Pith. sign in

REVIEW 2 major objections 5 minor 16 references

A comparative study of sum-connectivity and product-connectivity Gourava indices for benzenoid hydrocarbons

T0 review · 2 major / 5 minor · reviewed 2026-08-12 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read The sum-connectivity Gourava index predicts pi-electronic energies of 30 benzenoid hydrocarbons with correlation 0.9997, outperforming the product-connectivity variant.

desk verdict Competent but overinterpreted: SGO's better fit is real, but the 'across edge types' superiority claim lacks uncertainty and relies on an in-sample reference with an insignificant intercept. read the letter →

arxiv 2608.08099 v1 pith:7GDAQJN6 submitted 2026-08-08 physics.chem-ph

Nagesh H.M , Vijaya Chandra Kumar U , Azghar Pasha B , Narahari N This is my paper · ORCID
classification physics.chem-ph MSC 05C9092E10
keywords sum-connectivityGouravaindexproduct-connectivitybenzenoidhydrocarbonspi-electronicenergyQSPRtopologicalregressionanalysisdegeneracy
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper argues that the sum-connectivity Gourava index (SGO), a vertex-degree-based descriptor, is a better predictor of pi-electronic energies for benzenoid hydrocarbons than its product-connectivity cousin (PGO). Using 30 benzenoid structures, it shows SGO correlates with Eπ at R = 0.9997 versus R = 0.9970 for PGO, and that the edge-type coefficients implied by SGO lie closer to the least-squares optimum than those implied by PGO. This matters because QSPR models rely on descriptors that combine strong correlation with the ability to tell similar molecules apart, and SGO also shows low degeneracy and high structural sensitivity on tree datasets. The study positions SGO as a practical, stable choice for future quantitative structure-property modeling of aromatic hydrocarbons.

What carries the argument

The machine that carries the argument is the decomposition of each Gourava index into the three edge types that occur in benzenoid graphs: e22, e23, and e33, with vertices of degree 2 or 3. SGO assigns weights 1/√8, 1/√11, and 1/√15 to these edges, while PGO assigns 1/4, 1/√30, and 1/√54. This turns each index into a simple linear combination of edge counts, which can be substituted into the regression Eπ = A + B e22 + C e23 + D e33 to obtain a predicted coefficient set that is then compared against the direct least-squares fit. The comparison of those coefficient sets is what supports the paper's claim that SGO's weighting scheme is superior.

What would settle it

Recompute the edge-type regression on a fresh set of benzenoid hydrocarbons with known Eπ (for instance, a subset of larger catacondensed or pericondensed systems) and check whether SGO-derived coefficients remain closer to the new least-squares optimum than PGO-derived coefficients. If the two coefficient sets alternate in closeness across new data, the paper's ranking is an artifact of the original 30 molecules.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is that the sum-connectivity Gourava index carries essentially all the predictive information that the product-connectivity variant does, plus a weighting of edge types that better matches the optimal linear model. When the pi-electronic energy is written as a linear function of the three edge counts (e22, e23, e33), a least-squares fit over the 30 benzenoids yields coefficients approximately 1.342, 1.145, and 1.061; substituting SGO's edge weights produces coefficients 1.388, 1.183, 1.013, while PGO's weights produce 1.683, 1.229, 0.916. The SGO-derived coefficients are closer in every edge class, and its regression intercept (-0.460) is far smaller in magnitude than PGO's (-2.538), which the paper reads as evidence that SGO's weighting scheme is the more faithful description of how these molecules store pi energy. The paper further supports SGO by showing it matches the best modern indices on degeneracy tests and offers a favorable sensitivity-stability trade-off.

Load-bearing premise

The ranking of the two indices rests on the in-sample least-squares fit of equation (12) to the same 30 molecules serving as the 'optimal' benchmark, and that benchmark's intercept is statistically insignificant; if those coefficients are noisy, the conclusion that SGO's weights are the better approximation may not transfer to new benzenoid sets.

Editorial extensions

If this is right

  • If the claim is right, the sum-connectivity Gourava index should be preferred over the product-connectivity variant in QSPR models for benzenoid hydrocarbons.
  • The edge-type coefficients in equation (13) provide a ready-made linear model for estimating pi-electronic energies of larger benzenoids without re-fitting the regression.
  • Since SGO matches modern indices like Sombor and diminished Sombor on degeneracy while offering higher sensitivity, it is a competitive choice for isomer discrimination in QSAR studies.
  • The near-perfect but non-unit correlation with established descriptors means SGO adds information to multi-descriptor QSAR models rather than merely duplicating existing indices.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The same edge-type comparison could be tested on other fused polycyclic aromatic systems beyond benzenoids, such as fluoranthenes or azulenes, where degree-2 and degree-3 vertices still dominate the graph.
  • Because the benchmark is in-sample, a natural extension is to train on a subset of the 30 molecules and validate on the remainder; the paper's leave-one-out results are reported, but the edge-type coefficient comparison is not itself cross-validated.
  • The 74% sensitivity advantage over DSO suggests SGO may be useful in similarity-based molecular searches, though this has not been tested directly on property prediction.
  • One could compute SGO and PGO for a larger set of benzenoids with known experimental properties such as boiling point or heat of formation to see whether the predictive advantage generalizes beyond pi-electronic energy.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

2 major / 5 minor

Summary. The paper compares the sum-connectivity Gourava index (SGO) and the product-connectivity Gourava index (PGO) as descriptors for the pi-electronic energy (E_pi) of 30 benzenoid hydrocarbons. For benzenoid graphs with only degree-2 and degree-3 vertices, the indices reduce to fixed linear combinations of the edge-type counts e22, e23, and e33 (Eqs. (1)-(2)), and closed forms in terms of n, h, and r are derived (Eqs. (6)-(7)). Linear regressions of E_pi on SGO and on PGO give R = 0.9997 and R = 0.9970, respectively (Eqs. (9)-(10)). The paper argues that SGO is superior 'across molecular edge types' by comparing the coefficient vectors implied by the one-index regressions (Eqs. (13)-(14)) to the coefficients of a least-squares model in e22, e23, e33 (Eq. (12)). The remainder of the paper validates SGO through intercorrelation analysis with eight standard indices, degeneracy tests on octane isomers, nonane isomers, and all trees of order 10, and structure-sensitivity analysis on trees of order 10.

Significance. If the central comparative claim is made statistically sound, the paper offers a useful and clearly presented comparison of two recently introduced Gourava indices. The algebraic derivations in Eqs. (1)-(7) are transparent and reproducible, the reported regression numbers are internally consistent with the table values, and the primary regressions are accompanied by leave-one-out cross-validation statistics. The degeneracy and structure-sensitivity analyses on standard datasets (octane, nonane, order-10 trees) provide concrete, falsifiable evidence about the discriminative power of SGO. The main weakness is that the paper's emphasized 'across edge types' superiority of SGO rests on a pointwise comparison of coefficients against an in-sample least-squares reference without uncertainty quantification; this specific part of the conclusion is not yet statistically supported, even though the direct fit comparison (R and cross-validation) does favor SGO.

major comments (2)
  1. [§5, Table 2, Eqs. (12)-(14)] The conclusion that SGO is superior 'across molecular edge types' rests on comparing the point estimates of the coefficients in Eqs. (13) and (14) with the 'optimal' coefficients obtained from the least-squares fit in Eq. (12). This comparison is purely pointwise: no standard errors or confidence intervals are given for the differences, and the coefficients in (13)-(14) inherit uncertainty from Eqs. (9)-(10) that is not propagated. The problem is compounded by the fact that Eq. (12) is fitted to the same 30 molecules used in the comparison and its intercept is statistically insignificant (-0.024 +/- 0.054); the paper itself states that the term 'can be omitted,' yet the retained model is used as the reference. If the intercept is dropped, the least-squares coefficients change, and the stated closeness ranking may not persist. Without a statistical measure of the distance between the coefficient vectors, the claim that the weighting scheme in (13) 'provides a superior representation of the relative contribution of edge-types' is not substantiated.
  2. [§5, Eqs. (9)-(10) and Eq. (8)] The direct comparison of the two one-index regressions is more secure: SGO yields R = 0.9997 versus R = 0.9970 for PGO, with better leave-one-out cross-validation statistics (R_cv = 0.9996 vs 0.9960; S_cv = 0.194 vs 0.644). However, the paper does not report any formal test of whether this difference is significant despite the very high intercorrelation between SGO and PGO (R = 0.9983, Eq. (8)). A formal comparison for correlated predictors (e.g., a Williams-type test or a test of residual variance ratio) would strengthen the abstract's claim of a 'markedly better fit' and would address the concern that the R difference might be within sampling variability. This is a load-bearing part of the paper's central comparative claim.
minor comments (5)
  1. [§5.2, Figure 6] The caption of Figure 6 reads 'Correlation matrix', but the figure presents degeneracy comparisons; the caption should be corrected.
  2. [§5.2, Table 3] The symbol N is used for the number of isomers in the datasets (octane N=18, nonane N=35, order-10 trees N=106) but earlier in the paper N=30 denotes the number of benzenoid molecules; this double use of N is confusing and should be disambiguated.
  3. [References] References [10] and [11] are incomplete: they lack article titles, which should be supplied for completeness and reproducibility.
  4. [Data Availability] The Data Availability statement says the data are cited within the text; since the full dataset (E_pi, SGO, PGO for 30 compounds) appears only in Table 1, a machine-readable version or explicit statement of reproducibility would be helpful.
  5. [§5.1] The statement that the correlations in [-0.9923, -0.8936] 'indicate complementary structural information' is suggestive but not statistically tested; reporting confidence intervals for these correlations, or a test of whether they differ from -1, would make the claim more precise.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the regression fits use external E-pi data and independently cited index definitions; the in-sample coefficient comparison is a benchmark, not a self-referential reduction.

full rationale

I walked the derivation chain and found no step where a claimed prediction reduces by construction to a fitted input or to a self-citation. The Gourava index definitions (Kulli, Refs. [4,5]) and the pi-electronic energy values (Ref. [7]) are external to this paper. Eq. (1) and Eq. (2) are direct evaluations of those definitions for the three possible benzenoid edge types. Eqs. (6)-(7) are algebraically derived from the standard edge-count identities (3)-(5), which are cited from the benzenoid literature. Equations (9) and (10) are ordinary least-squares fits of E-pi against SGO and PGO on the 30-molecule dataset; they are statistical fits, but the paper does not disguise them as parameter-free predictions. Eq. (12) is a separate least-squares fit of E-pi against the three edge-type counts on the same dataset, and Eqs. (13)-(14) are obtained by substituting the fixed edge-weight expressions (1)-(2) into the fitted equations (9)-(10). The comparison in Section 5 measures how close those implied edge-type coefficients lie to the unconstrained benchmark Eq. (12). This is an in-sample comparison and its statistical strength is limited, especially because the intercept of Eq. (12) is statistically insignificant and the comparison is pointwise without propagated uncertainties; however, that is a robustness concern, not circularity. No load-bearing argument relies on a citation to the authors' own prior work, and no quantity is defined in terms of the conclusion it is used to support. The degeneracy, sensitivity, and intercorrelation analyses are independent calculations on external datasets. The finding is therefore that the central comparative claim, while having in-sample limitations, is not circular.

Assumptions & free parameters 3 free parameters · 5 assumptions · 0 invented entities

The central claim depends on three fitted regression models (SGO vs E_pi, PGO vs E_pi, and the four-parameter benchmark Eq. 12) and on standard benzenoid graph assumptions. No new entities are introduced. The main methodological axiom, that closeness to an in-sample least-squares benchmark ranks descriptors, carries the Table 2 conclusion.

free parameters (3)
  • Optimal regression coefficients A, B, C, D in Eq. (12) = A=-0.024018, B=1.342149, C=1.145181, D=1.060790
    Fitted to the 30 E_pi values and used as the in-sample 'optimal' reference in Table 2 to rank SGO over PGO.
  • Regression slope and intercept for SGO, Eq. (9) = slope 3.925, intercept -0.460
    Fitted to E_pi and used to derive the edge-coefficient model Eq. (13) that is compared in Table 2.
  • Regression slope and intercept for PGO, Eq. (10) = slope 6.731, intercept -2.538
    Fitted to E_pi and used to derive the edge-coefficient model Eq. (14) that is compared in Table 2.
assumptions (5)
  • domain assumption Benzenoid graphs have vertices only of degree 2 and 3, so edge types are e22, e23, e33.
    Invoked in Section 2 to derive the simplified formulas (1)-(2). Standard structural fact for benzenoid hydrocarbons.
  • domain assumption The edge-count formulas e22=n-2h-r+2, e23=2r, e33=3h-r-3 are correct for all benzenoids.
    Cited from refs [10,11] and used to derive the vertex-based formulas (6)-(7) that generate Table 1.
  • domain assumption The pi-electronic energies from ref. [7] are accepted as ground truth for the regression analyses.
    The dataset of 30 E_pi values is taken as published without independent verification or error bars.
  • ad hoc to paper Comparing the implied edge-coefficients of a one-index regression with the coefficients of the full least-squares model is a valid way to rank descriptors.
    This criterion is introduced in Section 5 via Table 2 but no formal statistical justification is given for why nearness to the coefficients of Eq. (12) measures descriptor quality.
  • domain assumption Degeneracy and structure sensitivity are appropriate validation metrics for descriptor quality.
    Follows the standard evaluation protocols of refs [12-16] used in the field.

how reviews work

0 comments
Cite this review

Pith. "Pith review of A comparative study of sum-connectivity and product-connectivity Gourava indices for benzenoid hydrocarbons." pith.science (2026). https://pith.science/paper/7GDAQJN6

@misc{pith2026260808099,
  author       = {Pith},
  title        = {Pith review of: A comparative study of sum-connectivity and product-connectivity Gourava indices for benzenoid hydrocarbons},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/7GDAQJN6}},
  note         = {Machine review of arXiv:2608.08099}
}
abstract

This study evaluates the sum-connectivity ($SGO$) and product-connectivity ($PGO$) Gourava indices as molecular descriptors for benzenoid hydrocarbons. Using a dataset of 30 benzenoid structures, we compare least-squares regression models for predicting $\pi$-electronic energies ($E_{\pi}$) and find that $SGO$ yields a markedly better fit than $PGO$ across molecular edge types. The indices are further assessed using three validation designs: (i) correlation analysis, in which $SGO$ exhibits strong yet non-perfect inverse correlations with standard descriptors ($M_1, M_2, SO, DSO,$ and $ABS$; $r\in[-0.9923,-0.8936]$), suggesting complementary structural information; (ii) degeneracy analysis on Octane, Nonane, and order-$10$ tree datasets, where $SGO$ attains low degeneracy rates (22.22\%, 40.00\%, and 42.45\%); and (iii) structure-sensitivity analysis on trees of order $n=10$, showing 74\% higher sensitivity than $DSO$ while maintaining a high structure-abruptness ratio ($SA = 0.474386$). Overall, $SGO$ offers a favorable balance between discriminative power and numerical stability, supporting its applicability in QSPR modeling and related theoretical studies.

Figures

Figures reproduced from arXiv: 2608.08099 by the authors.

Figure 1
Figure 1. Representation of bays, fjords, fissures, and coves in benzenoid systems. [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Structures of 30 lower benzenoids. The structure of benzenoid hydrocarbons is most easily shown and studied using ben￾zenoid graphs, which are mathematical drawings made of hexagonal rings. In benzenoid graphs, every vertex has a degree of either 2 or 3 [9], resulting in only three types of edges: e22, e23, and e33. By introducing the number of these specific edges into the general def￾initions for Gourava indices, … view at source ↗
Figure 3
Figure 3. Scatter plot illustrating the linear correlation between the product-connectivity [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Scatter plots illustrating the relationships between [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]
Figure 5
Figure 5. Figure 5: Correlation matrix. The first observation is that the indices M1, M2, SO, DSO, ABS, and N are highly correlated with each other. This is particularly expressed in the pair M1 and SO, where the correlation coefficient is 0.9983, indicating that they are very similar to …
Figure 6
Figure 6. Figure 6: Correlation matrix. As established in [12], a lower degeneracy percentage indicates a higher capability of a topological index to discriminate among non-isomorphic graphs. The comparative degeneracy percentages obtained from the graphical data are organized in [PITH_F…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

16 extracted references · 13 canonical work pages

  1. [1]

    A. Ali, I. Gutman, B. Furtula, I. Redˇ zepovi´ c, T. Doˇ sli´ c, Z. Raza, Extremal Results and Bounds for Atom–Bond Sum–Connectivity Index, MATCH Commu- nications in Mathematical and in Computer Chemistry.92(2)(2024) 271–314. https://doi.org/10.46793/match.92-2.271A

    work page doi:10.46793/match.92-2.271a 2024

  2. [2]

    Gutman, Geometric approach to degree-based topological indices: Sombor indices, MATCH Communications in Mathematical and in Computer Chemistry.86(1)(2021) 11–16

    I. Gutman, Geometric approach to degree-based topological indices: Sombor indices, MATCH Communications in Mathematical and in Computer Chemistry.86(1)(2021) 11–16

    work page 2021

  3. [3]

    Lin, Sombor-Type Indices for Certain Interconnection Networks, MATCH Com- munications in Mathematical and in Computer Chemistry.94(3)(2025) 701–724

    Z. Lin, Sombor-Type Indices for Certain Interconnection Networks, MATCH Com- munications in Mathematical and in Computer Chemistry.94(3)(2025) 701–724. https://doi.org/10.46793/match.94-3.701L

    work page doi:10.46793/match.94-3.701l 2025

  4. [4]

    V. R. Kulli, The Product Connectivity Gourava Index, Journal of Computer and Mathematical Sciences.8(6)(2017) 235-242. 14

    work page 2017

  5. [5]

    V. R. Kulli, On the Sum Connectivity Gourava Index, International Journal of Math- ematical Archive.8(6)(2017) 211-217

    work page 2017

  6. [6]

    Sarwar, S

    K. Sarwar, S. Kanwal, A. Razzaque, QSPR analysis of amino acids for the family of Gourava indices, PLoS One.20(4)(2025). https://doi.org/10.1371/journal.pone.0319029

    doi:10.1371/journal.pone.0319029 2025

  7. [7]

    Luˇ ci´ c, N

    B. Luˇ ci´ c, N. Trinajsti´ c, B. Zhou, Comparison between the sum-connectivity index and product-connectivity index for benzenoid hydrocarbons, Chemical Physics Letters. 475(2009) 146–148. https://doi.org/10.1016/j.cplett.2009.05.022

    doi:10.1016/j.cplett.2009.05.022 2009

  8. [8]

    Hayat, S

    S. Hayat, S. J. F. Alanazi, M. B. Belay, et al., Predictive power of para- metric temperature-based topological indices with applications in structure– property modeling of anti-tuberculosis drugs, Scientific Reports.15(2025) 39828. https://doi.org/10.1038/s41598-025-23529-3

    doi:10.1038/s41598-025-23529-3 2025

Show all 16 references
  1. [9]

    Gutman, S

    I. Gutman, S. J. Cyvin, Introduction to the Theory of Benzenoid Hydrocarbons, Springer, Berlin, (1989)

    1989

  2. [10]

    J. Rada, O. Araujo, I. Gutman, Croat. Chem. Acta.74(2001) 225–235

    2001

  3. [11]

    S. J. Cyvin, B. N. Cyvin, J. Brunwoll, Croat. Chem. Acta.66(1993) 101

    1993

  4. [12]

    E. V. Konstantinova, The discrimination ability of some topological and information distance indices for graphs of unbranched hexagonal systems, J. Chem. Inf. Comput. Sci.36(1)(1996) 54–57. https://doi.org/10.1021/ci9502461

    1996 doi

  5. [13]

    Furtula, I

    B. Furtula, I. Gutman, M. Dehmer, On structure-sensitivity of degree- based topological indices, Appl. Math. Comput.219(17)(2013) 8973-8978. https://doi.org/10.1016/j.amc.2013.03.072

    2013 doi

  6. [14]

    Raki´ c, B

    M. Raki´ c, B. Furtula, A novel method for measuring the structure sensitivity of molecular descriptors, J. Chemom.33(7)(2019) e3138. https://doi.org/10.1002/cem.3138

    2019 doi

  7. [15]

    Kumar, S

    V. Kumar, S. Das, Comparative study between Nirmala and Sombor indices based on their applicability, degeneracy and smoothness, Eur. Phys. J. Plus.139(2024). https://doi.org/10.1140/epjp/s13360-024-05768-x

    2024 doi

  8. [16]

    Brezovnik, M

    S. Brezovnik, M. Dehmer, N. Tratnik, P. ˇZigert Pleterˇ sek, On the Wiener-like root-indices of graphs, Computational and Applied Mathematics.44:333(2025). https://doi.org/10.1007/s40314-025-03296-7. 15

    2025 doi

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.