differences in rainfall, temperature, and fruit maturity) that may influence postharvest processing outcomes and the resulting physicochemical and bioactive compound profiles. Future studies should expand the sampling design by including multiple farmer groups, farms, trees, and independent processing batches within each geographical origin to improve the robustness and representativeness of geographical comparisons. Furthermore, drying temperature was not systematically recorded at regular intervals, limiting the ability to fully relate variations in bioactive compounds to specific processing conditions. Future studies should therefore incorporate standardized monitoring of drying conditions to enable more controlled comparisons. Finally, this study focused only on the major bioactive compounds in coffee; therefore, future studies are recommended to employ LC-MS and GC–MS analyses to obtain a more comprehensive characterization of non-volatile and volatile compounds, respectively.

5. Conclusion

Liberica coffee from Kayong, Jambi, and Meranti exhibited distinct physicochemical characteristics, bioactive compound profiles, and antioxidant capacities, reflecting the combined influence of geographical origin and postharvest processing within the scope of the present dataset. Liberica coffee from Kayong exhibited the most distinctive chemical profile, with lower caffeine content and higher diterpene content, suggesting potential differences in stimulant-related characteristics compared with the other geographical origins. Liberica coffee from Jambi and Meranti exhibited similar physicochemical characteristics, which may reflect shared compositional features despite representing different certified geographical indications. Natural and honey processing significantly affected the levels of caffeoylquinic acids, alkaloids, diterpenes, and antioxidant activity, highlighting the sensitivity of Liberica coffee to postharvest interventions. Overall, these findings provide valuable baseline evidence that geographical origin may contribute to the differentiation of Indonesian Liberica coffee, while indicating that further validation using larger sample sets and complementary untargeted metabolomic approaches is required before establishing definitive geographical authentication or chemical fingerprinting markers.

CRediT authorship contribution statement

A. Ita Juwita: Writing – original draft, Visualization, Methodology, Formal analysis, Data curation. Dian Herawati: Writing – review & editing, Validation, Supervision, Software, Resources, Methodology, Investigation, Data curation, Conceptualization. Didah Nur Faridah: Writing – review & editing, Validation, Supervision, Methodology, Investigation, Data curation, Conceptualization. Widiastuti Setyaningsih: Writing – review & editing, Resources, Project administration, Funding acquisition, Conceptualization. Ketut Wikantika: Writing – review & editing, Resources, Project administration, Funding acquisition, Conceptualization. Nuri Andarwulan: Writing – review & editing, Validation, Supervision, Resources, Project administration, Methodology, Funding acquisition, Data curation, Conceptualization.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Data availability

Data will be made available on request.