Predictive topological modeling of the structure–property relationships in naturally occurring anti-cancer chalcones Authors: Abdullah Abdul Razzaq Obaid, Abdul Hakeem, Asad Ullah, Shahid Zaman, Hijaz Ahmad, Parvez Ali, Melaku Berhe Belay Published: 2025 ABSTRACT AbstractIn this study, a comprehensive predictive topological modeling framework was employed to investigate the quantitative structure–property relationships (QSPR) of naturally occurring anticancer chalcones. A series of degree-based topological indices, including the first and second Zagreb indices, first and second redefined Zagreb indices, sum-connectivity index, symmetric division index, Randić index, and harmonic index, were computed to represent key molecular structural features. Statistical and regression analyses were performed to correlate these indices with experimentally reported physicochemical properties of chalcones. The study focuses on eight naturally derived anticancer Chalcones, namely, Isoliquiritigenin, Butein, Cardamonin, Sappanchalcone, Licochalcone A, Millepachine, Xanthohumol, and Curcumin—selected based on their structural diversity and well-documented pharmacological potential. Correlations between the computed indices and several physicochemical parameters, such as molecular complexity, molecular weight, molar refractivity, and polarizability, were examined. The study also includes a comprehensive statistical evaluation of the regression models to assess the predictive accuracy, robustness, and reliability of the proposed descriptors.Keywords: Graph theory, Anti-cancer drugs, QSPR analysis, Regression modelSubject terms: Chemistry, Computational biology and bioinformatics, Mathematics and computing Introduction Cancer remains one of the leading causes of death worldwide, posing a major challenge to global public health. According to the World Health Organization (WHO), cancer accounted for nearly 10 million deaths in 2022, with breast, lung, colorectal, and prostate cancers being the most prevalent types. The rising incidence and mortality rates highlight the urgent need for discovering and developing effective, low-toxicity anti-cancer agents. In recent years, naturally occurring compounds have gained significant attention due to their structural diversity and therapeutic potential. Among these, Chalcones aromatic ketones belonging to the flavonoid family have emerged as promising candidates because of their wide range of biological activities, including anti-inflammatory, antioxidant, and anti-cancer effects. Cancer involves uncontrolled cell growth; anti-cancer treatments include surgery, chemotherapy, radiation, immunotherapy, targeted therapy, and hormonal therapy.These treatments aim to destroy cancer cells, stop their growth, or prevent them from spreading to other body parts. Some anti-cancer drugs, isoliquiritigenin, butein, cardamonin, Sappanchalcone, licochalcone A, millepachine, xanthohumol,and Curcumin, are all-natural compounds studied for their potential anti-cancer properties. These are taken from PubChem which are depicted in Fig. 1. Flavonoids show anti-cancer possible, cheminformatics and graph theory use mathematics to model and care medicine discovery1. The combination of mathematics within cheminformatics promotes effective examination and analysis of chemical information, so driving progress in drug development and molecular investigation2–8. Quantitative Structure–Property Relationship (QSPR) analysis establishes mathematical models linking molecular structure with physicochemical or biological properties. By using topological indices and computational descriptors, QSPR and QSAR help predict compound behavior, optimize drug design, and reduce experimental efforts. It provides an efficient, reliable approach for understanding and forecasting molecular activity and functionality9–16. Chemical graph theory is vital in medicine design, computational chemistry, and materials science, aiding in the discovery and comprehension of novel chemical compounds3,17–38. These indices maintenance structure–property relationship studies, medicine development, and the estimate of molecules’ biological efficacy and physical characteristics14,39–49. Harry Wiener first introduced the idea of topological indices in 194750. Investigations are conducted on topological indices related to distance, resistance, and eccentricity51–60. Some of the topological indices examined in this paper include the first and second Zagreb indices (M1 and M2) indices, hyper (H) index, Forgotten (F) index, first and second Redefined (ReZG1, ReZG2) index, symmetric division (SSD) index, harmonic (H) index, (ABC) index, These indices yield numerical values derived from the molecular structures of the drugs61. Molecular structures of naturally occurring anti-cancer chalcones drugs. Gutman, in 197262, defined and formulated the first and second Zagreb indices as In 2016, Gao63 presented first andsecond redefined Zagreb indices as follows: We defined CReZG1(G), and CReZG2(G), redefined reverse Zagreb index Sum connectivity index (SCI(G))64 The Randic index was invented in 1976 by Milan Randi ´c65 and is defined a Vukičević et al. suggested the symmetric division degree by66. The harmonic index (H) familiarized by Zhong in 201267 Similarly, many other indices are essential in quantitative structure–property relationship (QSPR) analysis. The reader is referred to these references for further information and an in-depth review of the topological indices68–74. The objective of this study is to develop a predictive QSPR framework using degree-based topological indices to correlate the molecular structures of naturally occurring anticancer chalcones with their physicochemical properties. The study aims to assess the accuracy, robustness, and reliability of these indices in modeling and predicting key molecular features relevant to their pharmacological potential. Materials and methods This study employs simple graphs to represent the molecular structures of naturally occurring anti-cancer chalcones. Each molecule is modeled as a simple connected graph, where atoms correspond to vertices and chemical bonds correspond to edges. The topological features of these molecular graphs were analyzed to derive various topological indices that quantitatively describe molecular structure. Using computational methods, vertex partitioning and edge partitioning techniques were applied to compute topological indices such as first and second Zagreb indices, first and second redefined Zagreb indices, Sum connectivity index, symmetric division indices, Randic indices, Harmonic indices. The molecular structures of selected chalcones were obtained from the PubChem. The eight chalcone derivatives were selected based on their well-documented anticancer activity in literature, structural diversity across the chalcone scaffold, availability of reliable molecular and physicochemical data for modeling, and their pharmacological relevance as representative natural anticancer agents. These optimized structures were then converted into graph-theoretical representations for index computation. All numerical calculations were performed using MS Excel software to ensure accuracy and consistency. The computed indices were statistically correlated with the known anti-cancer activities of the compounds through regression analysis in MS Excel, enabling the development of predictive structure–property models. Cross-validation and statistical parameters such as the correlation coefficient (R2), and standard error (SE) were used to evaluate the robustness, accuracy, and predictive performance of the developed models. Results and discussion Drugs made from naturally occurring anti-cancer chalcones are modeled using degree-based topological indices. These indices and the compounds’ physicochemical characteristics are strongly correlated, according to QSPR research. The naturally found chalcone-based drugs that have anti-cancer properties exhibit five significant physicochemical characteristics: molecular weight, complexity, molar refractivity, polarizability, and surface tension. These properties are evaluated for each of the selected compounds. A linear regression model is applied in this study to examine the relationship between the properties and the calculated topological indices. where A, B → constants, P → Physical property of the drug, TI → topological descriptor. The linear regression equation discussed in this study is used to construct a regression model based on the previously computed degree-based topological indices. These indices, derived from the molecular networks of eight naturally occurring anti-cancer chalcones, serve as independent variables, while the corresponding physicochemical properties such as molecular weight, complexity, molar refractivity, polarizability, and surface tension are treated as dependent variables. The regression model is developed using Microsoft Excel, with the constants A and B in Eq. (1) determined from the data summarized in Tables 1 and 2. Table 1Topological index values for each drug molecule.NameM1(G)M2(G)ReZG1(G)ReZG2(G) RA(G)S(G)SSD(G)H(G)Isoliquiritigenin1601983137.114.45214.60579.33313.533Butein1642033238.21614.99015.13080.66614.1Cardamonin1782163741.515.59915.77892.08314.466Sappanchalcone1822213541.28316.13716.30293.41614.966Licochalcone A2523114755.39521.16421.332134.41619.204Millepachine27433748.7559.49721.77822.274146.2519.619Xanthohumol2563104856.04021.62521.800137.7519.642Curcumin2482984754.62821.25921.442132.83319.371 Topological index values for each drug molecule. Table 2Values of the properties.DrugsMolecular weightComplexityMolar RefractivityPolarizabilitySurface tensionIsoliquiritigenin256.2533172.828.870.3Butein272.2536774.629.681.2Cardamonin270.2834677.630.7210.7Sappaanchalcone286.2838179.431.5209.1Licochalcone A338.448810039.8289Millepachine350.4582103.240.9307.4Xanthohumol354.4515102.840.7284.9Curcumin368.450710441.2287.9 Values of the properties. Statistical parameters used in the QSPR analysis are presented in Tables 3, 4, 5, 6, 7, 8, 9, and 10. The standard error estimations are listed in Table 11, and a comparison between predicted and actual values is provided in Tables 12, 13, 14, 15, and 16. Figure 1 offers a visual representation of the molecular structures of the studied compounds. Table 3Statistical metrics of the linear QSPR model for M1 (G).Phy. pnABrr2FPIndicatorMW8116.54390.9126660.9582830.9183060.9183060.000176SignificantCom822.29161.9478810.9793360.9591140.69830.000021SignificantMR825.549720.2975510.9861560.972503212.20610.000000SignificantPo810.052190.118310.9864210.973025216.43180.000000SignificantST8-172.9131.8225210.9093980.82700528.683140.001735Significant Statistical metrics of the linear QSPR model for M1 (G). Table 4Statistical metrics of the linear QSPR model for M1 (G).Phy. pnAbRr2fpIndicatorMW8117.1740.7446360.9463290.89553951.437830.000371SignificantCom818.093071.6104370.9800080.960416145.57570.000019SignificantMR825.438440.2439790.9787090.95787136.41760.000023SignificantPo89.9983170.0970460.9793420.959112140.74070.000021SignificantST8-170.9221.4841810.8963650.8034724.529620.002571Significant Statistical metrics of the linear QSPR model for M1 (G). Table 5Statistical metrics of the linear QSPR model for ReZG1 (G).Phy. pnAbRr2fpIndicatorMW882.813635.6305480.959380.9204169.386010.000162SignificantCom8-32.305911.590010.9456090.89417650.697880.000386SignificantMR814.632161.8337460.9862370.972664213.48750.000000SignificantPo85.7224570.7288420.9861290.972451211.79510.000000SignificantST8-254.83211.601410.9393990.8824745.050910.000531Significant Statistical metrics of the linear QSPR model for ReZG1 (G). Table 6Statistical metrics of the linear QSPR model for ReZG2 (G).Phy. pnAbrr2fpIndicatorMW887.924144.6741160.9572940.91641365.78110.000189SignificantCom8-37.60649.9511570.9759040.952389120.02080.000034SignificantMR816.148751.5253390.9860850.972365211.11270.000000SignificantPo86.315520.6064650.9863080.972803214.60880.000000SignificantST8-232.8059.3909910.9140220.83543730.460230.001488Significant Statistical metrics of the linear QSPR model for ReZG2 (G). Table 7Statistical metrics of the linear QSPR model for RA (G).Phy. pnAbrr2fpIndicatorMW868.4832413.256740.9798750.960156144.5870.000020SignificantCom8-61.50327.271530.9652310.9316781.809470.000102SignificantMR810.760034.2741680.9972130.9944341071.90.000000SignificantPo84.1670021.6997090.9976290.9952641260.7910.000000SignificantST8-255.73625.757020.9047510.81857527.071430.002009Significant Statistical metrics of the linear QSPR model for RA (G). Table 8Statistical metrics of the linear QSPR model for S (G).Phy. pnABRr2fpIndicatorMW870.5302912.998650.9774510.95541128.55860.000028SignificantCom8-61.818226.984150.9716130.944031101.20210.000055SignificantMR811.30284.1972620.9962430.992499793.91160.000000SignificantPo84.3855131.6689820.9965720.993157870.7720.000000SignificantST8-253.43925.3460.9057440.82037327.40250.001948Significant Statistical metrics of the linear QSPR model for S (G). Table 9Statistical metrics of the linear QSPR model for SSD (G).Phy. pnAbrr2fPIndicatorMW8139.83221.5366680.9604950.92255171.470510.00015SignificantCom75.281063.2503610.9728240.946387105.91230.000049SignificantMR833.159690.5008350.9881260.976393248.1570.000004SignificantPo13.081490.1991060.9882360.976611250.53430.000000SignificantST888-130.0863.1014210.9212450.84869333.654520.00115Significant Statistical metrics of the linear QSPR model for SSD (G). 8 8 8 Table 10Statistical metrics of the linear QSPR model for H (G).Phy. pnABrr2FpIndicatorMW845.7388315.794910.9829980.966285171.96460.000012SignificantCom8-105.20632.309990.9628530.92708576.287970.000125SignificantMR83.6577435.0788210.9977020.9954091300.7650.000000SignificantPo81.3417892.0197460.9981430.9962891610.7480.000000SignificantST8-295.45230.423180.8997870.80961625.51530.002331Significant Statistical metrics of the linear QSPR model for H (G). Table 11Standard error of estimation for the drug’s physicochemical properties.DrugsMolecular weightComplexityMolar RefractivityPo larizabilitySurface tensionM1(G)13.9245220.575992.5593241.00763242.63853M2(G)15.745720.242243.1679391.24058145.44653ReZG1(G)13.7440733.09712.5518421.01830135.14469ReZG2(G)14.0849722.199962.565761.01178741.58644RA(G)9.72449726.595161.1515150.42222943.66515SCI(G)10.2873924.069761.3367110.50752543.44821SSD(G)13.5579123.557782.3714150.9382739.87632H(G)8.94528327.472941.0458270.37374844.73019 Standard error of estimation for the drug’s physicochemical properties. Table 12Comparison between molecular weight predicted by the regression model and the actual values.NameMwM1(G)M2(G)ReZG1(G)ReZG2(G)RA(G)SCI(G)SSD(G)H(G)Isoliquiritigenin256.25262.570264.612257.361261.334260.070260.376261.741259.491Butein272.25266.221268.335262.991266.550267.202267.200263.789268.447Cardamonin270.28278.998278.015291.144281.900275.275275.623281.333274.228Sappanchalcone286.28282.649281.739279.883280.886282.407282.434283.382282.125Licochalcone A338.4346.536348.756347.449346.847349.049347.817346.385349.064Millepachine350.4366.614368.116357.303366.020357.189360.062364.570355.619Xanthohumol354.4350.186348.011353.080349.862355.160353.901351.508355.982Curcumin368.4342.885339.076347.449343.262350.308349.247343.952351.702 Comparison between molecular weight predicted by the regression model and the actual values. Table 13Comparison between complexity predicted by the regression model and the actual values.NameComM1(G)M2(G)ReZG1(G)ReZG2(G)RA(G)SCI(G)SSD(G)H(G)Isoliquiritigenin331333.953336.960326.984331.582332.625332.285182.580332.045Butein367341.744345.012338.574342.687347.297346.452186.913350.365Cardamonin346369.014365.947396.524375.367363.906363.938224.022362.190Sappanchalcone381376.806374.000373.344373.207378.578378.077228.355378.345Licochalcone A488513.158518.939512.425513.638515.672513.808361.619515.275Millepachine582556.011560.810532.707554.458532.416539.227400.084528.684Xanthohumol515520.949517.329524.015520.056528.244526.436372.456529.427Curcumin507505.366498.003512.425506.005518.262516.776356.474520.671 Comparison between complexity predicted by the regression model and the actual values. Table 14Comparison between molar Refractivity predicted by the regression model and the actual values.NameMRM1(G)M2(G)ReZG1(G)ReZG2(G)RA(G)SCI(G)SSD(G)H(G)Isoliquiritigenin72.873.15873.74671.47872.73972.53072.60472.89272.389Butein74.674.34874.96673.31274.44174.83074.80773.56075.269Cardamonin77.678.51478.13882.48179.45077.43377.52779.27877.128Sappanchalcone79.479.70479.35878.81379.11979.73279.72779.94679.667Licochalcone A100100.533101.316100.818100.645101.219100.839100.480101.191Millepachine103.2107.079107.659104.027106.902103.843104.793106.407103.299Xanthohumol102.8101.723101.072102.652101.629103.189102.803102.150103.416Curcumin10499.34298.144100.81899.475101.625101.30099.687102.040 Comparison between molar Refractivity predicted by the regression model and the actual values. Table 15Comparison between polarizability predicted by the regression model and the actual values.NamePoM1(G)M2(G)ReZG1(G)ReZG2(G)RA(G)SCI(G)SSD(G)H(G)Isoliquiritigenin28.828.98229.21328.31728.81528.73128.76128.87728.675Butein29.629.45529.69929.04529.49229.64629.63729.14329.820Cardamonin30.731.11130.96032.69031.48430.68130.71931.41630.559Sappanchalcone31.531.58531.44531.23231.35231.59531.59331.68131.569Licochalcone A39.839.86640.18039.97839.91140.14039.98839.84540.129Millepachine40.942.46942.70341.25442.39841.18341.56042.20140.967Xanthohumol40.740.34040.08340.70740.30240.92340.76940.50841.014Curcumin41.239.39338.91839.97839.44540.30140.17239.52940.466 Comparison between polarizability predicted by the regression model and the actual values. Table 16Comparison between Surface Tension predicted by the regression model and the actual values.NameStM1(G)M2(G)ReZG1(G)ReZG2(G)RA(G)SCI(G)SSD(G)H(G)Isoliquiritigenin70.3118.690122.946104.812115.601116.504623.617115.959116.265Butein81.2125.980130.367116.413126.081130.362636.924120.093133.515Cardamonin210.7151.496149.661174.420156.921146.048653.348155.502144.650Sappanchalcone209.1158.786157.082151.217154.883159.905666.629159.636159.861Licochalcone A289286.362290.658290.434287.409289.386794.120286.795288.795Millepachine307.4326.458329.247310.737325.931305.200817.996323.497301.420Xanthohumol284.9293.652289.174302.036293.466301.260805.982297.135302.120Curcumin287.9279.072271.364290.434280.206291.832796.908281.885293.875 Comparison between Surface Tension predicted by the regression model and the actual values. Regression model of M1(G) Molecular weight = 116.5439 + 0.912666 [M1 (G)]. Complexity = 22.2916 + 1.947881 [M1 (G)]. Molar Refractivity = 25.54972 + 0.297551[M1 (G)]. Polarizability = 10.05219 + 0.11831[M1 (G)]. Surface tension = −172.913 + 1.822521[M1 (G)]. Regression model of M2(G) Molecular weight = 117.174 + 0.744636[M2 (G)]. Complexity = 18.09307 + 1.610437[M2 (G)]. Molar Refractivity = 25.43844 + 0.243979[M2 (G)]. Polarizability = 9.998317 + 0.097046[M2 (G)]. Surface tension = −170.922 + 1.484181[M2 (G)]. Regression model of ReZG1(G) Molecular weight = 82.81363 + 5.630548[ReZG1 (G)]. Complexity = −32.3059 + 11.59001[ReZG1 (G)]. Molar Refractivity = 14.63216 + 1.833746[ReZG1 (G)]. Polarizability = 5.722457 + 0.728842[ReZG1 (G)]. Surface tension = −254.832 + 11.60141[ReZG1 (G)]. Regression model of ReZG2(G) Molecular weight = 87.92414 + 4.674116[ReZG2 (G)]. Complexity = −37.6064 + 9.951157[ReZG2 (G)]. Molar Refractivity = 16.14875 + 1.525339[ReZG2 (G)]. Polarizability = 6.31552 + 0.606465[ReZG2 (G)]. Surface tension = −232.805 + 9.390991[ReZG2 (G)]. Regression model of R(G) Molecular weight = 68.48324 + 13.25674 [RA (G)]. Complexity = −61.503 + 27.27153[RA (G)]. Molar Refractivity = 10.76003 + 4.274168[RA (G)]. Polarizability = 4.167002 + 1.699709[RA (G)]. Surface tension = −255.736 + 25.75702[RA (G)]. Regression model of SCI(G) Molecular weight = 70.53029 + 12.99865[S (G)]. Complexity = −61.8182 + 26.98415[S (G)]. Molar Refractivity = 11.3028 + 4.197262[S (G)]. Polarizability = 4.385513 + 1.668982[S (G)]. Surface tension = –253.439 + 25.346[S (G)]. Regression model of SSD (G) Molecular weight = 139.8322 + 1.536668 [SSD (G)]. Complexity = −75.28106 + 3.250361 [SSD (G)]. Molar Refractivity = 33.15969 + 0.500835 [SSD (G)]. Polarizability = 13.08149 + 0.199106 [SSD (G)]. Surface tension = −130.086 + 3.101421 [SSD (G)]. Regression model of H(G) Molecular weight = 45.73883 + 15.7949[H(G)]. Complexity = −105.206 + 32.30999[H (G)]. Molar Refractivity = 3.657743 + 5.078821[H (G)]. Polarizability = 1.341789 + 2.019746[H (G)]. Surface tension = −295.452 + 30.42318[H (G)]. Table 1 shows the results of the study that computed nine degree-based topological indices for eight naturally occurring anti-cancer chalcones drugs. Table 17 displays the coefficients of correlation for the indices mentioned earlier versus five physicochemical parameters. When the first Zagreb index is examined row-wise, the correlation coefficient is higher concerning the polarizability (0.986421) for M1 (G). The second Zagreb index has a higher correlation with complexity (0.979336), and the redefined first Zagreb index has a higher correlation with molar refractivity (0.986237); and the redefined second Zagreb index has a higher correlation with Polarizability (0.986308); and the randic index has a higher correlation with Polarizability (0.997629), Sum connectivity index has a higher correlation with Polarizability (0.996572), symmetric division degree index has a higher correlation with Polarizability (0.988236), and harmonic index has a higher correlation with Polarizability (0.998143). Figure 2 presents graphical visualization of the correlation coefficients of physico-chemical properties and topological descriptors. Table 17Correlation coefficients of the properties and topological indices.DrugsMolecular weightComplexityMolar RefractivityPolarizabilitySurface tensionM1(G)0.9582830.9793360.9861560.9864210.909398M2(G)0.9463290.9800080.9787090.9793420.896365ReZG1(G)0.959380.9456090.9862370.9861290.939399ReZG2(G)0.9572940.9759040.9860850.9863080.914022RA(G)0.9798750.9652310.9972130.9976290.904751SCI(G)0.9774510.9716130.9962430.9965720.905744SSD(G)0.9604950.9728240.9881260.9882360.921245H(G)0.9829980.9628530.9977020.9981430.899787 Correlation coefficients of the properties and topological indices. Physicochemical properties and T1s. Conclusions A comprehensive predictive topological modeling framework was employed to investigate the quantitative structure–property relationships (QSPR) of naturally occurring anticancer chalcones. A series of degree-based topological indices, including the first and second Zagreb indices, first and second redefined Zagreb indices, sum-connectivity index, symmetric division index, Randić index, and harmonic index, were computed to represent key molecular structural features. Statistical and regression analyses were performed to correlate these indices with experimentally reported physicochemical properties of chalcones. The application of these degree-based topological indices combined with statistical parameters in linear QSPR models has revealed strong correlations between molecular structure and various physicochemical properties of naturally occurring anticancer chalcones. These results underscore the predictive capability of topological indices and their potential utility in quantitative molecular design. The significant correlation coefficients obtained demonstrate the reliability of these descriptors in capturing structural features that influence physicochemical behavior. Overall, this study establishes a robust and computationally efficient framework for predicting key molecular properties of bioactive compounds. The insights gained can assist pharmaceutical researchers in rational drug design, facilitating the development and optimization of novel anticancer agents derived from natural chalcone scaffolds. Author contributions The authors Abdullah Abdul Razzaq Obaid, Abdul Hakeem, Asad Ullah, Shahid Zaman, Hijaz Ahmad, Parvez Ali, and Melaku Berhe Belay have equally contributed to this manuscript in all stages, from conceptualization to the write-up of the final draft. Data availability All data generated or analyzed during this study are included within this article. Competing interests The authors declare no competing interests. Consent for publication All authors have approved the manuscript and given consent for publication. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Contributor Information Asad Ullah, Email: dr.asadullah@kiu.edu.pk. Shahid Zaman, Email: zaman.ravian@gmail.com. Melaku Berhe Belay, Email: melaku.berhe@aastu.edu.et. References Associated Data Data Availability Statement All data generated or analyzed during this study are included within this article.