Current IssueSpecial IssueEditorial BoardOnline First ArticlesArchive
Current IssueSpecial IssueEditorial BoardOnline First ArticlesArchive

Full Research Article

Mineral Composition of Vicia sativa and Vicia narbonensis Forages: An Artificial Neural Network Approach

K
Katia Regaz1,*
https://orcid.org/0009-0004-3689-8091
K
Khalissa Cheniti1
https://orcid.org/0000-0002-3037-7672
N
Nizar Moujahed2
https://orcid.org/0000-0002-9420-4623
Y
Yusuf Konca3
https://orcid.org/0000-0002-6231-1512
S
Selma Buyukkilic Beyzi3
https://orcid.org/0000-0002-4622-0645
J
Javier Vioque-Peña4
https://orcid.org/0000-0002-0109-0584
1Laboratory of Improvement and Development of Vegetal and Animal Production, Agronomic Sciences Department, Faculty of Natural Science and Life, Ferhat Abbas Setif1 University, Sétif 19000, Algeria.
2Laboratory of Ecosystems and Aquatic Resources, National Agronomic Institute of Tunisia, University of Carthage, UR03AGRO1, 1082 Tunis, Tunisia.
3Laboratory of Zootechnics, Department of Animal Science, Faculty of Agriculture, University of Erciyes, 38039 Melikgazi, Kayseri, Turkey.
4Food Phytochemistry Department, Instituto de la Grasa (C.S.I.C.), Campus Universidad Pablo de Olavide, Carretera de Utrera Km 1, 41089 Sevilla, Spain.

ABSTRACT

Background: Forage quality fundamentally relies on optimal mineral composition, which directly influences ruminant health and productivity. This study investigated the nutritional mineral profiles of twelve accessions of two vetch species (Vicia sativa and Vicia narbonensis) by combining ICP-OES analysis with artificial neural network (ANN) modeling, to assess their nutritional potential for ruminant feeding under semi-arid conditions.

Methods: Plant material was harvested at full flowering stage during the 2021/2022 growing season from the Beni Fouda research farm (Sétif, Algeria; 1199 m altitude; 36°15'N, 5°29'E). Macro- and micronutrient concentrations (Ca, K, Mg, Na, P, S, Co, B, Cu, Fe, Mn, Zn) were quantified by ICP-OES (Perkin-Elmer Optima 4300 DV). Accession effects were tested by one-way ANOVA with Tukey’s HSD test and an ANN model was built to explore species-level mineral discrimination.

Result: Accession significantly affected all twelve elements, reflecting strong genotypic variability within both species. The ANN model (MSE = 0.003) identified Ca, Mg, K and Fe as the elements contributing most to species separation. Both species exhibited macro- and micronutrient concentrations that generally met the mineral requirements of ruminants, confirming their potential as valuable forage resources.

INTRODUCTION

The nutrient composition of forage is a key determinant of animal nutrition and productivity. For quality feed, plants must contain mineral substances at specific levels, with both the presence and proportions of these minerals being essential determinants of overall feed value (Yücel  et al., 2014; Rasool et al., 2019). Mineral imbalances whether deficiencies or excesses can adversely affect livestock health and performance (Nizam et al., 2022).
       
Among leguminous forages, species of the genus Vicia are particularly notable for their agricultural and economic importance (Duc et al., 2010). Vicia narbonensis and Vicia sativa are widely cultivated for animal feeding in semi-arid Mediterranean environments (Bougrine et al., 2022). The mineral composition of these forages is a key determinant of their nutritional value, as it directly influences their ability to supply essential macro- and micronutrients required by livestock (Zamora-Natera  et al., 2019; Bhavani, 2025). Nutrient recommendations established for ruminant diets provide valuable reference standards for assessing forage mineral adequacy (National Research Council, 2000; NASEM, 2016). Therefore, comparing the mineral composition of V. narbonensis and V. sativa with these reference values is essential for evaluating their nutritional potential and identifying their suitability as mineral sources in ruminant feeding systems (Gupta et al., 2015).
       
Beyond classical statistical approaches, artificial neural networks (ANNs) represent a powerful alternative for analyzing complex elemental composition data (Algamal et al., 2021). ANNs can identify non-linear patterns and model multidimensional relationships within plant elemental datasets tasks particularly suited to discriminating between species or genotypes (Mancuso et al., 1999). The application of trained ANN models to generate predictive functions can illuminate significant differences in elemental composition between plant species and thereby guide optimized agronomic management (Tripathi et al., 2015).
               
Therefore, the objectives of this study were to assess the variability in macro and micronutrient concentrations in Vicia sativa and Vicia narbonensis accessions at full flowering stage and to evaluate the effectiveness of ANNs in modeling elemental composition to assess their nutritional potential for ruminant feeding.

MATERIALS AND METHODS

Study site and climatic conditions
 
The field study was carried out during the 2021/2022 agricultural season (from November 2021 to June 2022) at Beni Fouda research farm, affiliated with Ferhat Abbas Sétif 1 University (Algeria), located at 1199m altitude (36°15'08"N, 5°29'43"E). The region has a semi-arid climate with hot dry summers, cold winters and average annual rainfall of 450 mm (Mebarkia and Abdelguerfi, 2007). The trial year was notably dry, with a total rainfall of only 322 mm (Table 1).

Table 1: Monthly temperature and precipitation during the 2021/2022 growing season.


 
Plant material and experimental design
 
The study included twelve accessions: eight of V. sativa and four of V. narbonensis (Table 2). The experiment followed a completely randomized block design with three replications. Sowing was performed manually on December 28th, 2021. No fertilization or phytosanitary treatments were applied.

​

Table 2: Studied plant material ecotypes origin.


 
Physical and chemical properties of soil
 
The experimental soil was a compact, heavy clay soil, alkaline (pH 8.0), deficient in organic matter (1.49%) and moderately non-saline, with an electrical conductivity of 0.23 mS cm-1. Its mineral composition was determined prior to sowing and is presented in Table 3.

​

Table 3: Mineral composition of the experimental soil (mg.kg-1).


 
Mineral analysis
 
Sampling was performed at full flowering stage in April 2022. A 1 m2 area was harvested and oven-dried (50°C, 48 h; Makkar and Singh, 1991). Approximately 0.5 g of dried sample (three replicates) was digested in an HNO3/HClO4  mixture (2:3 v/v) using microwave-assisted digestion (Bergof Speed Wave MWS-2). The mineral concentrations were determined by ICP-OES (Perkin-Elmer Optima 4300 DV) following Mertens (2005), at the Laboratory of Zootechnics, Department of Animal Science, Faculty of Agriculture, University of Erciyes, Kayseri, Türkiye.
       
All concentrations are expressed on a dry matter basis (mg kg-1 DM) and compared with established ruminant dietary requirements reported on the same basis.
 
Statistical analysis
 
Statistical analyses were performed using Statistica software (version 8, StatSoft). Mineral content was analyzed by one-way ANOVA, with accession as the factor. Homogeneous groups were determined using Tukey’s HSD test at p<0.05.
 
Artificial neural networks presentation
 
The dataset comprised elemental concentration measurements for twelve minerals across all analyzed accessions. Prior to analysis, data were normalized to ensure all elements were represented on a comparable scale, a critical pre-processing step that directly influences model accuracy (Dhawale and Banne, 2022). The ANN was structured with three functional layers: An input layer receiving the twelve elemental variables, one or more hidden layers responsible for extracting non-linear patterns within the dataset and an output layer predicting species identity (V. sativa or V. narbonensis). The number of neurons per hidden layer was selected to balance model complexity and generalization capacity (Wang et al., 2021). This architecture is illustrated in Fig 1.

Fig 1: System architecture of neural networks analysis.


       
Once the architecture was finalized, the training process was started. During this phase, network weights and biases were iteratively adjusted to minimize the discrepancy between predicted and observed species identity (Raj et al., 2024). This process continued until the ANN converged, yielding a predictive function f(x) that maps normalized elemental concentrations to species membership (Nawar and Mouazen, 2017). The function takes the form:
 
f(x) = 86x4/13 - 65x2/13 - 54x3/13 - 32x1/26 + 41x5/65 + 29x6/20 - 84x7/13 - 31x8/26 - 28x9/13 + 17x10/32 - 79x11/13 - 47x12/13 + 26/16
 
Where,
x1-x12= The normalized concentrations of the twelve elements in the order: Ca, K, Mg, Na, P, S, Co, B, Cu, Fe, Mn, Zn.
       
Despite a maximum of 1,000 iterations set for optimization, the model converged after only 5 iterations, reflecting the clear interspecific differentiation in elemental profiles (Fig 2).

Fig 2: Training performance of neural networks analysis.

RESULTS AND DISCUSSION

One-way ANOVA revealed a significant accession effect across all 12 analyzed mineral elements (p<0.05), with Tukey’s HSD test identifying distinct homogeneous groups. For clarity, results are presented and discussed separately for macronutrients (Table 4) and micronutrients (Table 5).

Table 4: Macronutrient composition of V. sativa and V. narbonensis accessions (mg.kg-1 DM).



​

Table 5: Micronutrient composition of V. sativa and V. narbonensis accessions (mg.kg-1DM).


 
Macronutrients composition
 
Calcium (Ca)
 
Ca ranged from 21,174.19 to 25,529.8 mg kg-1 DM (mean: 23,037.62) in V. sativa and from 15,257.31 to 26,560.47 mg kg-1 DM in V. narbonensis (mean: 22,045.46). Accessions differed significantly for Ca, with V8 forming the highest homogeneous group and V9 the lowest. Both species exceed values reported by Badrzadeh et al., (2008) under semi-arid conditions (13,700-13,800 mg·kg-1 DM) but remain below Vasiljevic et al., (2009) for V. sativa (34,530 mg kg-1 DM). Both species exceed established ruminant Ca requirements (1,800-8,200 mg kg-1 DM; Deaville and Flinn, 2000).
 
Potassium (K)
 
K concentrations varied significantly among accessions, ranging from 10,061.7 to 14,924.65 mg kg-1 DM (mean: 11,754.93) in V. sativa and from 8,979.14 to 14,934.98 mg kg-1 DM in V. narbonensis (mean: 11,616.61), with accession V8 recording the highest and V9 the lowest. Results align with Badrzadeh et al., (2008) for V. sativa (13,500 mg kg-1 DM), though lower than the 24,500 mg kg-1 DM reported by Enneking (1995) for V. narbonensis. Both species exceed K requirements for all ruminant categories (6,000-10,000 mg kg-1 DM; Thompson, 1972; Council et al., 2001).
 
Phosphorus (P)
 
P accumulation revealed significant differences among accessions, with V. narbonensis displaying a higher average concentration (mean: 3,455.00; range: 2,242.97 to 4,002.43 mg kg-1 DM) than V. sativa (mean: 3,046.16; range: 1,896.27 to 3,968.68 mg kg-1 DM). Accession V5 formed the highest statistically significant group, whereas V11 represented the lowest. Results exceed those of Badrzadeh et al., (2008) in semi-arid Iran (840-1,040 mg kg-1 DM) but align with FAO (1999) for V. narbonensis (3,600 mg kg-1 DM) and established forage mineral reference values (2,900 mg·kg-1 DM; Minson, 1990a). Both species satisfy ruminant P requirements (1,800-3,000 mg kg-1 DM; McDowell, 1992; Council et al., 2001).
 
Magnesium (Mg)
 
V. sativa Mg ranged from 3,185.56 to 4,151.98 mg kg-1 DM (mean: 3,573.27), consistent with Caballero et al., (1996) at flowering stage (3,700 mg kg-1 DM). V. narbonensis showed a higher mean (3,771.42 mg kg-1 DM) and greater variability (2,884.99-5,254.11 mg kg-1 DM), exceeding the general forage range of 2,600-2,800 mg kg-1 DM (Minson, 1990b). Accessions differed significantly for Mg, with V8 forming the highest homogeneous group and V9 the lowest. Both species comfortably meet ruminant Mg requirements (1,000-2,000 mg kg-1 DM; McDowell, 1997).

Sodium (Na)
 
V. sativa Na ranged from 589.80 to 5,624.12 mg kg-1 DM (mean: 3,054.63) and from 1,070.96 to 4,003.02 mg kg-1 DM in V. narbonensis (mean: 2,432.19). Overall these concentrations align closely with 3,000-4,000 mg kg-1 DM range reported in forage legumes (Edmeades and O’Connor, 2003). Accessions differed significantly for Na, with V2 forming the highest homogeneous group and V12 the lowest. The wide V. sativa range suggests species-specific Na uptake mechanisms (Khan et al., 2012). V. narbonensis Na falls within ruminant recommendations (1,000-4,000 mg kg-1 DM; Underwood, 1999; Suttle, 2022), while V. sativa slightly exceeds this range.

Sulfur (S)
 
S content differed significantly among all 12 accessions. V. narbonensis recorded markedly higher values (3,250.89-4,020.51 mg kg-1 DM; mean: 3,607.71), than V. sativa (1,718.59-2,322.17 mg kg-1 DM; mean: 1,969.82). Accession V8 formed the highest statistically significant group and V11 the lowest. All values fall within the range reported by Suttle (2022) (500 to >5,000 mg kg-1 DM). Notably S levels in V. narbonensis exceed the 2,000 mg kg-1 DM optimum level for effective fiber digestion in ruminants (Miles and McDowell, 1983).
 
Micronutrients composition
 
Manganese (Mn)
 
Mn ranged from 38.45 to 52.00 mg kg-1 DM (mean: 45.59) in V. sativa and from 30.62 to 55.02 mg·kg-1 DM in V. narbonensis (mean: 43.83), consistent with 50.3 mg·kg-1 DM reported for common vetch by Caballero et al., (1996) and the 42.80 mg kg-1 DM reported for hairy vetch by Türk  et al. (2009). Accessions differed significantly for Mn, with V5 forming the highest homogeneous group and V9 the lowest. Both species exceed the 40 mg kg-1 DM threshold for ruminants (Tejada et al., 1987) and fall within the optimal guidelines of 50 mg kg-1 DM (max 150 mg kg-1 DM) reported by Vigh et al., (2023).
 
Copper (Cu)
 
V. narbonensis showed a wide Cu range (9.01-25.33 mg kg-1 DM; mean: 17.72), substantially higher than V. sativa (8.80-11.75 mg kg-1 DM; mean: 10.08). Accessions differed significantly for Cu, with V5 forming the highest homogeneous group and V11 the lowest. These values exceed the 7.3 mg kg-1 DM reported for common vetch by Caballero et al., (1996) but remain below the 19.76 mg kg-1 DM reported for V. villosa by Türk  et al. (2009). Copper content in both species aligned with current ruminant feeding guidelines (optimal: 10 mg kg-1 DM; max: 35 mg kg-1 DM) reported by Nozièreet_al(2018).
 
Zinc (Zn)
 
Zinc content ranged from 39.78 to 59.39 mg kg-1 DM (mean: 48.62) in V. sativa and from 36.96 to 76.00 mg kg-1 DM in V. narbonensis (mean: 61.72), consistent with the 30.2 mg kg-1 DM reported for common vetch by Caballero et al., (1996) and within Minson’s (1990a) global pasture range (7 to >100 mg kg-1 DM). Accessions differed significantly for Zn, with V8 forming the highest homogeneous group and V9 the lowest. Zn levels in both species meet the minimum requirement of 30 mg·kg-1 DM and remain below the regulatory maximum of 120 mg kg-1 DM (NASEM, 2016; Vigh et al., 2023).
 
Iron (Fe)
 
V. sativa Fe ranged from 252.73 to 1,461.90 mg kg-1 DM (mean: 644.38) and from 378.05 to 2,172.18 mg kg-1 DM in V. narbonensis (mean:1,010.08), consistent with Nizam et al., (2022) for common vetch (43.00-2,295.20 mg kg-1 DM). Accessions differed significantly for Fe, with V5 forming the highest homogeneous group and V12 the lowest. Several accessions exceeded the maximum tolerable Fe concentration of 500 mg kg-1 DM for grazing animals (Council, 2005), partly due to soil dust adhering to leaf surfaces in the clay-textured soil (Kabata-Pendias, 2000).
 
Boron (B)
 
 B ranged from 24.20 to 34.70 mg kg-1 DM in V. sativa (mean: 30.50) and from 24.90 to 45.44 mg kg-1 DM in V. narbonensis (mean: 30.95), within the range of 14-78 mg kg-1 DM reported for forage legumes by Kabata-Pendias (2000) and comparable to the 24 mg kg-1 DM reported for alfalfa by Gupta (1991). Accessions differed significantly for B, with V8 forming the highest homogeneous group and V12 the lowest. Both species remain well below the ruminant tolerance threshold of 800 mg kg-1 DM (Council, 2005), indicating no toxicity risk.
 
Cobalt (Co)
 
Cobalt content ranged from 0.28 to 0.73 mg kg-1 DM in V. sativa (mean: 0.52 mg kg-1 DM) and from 0.35 to 1.09 mg kg-1 DM in V. narbonensis (mean: 0.63 mg kg-1 DM), consistent with the levels reported for red clover (0.35 mg kg-1 DM) and lucerne (0.3-0.4 mg kg-1 DM) by Underwood (1999). Accessions differed significantly for Co, with V5 forming the highest homogeneous group and V3 the lowest. Both species exceeded the Co requirement of 0.20 mg kg-1 DM reported for rumen B12 synthesis by González-Montaña  et al. (2020), while remaining well below the maximum tolerable limit of 25 mg kg-1 DM (Council, 2005), with no evidence of deficiency or toxicity in any accession.
       
Despite uniform soil and management conditions (Table 3), accessions differed significantly for all twelve mineral elements, indicating that genetic background is the primary driver of the observed variability. Beyond genotype, seasonal and climatic conditions during plant growth are also known to influence forage mineral concentration, sometimes more strongly than soil characteristics (Jumba et al., 1995; McDowell, 1997), underscoring the importance of site-specific mineral analyses in forage evaluation (Suttle, 2022).
 
ANN analysis of mineral composition
 
The predictive function, f(x), demonstrates its effectiveness in revealing the differences in mineral composition between V. narbonensis and V.  sativa accessions. Optimal performance was achieved by minimizing prediction error (Fig 3); during the training phase, logistic regression was optimized using the least squares method with the mean squared error (MSE) on the test set:  0.0030044 (Fig 4). The model converged after only 5 iterations out of a maximum of 1,000, reflecting the clear mineral differentiation between the two species. 

Fig 3: Function optimization F(x).



Fig 4: Training phase optimization of ANN.

    
       
Among the twelve elements analyzed, Ca, Mg, K and Fe were the strongest contributors to species discrimination in the ANN model, whereas Mn, B and Co contributed less to the separation between V. sativa and V. narbonensis. These results highlight the capacity of ANNs to identify and model complex, non-linear relationships within multidimensional datasets such as those characterizing plant mineral composition (Watanabe et al., 2007).
       
ANOVA revealed significant differences among the 12 accessions for all twelve minerals, while ANN identified Ca, Mg, K and Fe as the most informative elements for discriminating between V. sativa and V. narbonensis. Thus, the two approaches provide complementary information on accession variation and species discrimination.

CONCLUSION

The present study demonstrates that both V. sativa and V. narbonensis provide adequate macro- and micronutrient levels for ruminant nutrition at full flowering stage under semi-arid Algerian conditions. Significant differences among accessions were observed for all twelve minerals, highlighting the influence of genetic background and environmental conditions on mineral composition.
               
The ANN model reliably discriminated between the two species based on their mineral profiles, with Ca, Mg, K and Fe being the most informative elements. Overall, combining ANOVA and ANN provided complementary insights into mineral variation and species discrimination. Further validation using independent datasets from different environments and species would help assess the robustness and broader applicability of the ANN approach.

ACKNOWLEDGEMENT

We sincerely thank Mr. MEHNAN Said, the manager of the farm where the trial was conducted, for his invaluable help in setting up and maintaining the experiment and his availability throughout the study. We would also like to thank Mr. Ahmet Yeşitaş for his contributions to the laboratory analyses, as well as Dr. BENTOUATI Insaf for her expert assistance with the statistical ANOVA computations.
 
Disclaimers
 
The views and conclusions expressed in this article are solely those of the authors and do not necessarily represent the views of their affiliated institutions. The authors are responsible for the accuracy and completeness of the information provided, but do not accept any liability for any direct or indirect losses resulting from the use of this content.
 

CONFLICT OF INTEREST

The authors declare that there are no conflicts of interest regarding the publication of this article. No funding or sponsorship influenced the design of the study, data collection, analysis, decision to publish, or preparation of the manuscript.

REFERENCES


  1. Algamal, Z.Y., Qasim, M.K., Lee, M.H. and Ali, H.T.M. (2021). Improving grasshopper optimization algorithm for hyperparameters estimation and feature selection in support vector regression. Chemometrics and Intelligent Laboratory Systems. 208: 104196. 

  2. Badrzadeh, M., Zaragarzadeh, F. and Esmaielpour, B. (2008). Chemical composition of some forage Vicia spp. in Iran. Journal of Food, Agriculture and Environment. 6(2): 178-180.

  3. Bhavani, R. (2025). Smart detection of macro nutrient deficiency in soybean plant using convolutional neural network. Legume Research. 48(11): 1855-1862. doi: 10.18805/LR-5534.

  4. Bougrine, H., Mebarkia, A. and Bechkri, S. (2022). Genetic diversity associated with eco-geographical parameters, morphological characteristics and soil analyses of common vetch (Vicia sativa L.) in Algeria. Agricultural Science Digest. 42(3): 317-321. doi: 10.18805/ag.DF-404.

  5. Caballero, R., Arauzo, M. and Hernaiz, P.J. (1996). Accumulation and redistribution of mineral elements in common vetch during pod filling. Agronomy Journal. 88(5): 801-805. 

  6. Council, N.R. (2001). Nutrient Requirements of Dairy Cattle. (7th rev. ed.).  National Academy Press.

  7. Council, N.R. (2005). Mineral Tolerance of Animals. (2nd rev. ed). National Academy Press.

  8. Deaville, E.R. and Flinn, P.C. (2000). Near-Infrared (NIR) Spectroscopy: An Alternative Approach for the Estimation of Forage Quality and Voluntary Intake. In Forage Evaluation in Ruminant Nutrition. CABI Publishing, Wallingford, UK. pp. 301-320. 

  9. Dhawale, A.W. and Banne, S.P. (2022). Comparative Study of Application of Artificial Neural Networks for Predicting Engineering Properties of Soil: A Review. In Proceedings of Fourth International Conference on Inventive Material Science Applications. Springer, Singapore. pp. 751-763. 

  10. Duc, G., Bao, S., Baum, M., Redden, B., Sadiki, M., Suso, M.J., Vishniakova, M. and Zong, X. (2010). Diversity maintenance and use of Vicia faba L. genetic resources. Field Crops Research. 115(3): 270-278. 

  11. Edmeades, D.C. and O’Connor, M.B. (2003). Sodium requirements for temperate pastures in New Zealand: A review. New Zealand Journal of Agricultural Research. 46(1): 37-47. 

  12. Enneking, D. (1995). The toxicity of Vicia species and their utilisation as grain legumes. CLIMA Occasional Publication No.6. Centre for Legumes in Mediterranean Agriculture, University of Western Australia.

  13. Food and Agriculture Organization of the United Nations. (1999). FAO Production Yearbook 1998. (Vol. 52, FAO Statistics Series No. 135). FAO. Rome.

  14. González-Montaña, J.R., Escalera-Valente, F., Alonso, A.J., Lomillos, J.M., Robles, R. and Alonso, M.E. (2020). Relationship between vitamin b12 and cobalt metabolism in domestic ruminant: An update. Animals. 10(10): 1-36. 

  15. Gupta, P.V., Kumar, V., Roy, D. and Kumar, M. (2015). Macro and micro-mineral profile of feeds, fodders and blood of livestock under farm condition in Mathura district of India. Indian Journal of Animal Research. 50(2): 203-206. doi: 10.18805/ijar.5724.

  16. Gupta, U.C. (1991). Boron, molybdenum and selenium status in different plant parts in forage legumes and vegetable crops. Journal of Plant Nutrition. 14(6): 613-621. 

  17. Jumba, I.O., Suttle, N.F. and Hunter, E.A. (1995). Effects of soil origin and mineral composition and herbage species on the mineral composition of forages in the Mount. Tropical Grasslands. 29: 40-46.

  18. Kabata-Pendias, A. (2000). Trace Elements in Soils and Plants. CRC Press, Boca Raton. 

  19. Khan, Z., Ahmad, K., Danish, M. and Mirzaei, F. (2012). Minerals profile of forages for grazing ruminants in Pakistan. International Journal of Livestock Research. 2(2): 74-82. 

  20. Makkar, H.P.S. and Singh, B. (1991). Composition, tannin levels and in-sacco dry matter digestibility of fresh and fallen oak (Quercus incana) leaves. Bioresource Technology. 37(2): 185-187. 

  21. Mancuso, S., Ferrini, F. and Nicese, F.P. (1999). Chestnut (Castanea sativa Mill.) genotype identification: An artificial neural network approach. Journal of Horticultural Science and Biotechnology. 74(6): 777-784. 

  22. McDowell, L.R. (1992). Minerals in Animal and Human Nutrition. (1st ed.). Academic Press, San Diego.

  23. McDowell, L.R. (1997). Minerals for Grazing Ruminants in Tropical Regions. (3rd ed.). University of Florida. 

  24. Mebarkia, A. and Abdelguerfi, A. (2007). Etude du potentiel agronomique de trois espèces de vesces (Vicia spp.) et variabilité dans la région semi-aride de Sétif (Algérie). Fourrages. 192: 495-506.

  25. Mertens, D. (2005). Plants: Preparation of Laboratory Sample (AOAC Official Method 922.02). In Official Methods of Analysis of AOAC International. (18th ed.). AOAC International, Gaithersburg, MD, USA. Chapter 3, pp. 1-2.

  26. Miles, W.H. and McDowell, L.R. (1983). Mineral deficiencies in the llanos rangeland of Colombia. World Animal Review. 46: 2-10.

  27. Minson, D.J. (1990a). Manganese. In Forage in Ruminant Nutrition (Ed. Minson D.J.). Academic Press, San Diego. pp. 359- 368. 

  28. Minson, D.J. (1990b). Zinc. In Forage in Ruminant Nutrition (Ed. Minson D.J.). Academic Press, San Diego.  pp. 346-358. 

  29. National Academies of Sciences, Engineering and Medicine (NASEM). (2016). Nutrient Requirements of Beef Cattle. (8th rev. ed.). National Academies Press, Washington, D.C.

  30. National Research Council (NRC). (2000). Nutrient Requirements of Beef Cattle: Update 2000. National Academy Press, Washington, D.C.

  31. Nawar, S. and Mouazen, A.M. (2017). Predictive performance of mobile vis-near infrared spectroscopy for key soil properties at different geographical scales by using spiking and data mining techniques. Catena. 151: 118- 129.

  32. Nizam, I., Tenikecier, H.S., Orak, A., Sen, C., Guler, N. and Ersoy, H. (2022). A preliminary study on the chemical structure of Vicia sativa L. accessions collected from natural flora of European part of Turkey. Journal of Tekirdag Agricultural Faculty. 19(2): 271-282. 

  33. Noziere, P., Sauvant, D. and Delaby, L. (2018). INRA 2018 : Alimentation des Ruminants. Editions Quae, Versailles.

  34. Raj, A.S., Badgujar, C.M., Lollato, R., Prasad, P.V.V. and Siliveru, K. (2024). Predicting rheological properties of wheat dough from flour properties using NIR coupled with artificial neural network. Journal of the ASABE. 67(4): 1023- 1035. 

  35. Rasool, S., Nazir, M., Sofi, A.P., Murtaza, I., Shikari, B.A., Shah, M.D., Nazir, N., Nehvi, F.A. and Zargar, M.S. (2019). Mineral profiling of common bean (Phaseolus vulgaris L.) germplasm. Indian Journal of Agricultural Research. 53(3): 270-276. doi: 10.18805/IJARe.A-5160.

  36. Suttle, N. (2022). Mineral Nutrition of Livestock. (5th ed.). CAB International, Wallingford. 

  37. Tejada, R., McDowell, L.R., Martin, F.G. and Conrad, J.H. (1987). Evaluation of the macromineral and crude protein status of cattle in specific regions in Guatemala. Nutrition Reports International. 35(5): 989-998.

  38. Thompson, D.J. (1972). Potassium in Animal Nutrition. International Minerals and Chemical Corporation, Libertyville, Illinois, USA.

  39. Tripathi, D.K., Singh, V.P., Prasad, S.M., Chauhan, D.K. and Dubey, N.K. (2015). Silicon nanoparticles (SiNp) alleviate chromium (VI) phytotoxicity in Pisum sativum (L.) seedlings. Plant Physiology and Biochemistry. 96: 189-198. 

  40. Türk, M., Albayrak, S. and Yüksel, O. (2009). Effects of fertilisation and harvesting stages on forage yield and quality of hairy vetch (Vicia villosa Roth). New Zealand Journal of Agricultural Research. 52(3): 269-275. 

  41. Underwood, E.J. (1999). The Mineral Nutrition of Livestock. (3rd ed.). CABI Publishing, Wallingford, UK.

  42. Vasiljevic, S., Milic, D. and Mikic, A. (2009). Chemical attributes and quality improvement of forage legumes. Biotechnology in Animal Husbandry. 25(5-6-1): 493-504.

  43. Vigh, A., Criste, A., Corcionivoschi, N. and Gerard, C. (2023). Rumen solubility of copper, manganese and zinc and the potential link between the source and rumen function: A systematic review. Agriculture. 13(12): 2198.

  44. Wang, S., Li, Y., Yuan, J., Song, L., Liu, X. and Liu, X. (2021). Recognition of cotton growth period for precise spraying based on convolution neural network. Information Processing in Agriculture. 8(2): 219-231. 

  45. Watanabe, T., Broadley, M.R., Jansen, S., White, P.J., Takada, J., Satake, K., Takamatsu, T., Tuah, S.J. and Osaki, M. (2007). Evolutionary control of leaf element composition in plants. New Phytologist. 174(3): 516-523. 

  46. Yücel, C., Yücel, D., Akkaya, M.R. and Anlarsal, A.E. (2014). Bazı Ümitvar Yaygın Fiğ (Vicia sativa L.) Genotiplerinde Kalite Özellikleri. KSÜ Doğa Bilimleri Dergisi. 17(1): 8-14.

  47. Zamora-Natera J.F., Rodriguez-Macias R., Salcedo-Perez E., García- Lopez P., Barrientos-Ramirez L., Vargas-Radillo J., Soto- Velasco C., Ruiz-López M.A. (2019). Forage potential of three wild species of genus Lupinus (Leguminosae) from Mexico. Legume Research. 43(1): 93-98. doi: 10.18805/LR-453.
Published In
Indian Journal of Agricultural Research
Article Metrics
Metrics Icon
16
Views
0
Citations
Reviewed By
Share Article
Download Article
In this Article
    Contact us
    Follow us

    Full Research Article

    Mineral Composition of Vicia sativa and Vicia narbonensis Forages: An Artificial Neural Network Approach

    K
    Katia Regaz1,*
    https://orcid.org/0009-0004-3689-8091
    K
    Khalissa Cheniti1
    https://orcid.org/0000-0002-3037-7672
    N
    Nizar Moujahed2
    https://orcid.org/0000-0002-9420-4623
    Y
    Yusuf Konca3
    https://orcid.org/0000-0002-6231-1512
    S
    Selma Buyukkilic Beyzi3
    https://orcid.org/0000-0002-4622-0645
    J
    Javier Vioque-Peña4
    https://orcid.org/0000-0002-0109-0584
    1Laboratory of Improvement and Development of Vegetal and Animal Production, Agronomic Sciences Department, Faculty of Natural Science and Life, Ferhat Abbas Setif1 University, Sétif 19000, Algeria.
    2Laboratory of Ecosystems and Aquatic Resources, National Agronomic Institute of Tunisia, University of Carthage, UR03AGRO1, 1082 Tunis, Tunisia.
    3Laboratory of Zootechnics, Department of Animal Science, Faculty of Agriculture, University of Erciyes, 38039 Melikgazi, Kayseri, Turkey.
    4Food Phytochemistry Department, Instituto de la Grasa (C.S.I.C.), Campus Universidad Pablo de Olavide, Carretera de Utrera Km 1, 41089 Sevilla, Spain.

    ABSTRACT

    Background: Forage quality fundamentally relies on optimal mineral composition, which directly influences ruminant health and productivity. This study investigated the nutritional mineral profiles of twelve accessions of two vetch species (Vicia sativa and Vicia narbonensis) by combining ICP-OES analysis with artificial neural network (ANN) modeling, to assess their nutritional potential for ruminant feeding under semi-arid conditions.

    Methods: Plant material was harvested at full flowering stage during the 2021/2022 growing season from the Beni Fouda research farm (Sétif, Algeria; 1199 m altitude; 36°15'N, 5°29'E). Macro- and micronutrient concentrations (Ca, K, Mg, Na, P, S, Co, B, Cu, Fe, Mn, Zn) were quantified by ICP-OES (Perkin-Elmer Optima 4300 DV). Accession effects were tested by one-way ANOVA with Tukey’s HSD test and an ANN model was built to explore species-level mineral discrimination.

    Result: Accession significantly affected all twelve elements, reflecting strong genotypic variability within both species. The ANN model (MSE = 0.003) identified Ca, Mg, K and Fe as the elements contributing most to species separation. Both species exhibited macro- and micronutrient concentrations that generally met the mineral requirements of ruminants, confirming their potential as valuable forage resources.

    INTRODUCTION

    The nutrient composition of forage is a key determinant of animal nutrition and productivity. For quality feed, plants must contain mineral substances at specific levels, with both the presence and proportions of these minerals being essential determinants of overall feed value (Yücel  et al., 2014; Rasool et al., 2019). Mineral imbalances whether deficiencies or excesses can adversely affect livestock health and performance (Nizam et al., 2022).
           
    Among leguminous forages, species of the genus Vicia are particularly notable for their agricultural and economic importance (Duc et al., 2010). Vicia narbonensis and Vicia sativa are widely cultivated for animal feeding in semi-arid Mediterranean environments (Bougrine et al., 2022). The mineral composition of these forages is a key determinant of their nutritional value, as it directly influences their ability to supply essential macro- and micronutrients required by livestock (Zamora-Natera  et al., 2019; Bhavani, 2025). Nutrient recommendations established for ruminant diets provide valuable reference standards for assessing forage mineral adequacy (National Research Council, 2000; NASEM, 2016). Therefore, comparing the mineral composition of V. narbonensis and V. sativa with these reference values is essential for evaluating their nutritional potential and identifying their suitability as mineral sources in ruminant feeding systems (Gupta et al., 2015).
           
    Beyond classical statistical approaches, artificial neural networks (ANNs) represent a powerful alternative for analyzing complex elemental composition data (Algamal et al., 2021). ANNs can identify non-linear patterns and model multidimensional relationships within plant elemental datasets tasks particularly suited to discriminating between species or genotypes (Mancuso et al., 1999). The application of trained ANN models to generate predictive functions can illuminate significant differences in elemental composition between plant species and thereby guide optimized agronomic management (Tripathi et al., 2015).
                   
    Therefore, the objectives of this study were to assess the variability in macro and micronutrient concentrations in Vicia sativa and Vicia narbonensis accessions at full flowering stage and to evaluate the effectiveness of ANNs in modeling elemental composition to assess their nutritional potential for ruminant feeding.

    MATERIALS AND METHODS

    Study site and climatic conditions
     
    The field study was carried out during the 2021/2022 agricultural season (from November 2021 to June 2022) at Beni Fouda research farm, affiliated with Ferhat Abbas Sétif 1 University (Algeria), located at 1199m altitude (36°15'08"N, 5°29'43"E). The region has a semi-arid climate with hot dry summers, cold winters and average annual rainfall of 450 mm (Mebarkia and Abdelguerfi, 2007). The trial year was notably dry, with a total rainfall of only 322 mm (Table 1).

    Table 1: Monthly temperature and precipitation during the 2021/2022 growing season.


     
    Plant material and experimental design
     
    The study included twelve accessions: eight of V. sativa and four of V. narbonensis (Table 2). The experiment followed a completely randomized block design with three replications. Sowing was performed manually on December 28th, 2021. No fertilization or phytosanitary treatments were applied.

    ​

    Table 2: Studied plant material ecotypes origin.


     
    Physical and chemical properties of soil
     
    The experimental soil was a compact, heavy clay soil, alkaline (pH 8.0), deficient in organic matter (1.49%) and moderately non-saline, with an electrical conductivity of 0.23 mS cm-1. Its mineral composition was determined prior to sowing and is presented in Table 3.

    ​

    Table 3: Mineral composition of the experimental soil (mg.kg-1).


     
    Mineral analysis
     
    Sampling was performed at full flowering stage in April 2022. A 1 m2 area was harvested and oven-dried (50°C, 48 h; Makkar and Singh, 1991). Approximately 0.5 g of dried sample (three replicates) was digested in an HNO3/HClO4  mixture (2:3 v/v) using microwave-assisted digestion (Bergof Speed Wave MWS-2). The mineral concentrations were determined by ICP-OES (Perkin-Elmer Optima 4300 DV) following Mertens (2005), at the Laboratory of Zootechnics, Department of Animal Science, Faculty of Agriculture, University of Erciyes, Kayseri, Türkiye.
           
    All concentrations are expressed on a dry matter basis (mg kg-1 DM) and compared with established ruminant dietary requirements reported on the same basis.
     
    Statistical analysis
     
    Statistical analyses were performed using Statistica software (version 8, StatSoft). Mineral content was analyzed by one-way ANOVA, with accession as the factor. Homogeneous groups were determined using Tukey’s HSD test at p<0.05.
     
    Artificial neural networks presentation
     
    The dataset comprised elemental concentration measurements for twelve minerals across all analyzed accessions. Prior to analysis, data were normalized to ensure all elements were represented on a comparable scale, a critical pre-processing step that directly influences model accuracy (Dhawale and Banne, 2022). The ANN was structured with three functional layers: An input layer receiving the twelve elemental variables, one or more hidden layers responsible for extracting non-linear patterns within the dataset and an output layer predicting species identity (V. sativa or V. narbonensis). The number of neurons per hidden layer was selected to balance model complexity and generalization capacity (Wang et al., 2021). This architecture is illustrated in Fig 1.

    Fig 1: System architecture of neural networks analysis.


           
    Once the architecture was finalized, the training process was started. During this phase, network weights and biases were iteratively adjusted to minimize the discrepancy between predicted and observed species identity (Raj et al., 2024). This process continued until the ANN converged, yielding a predictive function f(x) that maps normalized elemental concentrations to species membership (Nawar and Mouazen, 2017). The function takes the form:
     
    f(x) = 86x4/13 - 65x2/13 - 54x3/13 - 32x1/26 + 41x5/65 + 29x6/20 - 84x7/13 - 31x8/26 - 28x9/13 + 17x10/32 - 79x11/13 - 47x12/13 + 26/16
     
    Where,
    x1-x12= The normalized concentrations of the twelve elements in the order: Ca, K, Mg, Na, P, S, Co, B, Cu, Fe, Mn, Zn.
           
    Despite a maximum of 1,000 iterations set for optimization, the model converged after only 5 iterations, reflecting the clear interspecific differentiation in elemental profiles (Fig 2).

    Fig 2: Training performance of neural networks analysis.

    RESULTS AND DISCUSSION

    One-way ANOVA revealed a significant accession effect across all 12 analyzed mineral elements (p<0.05), with Tukey’s HSD test identifying distinct homogeneous groups. For clarity, results are presented and discussed separately for macronutrients (Table 4) and micronutrients (Table 5).

    Table 4: Macronutrient composition of V. sativa and V. narbonensis accessions (mg.kg-1 DM).



    ​

    Table 5: Micronutrient composition of V. sativa and V. narbonensis accessions (mg.kg-1DM).


     
    Macronutrients composition
     
    Calcium (Ca)
     
    Ca ranged from 21,174.19 to 25,529.8 mg kg-1 DM (mean: 23,037.62) in V. sativa and from 15,257.31 to 26,560.47 mg kg-1 DM in V. narbonensis (mean: 22,045.46). Accessions differed significantly for Ca, with V8 forming the highest homogeneous group and V9 the lowest. Both species exceed values reported by Badrzadeh et al., (2008) under semi-arid conditions (13,700-13,800 mg·kg-1 DM) but remain below Vasiljevic et al., (2009) for V. sativa (34,530 mg kg-1 DM). Both species exceed established ruminant Ca requirements (1,800-8,200 mg kg-1 DM; Deaville and Flinn, 2000).
     
    Potassium (K)
     
    K concentrations varied significantly among accessions, ranging from 10,061.7 to 14,924.65 mg kg-1 DM (mean: 11,754.93) in V. sativa and from 8,979.14 to 14,934.98 mg kg-1 DM in V. narbonensis (mean: 11,616.61), with accession V8 recording the highest and V9 the lowest. Results align with Badrzadeh et al., (2008) for V. sativa (13,500 mg kg-1 DM), though lower than the 24,500 mg kg-1 DM reported by Enneking (1995) for V. narbonensis. Both species exceed K requirements for all ruminant categories (6,000-10,000 mg kg-1 DM; Thompson, 1972; Council et al., 2001).
     
    Phosphorus (P)
     
    P accumulation revealed significant differences among accessions, with V. narbonensis displaying a higher average concentration (mean: 3,455.00; range: 2,242.97 to 4,002.43 mg kg-1 DM) than V. sativa (mean: 3,046.16; range: 1,896.27 to 3,968.68 mg kg-1 DM). Accession V5 formed the highest statistically significant group, whereas V11 represented the lowest. Results exceed those of Badrzadeh et al., (2008) in semi-arid Iran (840-1,040 mg kg-1 DM) but align with FAO (1999) for V. narbonensis (3,600 mg kg-1 DM) and established forage mineral reference values (2,900 mg·kg-1 DM; Minson, 1990a). Both species satisfy ruminant P requirements (1,800-3,000 mg kg-1 DM; McDowell, 1992; Council et al., 2001).
     
    Magnesium (Mg)
     
    V. sativa Mg ranged from 3,185.56 to 4,151.98 mg kg-1 DM (mean: 3,573.27), consistent with Caballero et al., (1996) at flowering stage (3,700 mg kg-1 DM). V. narbonensis showed a higher mean (3,771.42 mg kg-1 DM) and greater variability (2,884.99-5,254.11 mg kg-1 DM), exceeding the general forage range of 2,600-2,800 mg kg-1 DM (Minson, 1990b). Accessions differed significantly for Mg, with V8 forming the highest homogeneous group and V9 the lowest. Both species comfortably meet ruminant Mg requirements (1,000-2,000 mg kg-1 DM; McDowell, 1997).

    Sodium (Na)
     
    V. sativa Na ranged from 589.80 to 5,624.12 mg kg-1 DM (mean: 3,054.63) and from 1,070.96 to 4,003.02 mg kg-1 DM in V. narbonensis (mean: 2,432.19). Overall these concentrations align closely with 3,000-4,000 mg kg-1 DM range reported in forage legumes (Edmeades and O’Connor, 2003). Accessions differed significantly for Na, with V2 forming the highest homogeneous group and V12 the lowest. The wide V. sativa range suggests species-specific Na uptake mechanisms (Khan et al., 2012). V. narbonensis Na falls within ruminant recommendations (1,000-4,000 mg kg-1 DM; Underwood, 1999; Suttle, 2022), while V. sativa slightly exceeds this range.

    Sulfur (S)
     
    S content differed significantly among all 12 accessions. V. narbonensis recorded markedly higher values (3,250.89-4,020.51 mg kg-1 DM; mean: 3,607.71), than V. sativa (1,718.59-2,322.17 mg kg-1 DM; mean: 1,969.82). Accession V8 formed the highest statistically significant group and V11 the lowest. All values fall within the range reported by Suttle (2022) (500 to >5,000 mg kg-1 DM). Notably S levels in V. narbonensis exceed the 2,000 mg kg-1 DM optimum level for effective fiber digestion in ruminants (Miles and McDowell, 1983).
     
    Micronutrients composition
     
    Manganese (Mn)
     
    Mn ranged from 38.45 to 52.00 mg kg-1 DM (mean: 45.59) in V. sativa and from 30.62 to 55.02 mg·kg-1 DM in V. narbonensis (mean: 43.83), consistent with 50.3 mg·kg-1 DM reported for common vetch by Caballero et al., (1996) and the 42.80 mg kg-1 DM reported for hairy vetch by Türk  et al. (2009). Accessions differed significantly for Mn, with V5 forming the highest homogeneous group and V9 the lowest. Both species exceed the 40 mg kg-1 DM threshold for ruminants (Tejada et al., 1987) and fall within the optimal guidelines of 50 mg kg-1 DM (max 150 mg kg-1 DM) reported by Vigh et al., (2023).
     
    Copper (Cu)
     
    V. narbonensis showed a wide Cu range (9.01-25.33 mg kg-1 DM; mean: 17.72), substantially higher than V. sativa (8.80-11.75 mg kg-1 DM; mean: 10.08). Accessions differed significantly for Cu, with V5 forming the highest homogeneous group and V11 the lowest. These values exceed the 7.3 mg kg-1 DM reported for common vetch by Caballero et al., (1996) but remain below the 19.76 mg kg-1 DM reported for V. villosa by Türk  et al. (2009). Copper content in both species aligned with current ruminant feeding guidelines (optimal: 10 mg kg-1 DM; max: 35 mg kg-1 DM) reported by Nozièreet_al(2018).
     
    Zinc (Zn)
     
    Zinc content ranged from 39.78 to 59.39 mg kg-1 DM (mean: 48.62) in V. sativa and from 36.96 to 76.00 mg kg-1 DM in V. narbonensis (mean: 61.72), consistent with the 30.2 mg kg-1 DM reported for common vetch by Caballero et al., (1996) and within Minson’s (1990a) global pasture range (7 to >100 mg kg-1 DM). Accessions differed significantly for Zn, with V8 forming the highest homogeneous group and V9 the lowest. Zn levels in both species meet the minimum requirement of 30 mg·kg-1 DM and remain below the regulatory maximum of 120 mg kg-1 DM (NASEM, 2016; Vigh et al., 2023).
     
    Iron (Fe)
     
    V. sativa Fe ranged from 252.73 to 1,461.90 mg kg-1 DM (mean: 644.38) and from 378.05 to 2,172.18 mg kg-1 DM in V. narbonensis (mean:1,010.08), consistent with Nizam et al., (2022) for common vetch (43.00-2,295.20 mg kg-1 DM). Accessions differed significantly for Fe, with V5 forming the highest homogeneous group and V12 the lowest. Several accessions exceeded the maximum tolerable Fe concentration of 500 mg kg-1 DM for grazing animals (Council, 2005), partly due to soil dust adhering to leaf surfaces in the clay-textured soil (Kabata-Pendias, 2000).
     
    Boron (B)
     
     B ranged from 24.20 to 34.70 mg kg-1 DM in V. sativa (mean: 30.50) and from 24.90 to 45.44 mg kg-1 DM in V. narbonensis (mean: 30.95), within the range of 14-78 mg kg-1 DM reported for forage legumes by Kabata-Pendias (2000) and comparable to the 24 mg kg-1 DM reported for alfalfa by Gupta (1991). Accessions differed significantly for B, with V8 forming the highest homogeneous group and V12 the lowest. Both species remain well below the ruminant tolerance threshold of 800 mg kg-1 DM (Council, 2005), indicating no toxicity risk.
     
    Cobalt (Co)
     
    Cobalt content ranged from 0.28 to 0.73 mg kg-1 DM in V. sativa (mean: 0.52 mg kg-1 DM) and from 0.35 to 1.09 mg kg-1 DM in V. narbonensis (mean: 0.63 mg kg-1 DM), consistent with the levels reported for red clover (0.35 mg kg-1 DM) and lucerne (0.3-0.4 mg kg-1 DM) by Underwood (1999). Accessions differed significantly for Co, with V5 forming the highest homogeneous group and V3 the lowest. Both species exceeded the Co requirement of 0.20 mg kg-1 DM reported for rumen B12 synthesis by González-Montaña  et al. (2020), while remaining well below the maximum tolerable limit of 25 mg kg-1 DM (Council, 2005), with no evidence of deficiency or toxicity in any accession.
           
    Despite uniform soil and management conditions (Table 3), accessions differed significantly for all twelve mineral elements, indicating that genetic background is the primary driver of the observed variability. Beyond genotype, seasonal and climatic conditions during plant growth are also known to influence forage mineral concentration, sometimes more strongly than soil characteristics (Jumba et al., 1995; McDowell, 1997), underscoring the importance of site-specific mineral analyses in forage evaluation (Suttle, 2022).
     
    ANN analysis of mineral composition
     
    The predictive function, f(x), demonstrates its effectiveness in revealing the differences in mineral composition between V. narbonensis and V.  sativa accessions. Optimal performance was achieved by minimizing prediction error (Fig 3); during the training phase, logistic regression was optimized using the least squares method with the mean squared error (MSE) on the test set:  0.0030044 (Fig 4). The model converged after only 5 iterations out of a maximum of 1,000, reflecting the clear mineral differentiation between the two species. 

    Fig 3: Function optimization F(x).



    Fig 4: Training phase optimization of ANN.

        
           
    Among the twelve elements analyzed, Ca, Mg, K and Fe were the strongest contributors to species discrimination in the ANN model, whereas Mn, B and Co contributed less to the separation between V. sativa and V. narbonensis. These results highlight the capacity of ANNs to identify and model complex, non-linear relationships within multidimensional datasets such as those characterizing plant mineral composition (Watanabe et al., 2007).
           
    ANOVA revealed significant differences among the 12 accessions for all twelve minerals, while ANN identified Ca, Mg, K and Fe as the most informative elements for discriminating between V. sativa and V. narbonensis. Thus, the two approaches provide complementary information on accession variation and species discrimination.

    CONCLUSION

    The present study demonstrates that both V. sativa and V. narbonensis provide adequate macro- and micronutrient levels for ruminant nutrition at full flowering stage under semi-arid Algerian conditions. Significant differences among accessions were observed for all twelve minerals, highlighting the influence of genetic background and environmental conditions on mineral composition.
                   
    The ANN model reliably discriminated between the two species based on their mineral profiles, with Ca, Mg, K and Fe being the most informative elements. Overall, combining ANOVA and ANN provided complementary insights into mineral variation and species discrimination. Further validation using independent datasets from different environments and species would help assess the robustness and broader applicability of the ANN approach.

    ACKNOWLEDGEMENT

    We sincerely thank Mr. MEHNAN Said, the manager of the farm where the trial was conducted, for his invaluable help in setting up and maintaining the experiment and his availability throughout the study. We would also like to thank Mr. Ahmet Yeşitaş for his contributions to the laboratory analyses, as well as Dr. BENTOUATI Insaf for her expert assistance with the statistical ANOVA computations.
     
    Disclaimers
     
    The views and conclusions expressed in this article are solely those of the authors and do not necessarily represent the views of their affiliated institutions. The authors are responsible for the accuracy and completeness of the information provided, but do not accept any liability for any direct or indirect losses resulting from the use of this content.
     

    CONFLICT OF INTEREST

    The authors declare that there are no conflicts of interest regarding the publication of this article. No funding or sponsorship influenced the design of the study, data collection, analysis, decision to publish, or preparation of the manuscript.

    REFERENCES


    1. Algamal, Z.Y., Qasim, M.K., Lee, M.H. and Ali, H.T.M. (2021). Improving grasshopper optimization algorithm for hyperparameters estimation and feature selection in support vector regression. Chemometrics and Intelligent Laboratory Systems. 208: 104196. 

    2. Badrzadeh, M., Zaragarzadeh, F. and Esmaielpour, B. (2008). Chemical composition of some forage Vicia spp. in Iran. Journal of Food, Agriculture and Environment. 6(2): 178-180.

    3. Bhavani, R. (2025). Smart detection of macro nutrient deficiency in soybean plant using convolutional neural network. Legume Research. 48(11): 1855-1862. doi: 10.18805/LR-5534.

    4. Bougrine, H., Mebarkia, A. and Bechkri, S. (2022). Genetic diversity associated with eco-geographical parameters, morphological characteristics and soil analyses of common vetch (Vicia sativa L.) in Algeria. Agricultural Science Digest. 42(3): 317-321. doi: 10.18805/ag.DF-404.

    5. Caballero, R., Arauzo, M. and Hernaiz, P.J. (1996). Accumulation and redistribution of mineral elements in common vetch during pod filling. Agronomy Journal. 88(5): 801-805. 

    6. Council, N.R. (2001). Nutrient Requirements of Dairy Cattle. (7th rev. ed.).  National Academy Press.

    7. Council, N.R. (2005). Mineral Tolerance of Animals. (2nd rev. ed). National Academy Press.

    8. Deaville, E.R. and Flinn, P.C. (2000). Near-Infrared (NIR) Spectroscopy: An Alternative Approach for the Estimation of Forage Quality and Voluntary Intake. In Forage Evaluation in Ruminant Nutrition. CABI Publishing, Wallingford, UK. pp. 301-320. 

    9. Dhawale, A.W. and Banne, S.P. (2022). Comparative Study of Application of Artificial Neural Networks for Predicting Engineering Properties of Soil: A Review. In Proceedings of Fourth International Conference on Inventive Material Science Applications. Springer, Singapore. pp. 751-763. 

    10. Duc, G., Bao, S., Baum, M., Redden, B., Sadiki, M., Suso, M.J., Vishniakova, M. and Zong, X. (2010). Diversity maintenance and use of Vicia faba L. genetic resources. Field Crops Research. 115(3): 270-278. 

    11. Edmeades, D.C. and O’Connor, M.B. (2003). Sodium requirements for temperate pastures in New Zealand: A review. New Zealand Journal of Agricultural Research. 46(1): 37-47. 

    12. Enneking, D. (1995). The toxicity of Vicia species and their utilisation as grain legumes. CLIMA Occasional Publication No.6. Centre for Legumes in Mediterranean Agriculture, University of Western Australia.

    13. Food and Agriculture Organization of the United Nations. (1999). FAO Production Yearbook 1998. (Vol. 52, FAO Statistics Series No. 135). FAO. Rome.

    14. González-Montaña, J.R., Escalera-Valente, F., Alonso, A.J., Lomillos, J.M., Robles, R. and Alonso, M.E. (2020). Relationship between vitamin b12 and cobalt metabolism in domestic ruminant: An update. Animals. 10(10): 1-36. 

    15. Gupta, P.V., Kumar, V., Roy, D. and Kumar, M. (2015). Macro and micro-mineral profile of feeds, fodders and blood of livestock under farm condition in Mathura district of India. Indian Journal of Animal Research. 50(2): 203-206. doi: 10.18805/ijar.5724.

    16. Gupta, U.C. (1991). Boron, molybdenum and selenium status in different plant parts in forage legumes and vegetable crops. Journal of Plant Nutrition. 14(6): 613-621. 

    17. Jumba, I.O., Suttle, N.F. and Hunter, E.A. (1995). Effects of soil origin and mineral composition and herbage species on the mineral composition of forages in the Mount. Tropical Grasslands. 29: 40-46.

    18. Kabata-Pendias, A. (2000). Trace Elements in Soils and Plants. CRC Press, Boca Raton. 

    19. Khan, Z., Ahmad, K., Danish, M. and Mirzaei, F. (2012). Minerals profile of forages for grazing ruminants in Pakistan. International Journal of Livestock Research. 2(2): 74-82. 

    20. Makkar, H.P.S. and Singh, B. (1991). Composition, tannin levels and in-sacco dry matter digestibility of fresh and fallen oak (Quercus incana) leaves. Bioresource Technology. 37(2): 185-187. 

    21. Mancuso, S., Ferrini, F. and Nicese, F.P. (1999). Chestnut (Castanea sativa Mill.) genotype identification: An artificial neural network approach. Journal of Horticultural Science and Biotechnology. 74(6): 777-784. 

    22. McDowell, L.R. (1992). Minerals in Animal and Human Nutrition. (1st ed.). Academic Press, San Diego.

    23. McDowell, L.R. (1997). Minerals for Grazing Ruminants in Tropical Regions. (3rd ed.). University of Florida. 

    24. Mebarkia, A. and Abdelguerfi, A. (2007). Etude du potentiel agronomique de trois espèces de vesces (Vicia spp.) et variabilité dans la région semi-aride de Sétif (Algérie). Fourrages. 192: 495-506.

    25. Mertens, D. (2005). Plants: Preparation of Laboratory Sample (AOAC Official Method 922.02). In Official Methods of Analysis of AOAC International. (18th ed.). AOAC International, Gaithersburg, MD, USA. Chapter 3, pp. 1-2.

    26. Miles, W.H. and McDowell, L.R. (1983). Mineral deficiencies in the llanos rangeland of Colombia. World Animal Review. 46: 2-10.

    27. Minson, D.J. (1990a). Manganese. In Forage in Ruminant Nutrition (Ed. Minson D.J.). Academic Press, San Diego. pp. 359- 368. 

    28. Minson, D.J. (1990b). Zinc. In Forage in Ruminant Nutrition (Ed. Minson D.J.). Academic Press, San Diego.  pp. 346-358. 

    29. National Academies of Sciences, Engineering and Medicine (NASEM). (2016). Nutrient Requirements of Beef Cattle. (8th rev. ed.). National Academies Press, Washington, D.C.

    30. National Research Council (NRC). (2000). Nutrient Requirements of Beef Cattle: Update 2000. National Academy Press, Washington, D.C.

    31. Nawar, S. and Mouazen, A.M. (2017). Predictive performance of mobile vis-near infrared spectroscopy for key soil properties at different geographical scales by using spiking and data mining techniques. Catena. 151: 118- 129.

    32. Nizam, I., Tenikecier, H.S., Orak, A., Sen, C., Guler, N. and Ersoy, H. (2022). A preliminary study on the chemical structure of Vicia sativa L. accessions collected from natural flora of European part of Turkey. Journal of Tekirdag Agricultural Faculty. 19(2): 271-282. 

    33. Noziere, P., Sauvant, D. and Delaby, L. (2018). INRA 2018 : Alimentation des Ruminants. Editions Quae, Versailles.

    34. Raj, A.S., Badgujar, C.M., Lollato, R., Prasad, P.V.V. and Siliveru, K. (2024). Predicting rheological properties of wheat dough from flour properties using NIR coupled with artificial neural network. Journal of the ASABE. 67(4): 1023- 1035. 

    35. Rasool, S., Nazir, M., Sofi, A.P., Murtaza, I., Shikari, B.A., Shah, M.D., Nazir, N., Nehvi, F.A. and Zargar, M.S. (2019). Mineral profiling of common bean (Phaseolus vulgaris L.) germplasm. Indian Journal of Agricultural Research. 53(3): 270-276. doi: 10.18805/IJARe.A-5160.

    36. Suttle, N. (2022). Mineral Nutrition of Livestock. (5th ed.). CAB International, Wallingford. 

    37. Tejada, R., McDowell, L.R., Martin, F.G. and Conrad, J.H. (1987). Evaluation of the macromineral and crude protein status of cattle in specific regions in Guatemala. Nutrition Reports International. 35(5): 989-998.

    38. Thompson, D.J. (1972). Potassium in Animal Nutrition. International Minerals and Chemical Corporation, Libertyville, Illinois, USA.

    39. Tripathi, D.K., Singh, V.P., Prasad, S.M., Chauhan, D.K. and Dubey, N.K. (2015). Silicon nanoparticles (SiNp) alleviate chromium (VI) phytotoxicity in Pisum sativum (L.) seedlings. Plant Physiology and Biochemistry. 96: 189-198. 

    40. Türk, M., Albayrak, S. and Yüksel, O. (2009). Effects of fertilisation and harvesting stages on forage yield and quality of hairy vetch (Vicia villosa Roth). New Zealand Journal of Agricultural Research. 52(3): 269-275. 

    41. Underwood, E.J. (1999). The Mineral Nutrition of Livestock. (3rd ed.). CABI Publishing, Wallingford, UK.

    42. Vasiljevic, S., Milic, D. and Mikic, A. (2009). Chemical attributes and quality improvement of forage legumes. Biotechnology in Animal Husbandry. 25(5-6-1): 493-504.

    43. Vigh, A., Criste, A., Corcionivoschi, N. and Gerard, C. (2023). Rumen solubility of copper, manganese and zinc and the potential link between the source and rumen function: A systematic review. Agriculture. 13(12): 2198.

    44. Wang, S., Li, Y., Yuan, J., Song, L., Liu, X. and Liu, X. (2021). Recognition of cotton growth period for precise spraying based on convolution neural network. Information Processing in Agriculture. 8(2): 219-231. 

    45. Watanabe, T., Broadley, M.R., Jansen, S., White, P.J., Takada, J., Satake, K., Takamatsu, T., Tuah, S.J. and Osaki, M. (2007). Evolutionary control of leaf element composition in plants. New Phytologist. 174(3): 516-523. 

    46. Yücel, C., Yücel, D., Akkaya, M.R. and Anlarsal, A.E. (2014). Bazı Ümitvar Yaygın Fiğ (Vicia sativa L.) Genotiplerinde Kalite Özellikleri. KSÜ Doğa Bilimleri Dergisi. 17(1): 8-14.

    47. Zamora-Natera J.F., Rodriguez-Macias R., Salcedo-Perez E., García- Lopez P., Barrientos-Ramirez L., Vargas-Radillo J., Soto- Velasco C., Ruiz-López M.A. (2019). Forage potential of three wild species of genus Lupinus (Leguminosae) from Mexico. Legume Research. 43(1): 93-98. doi: 10.18805/LR-453.
    In this Article
      Contact us
      Follow us
      Published In
      Indian Journal of Agricultural Research

      Editorial Board

      View all (0)