nabu.science
MethodologyAboutContact

 

Reliability concern: This paper was retracted on 2026-05-26. See notice: 10.1038/s41598-026-54594-x.

  • Snapshot
  • Key Claims
  • Strengths & Limitations
  • Commentary
  • From the Authors
  • Author Note / Field Feedback
  • Dimension breakdown
  • Impact Profile
Free · About 5 minutes
Report #NBU-3DED85D9·Framework v5.5·Pipeline v0.4·Evaluated 22 Jul 2026·OECD: 1.2 Computer and information sciences·Subfield: Computational Theory and Mathematics·Basis: Full text·Type: Machine learning or AI study

Mathematical Modeling and Computation of NM-Polynomial Indices for Physicochemical Properties Prediction

Qasem M Tawhari et al.Qasem M Tawhari, Muhammad Naeem, Ali N. A. Koam et al.·2026·10.1038/s41598-026-39562-9 ↗

Primary claim

key physicochemical characteristics–including flash point, enthalpy of vaporization, boiling point, molar refractivity, molecular complexity, molar volume, surface tension and polarizability–can be accurately predicted using NM-polynomial indices

Tentative · strong reported metrics for eight endpoints, but small sample and reliability irregularities temper confidence
In brief

An incremental NM-polynomial QSPR study with useful worked descriptors, constrained by unresolved data and metric-reporting errors.

ReliabilityConcerns4 of 4 checks completeRead the 2 entries ↓
Fragile — the claim's weakest requirement is that the effect exists at all. Experimental. Every point below is traceable to a quoted passage, but this forecast has not yet been validated against known replication outcomes.
Abstract

Mitoxantrone and Doxorubicin are anthracycline-based chemotherapeutic agents widely used in cancer treatment. Mitoxantrone is primarily applied in the treatment of leukemia, lymphoma, and prostate cancer, whereas Doxorubicin is effective against breast cancer, lung cancer, and various solid tumors. Both drugs act by intercalating DNA and inhibiting topoisomerase II, resulting in cell cycle arrest and apoptosis in cancer cells. NM-polynomial indices provide a robust framework for predicting the physicochemical characteristics of drug molecules. In this study, NM-polynomial indices for Mitoxantrone and Doxorubicin were determined using the edge/connectivity partition approach. A comprehensive statistical analysis was performed to evaluate their predictive capability for physicochemical properties of polycyclic drug structures. Lasso, Ridge, Elastic Net, and Multiple Linear Regression models were implemented in Python, and model performance was assessed using correlation coefficients (R, [Formula: see text]), root mean squared error (RMSE), p-values, and 95% confidence intervals. Cross-validation metrics and external test set evaluations confirmed the predictive strength of the models. The results indicate that key physicochemical characteristics–including flash point, enthalpy of vaporization, boiling point, molar refractivity, molecular complexity, molar volume, surface tension and polarizability–can be accurately predicted using NM-polynomial indices. A Python-based computational tool was developed for efficient calculation of these indices, substantially reducing computation time and minimizing human error. Furthermore, predictive models were proposed, retaining only the most statistically significant ones for estimating uncalculated physicochemical properties.

Key Claims

Primary finding · Tentative

key physicochemical characteristics–including flash point, enthalpy of vaporization, boiling point, molar refractivity, molecular complexity, molar volume, surface tension and polarizability–can be accurately predicted using NM-polynomial indices

Fragile — the claim's weakest requirement is that the effect exists at all. Experimental. Every point below is traceable to a quoted passage, but this forecast has not yet been validated against known replication outcomes.

strong reported metrics for eight endpoints, but small sample and reliability irregularities temper confidence

Abstract; Tables 9–17
Secondary · Supported

NM-polynomial indices for Mitoxantrone and Doxorubicin were determined using the edge/connectivity partition approach

explicit edge partitions, formulas, and theorem-proof derivations are provided

Theorems 1–4; Tables 3–5
Secondary · Tentative

Cross-validation metrics and external test set evaluations confirmed the predictive strength of the models

validation metrics are reported, but protocol and CI inconsistencies weaken confirmation

Material and methods; Tables 9–17
Secondary · Unsupported

A Python-based computational tool was developed for efficient calculation of these indices, substantially reducing computation time and minimizing human error

tool is described and named, but no runtime or error-reduction benchmark is shown

Abstract; Python code section
Exploratory · Unsupported

the neighborhood-degree-based M-polynomial indices (NM-polynomial indices) exhibited superior predictive performance

comparative descriptor results are asserted but not displayed in the supplied results

Motivation

Derived from the full evaluation — not a separate score.

Strengths · Limitations

Strengths

#contribution

Worked NM-polynomial descriptor derivations

The paper gives explicit NM-polynomial formulas and derived indices for Mitoxantrone and Doxorubicin using edge/connectivity partitions. The theorem-proof format makes the graph-theoretic descriptor calculations inspectable within the article.

↳ Theorems 1–4; Tables 3–5

#methodological rigour

Recognizable validation workflow elements

The methods describe an 80/20 train-test split, VIF filtering, within-fold scaling, regularized regressions, and both cross-validation and external test metrics. These are appropriate components for an exploratory QSPR modelling pipeline.

↳ Material and methods, Model Development and Validation

#reporting

Failure case discussed explicitly

The Index of Refraction endpoint is reported as poorly predicted, with low or undefined correlations and an explanation that penalized models collapsed to constant predictions. This prevents the results section from being uniformly selective, even though the abstract remains more upbeat.

↳ Table 13; Discussion of results

Limitations

#methodological rigour

Data and metric irregularities affect reliability

The Amoxicillin row in Table 8 contains values that appear shifted across property columns, and several reported confidence intervals do not contain the corresponding R values. These issues affect the interpretability of the model-performance tables.

↳ Table 8; Tables 9, 13–15

#methodological rigour

Small validation base limits generalisation

The dataset contains 45 compounds and the external test set contains only 9 compounds, with no independent external dataset or scaffold-aware validation. Reported holdout metrics therefore have high uncertainty for claims about broader polycyclic drug structures.

↳ Material and methods; Table 6; Statistical analysis of NM-polynomial indices

#positioning

Benchmarking claims exceed displayed evidence

The motivation states that NM-polynomial indices outperformed other descriptor families, but the supplied results do not present the comparative tables or benchmarks needed to substantiate that claim. The work also does not compare against established QSPR toolkits or simpler molecular descriptors.

↳ Motivation; Tables 9–17

Does this match your read? Add your assessment ↓

Withdrawn — shown for reference only

Nabu Commentary

The paper’s strongest element is the transparent graph-theoretic derivation of NM-polynomial descriptors for two exemplar drug graphs, followed by a structured regression workflow for nine physicochemical endpoints. The empirical contribution is limited by the 45-compound dataset and the absence of direct comparisons to standard QSPR descriptors or tools. The most serious issue is internal coherence: Table 8 contains an apparent data-entry error, several confidence intervals are inconsistent with reported R values, and the validation description is not fully stable across sections. Those reliability concerns cap the Methodological Rigour score and make the reported predictive performance provisional rather than practice-ready.

Evaluate another paper

Free · About 5 minutes · Research preview

Author Note / Field Feedback

Nabu’s assessment, alongside the field’s view.

Are you an author of this paper?

Dimension breakdown

Evidence Score

Out of 5
012345
2.5±0.2

Contribution
2.7

The paper contributes an applied NM-polynomial descriptor workflow and worked derivations for Mitoxantrone and Doxorubicin, but the advance is incremental within chemical graph theory and small-sample QSPR. Claims of superior descriptor performance and computational efficiency are not backed by displayed comparative tables or timing benchmarks.

Abstract; Motivation; Theorems 1–4; Tables 7–17

"NM-polynomial indices for Mitoxantrone and Doxorubicin were determined"

Methodological Rigour
2.0

The study uses a recognizable regression workflow with VIF filtering, train/test splitting, within-fold scaling, cross-validation, and external test metrics, which supports a merit-based craft score above the lowest range. The score is capped because the reliability screen identifies data and statistical-reporting concerns in Table 8 and Tables 9–16 that affect confidence in execution.

Material and methods, Model Development and Validation; Table 8; Tables 9–16

"Scaling was fitted on the training subset within each cross-validation fold"

Reporting
3.1

The article’s structure is generally legible, with definitions, theorem-proof derivations, and property-by-property model discussion. Reporting is reduced by undefined corrected R² terminology, repetitive table discussion, and broad claims of robustness despite a clearly failed Index of Refraction endpoint.

Material and methods; Tables 9–17; Discussion of results

"none of the tested regression models achieved satisfactory predictive performance"

Positioning
2.8

The paper situates NM-polynomials within M-polynomial and topological-index literature and explains why neighbourhood-degree descriptors might encode molecular structure. It gives less context for mainstream QSPR validation, simpler descriptor baselines, scaffold-aware testing, or reporting guidelines.

Introduction; Motivation; Applicability domain via Williams plots

"This comparative analysis guided our choice of NM-polynomial indices"

Lower confidence on Methodological Rigour — domain match limited.

Reliability

Concerns4 of 4 checks


Internal CoherenceConcern

Core statistical and data-reporting irregularities affect interpretation of the predictive-performance claims. These include a visibly scrambled row in the physicochemical-property table, unclear validation reporting, confidence-interval anomalies, and duplicated or near-duplicated outcome metrics/text.

Research ConductClean

No research-conduct violation is evident for a computational mathematical/QSPR study. Ethics approval and preregistration are not applicable or not expected for this design, and the uninspected supplementary notebook is an assessability gap rather than a conduct concern.

Flags: 0 declared / 5 total

Reference IntegrityClean

26 references verified via PubMed, Crossref, or OpenAlex.

References not verified automatically are resolved manually within 24 hours.

Post-Publication RecordConcern

This paper was retracted on 2026-05-26. See notice: 10.1038/s41598-026-54594-x.

Last checked: 17 Oct 2026

Sources: Crossref (incl. Retraction Watch) ✓

Impact Profile

Impact areaHealth · Generic health relevance
Type of advanceMethodological
Route to useSuggested
AdvanceExtends
CitationsToo early
Policy · patentNot yet traced
ReachedNot reachedNot assessable or unverified
ReadinessPiloted
Foundational
Controlled
Piloted
Validated
Implementable
FoundationalImplementable

The predictive models were demonstrated on a small dataset of 45 compounds with a 9-compound holdout split by the same authors, which is a proof of concept rather than independent confirmation.

From the paper
“predictive models were proposed, retaining only the most statistically significant ones for estimating uncalculated physicochemical properties”

The paper names a real-world use of its findings, namely estimating physicochemical properties of drug molecules, and mentions the pharmaceutical sector only vaguely without specifying who would act or how.

Pathway

Cheminformatics and pharmaceutical researchers could adopt the Python tool and regression models to estimate physicochemical properties of drug candidates during early drug development.

Barrier

The models rest on only 45 compounds with a 9-compound internal test split, so independent validation on larger and more diverse chemical datasets is needed before routine use.

Evaluated 13 Nov 2026 · GBD 2023 · glm_v1_3_3 · impact_profile_v1_3_3

Report an issue →

AI-generated, human-governed. Something look off? Contact us to request a review.

Keep reading

Related evaluations

Other papers Nabu has evaluated blind, starting with the same field.

  1. An Integrated Data Driven Approach to Drug Repositioning Using Gene-Disease Associations

    Mullen, Cockell, Woollard, et al. · 2016 · Basic medicine

    Evidence Score
    3.2±0.2
    Reliability
    Caveats
  2. AI Agents That Matter

    Kapoor, Stroebl, Siegel, et al. · 2024 · Computer and information sciences

    Primary claim

    “State-of-the-art” agent architectures for HumanEval do not outperform simple baselines.

    Supported
    Evidence Score
    3.9±0.2
    Reliability
    Caveats
  3. Recursa: Toward Recursively Self-Improving Foundation Model Systems

    Zhu, Lin, Shi, et al. · 2026 · Computer and information sciences

    Primary claim

    On Harvey LAB's Firm Knowledge benchmark, Recursa raises the all-pass rate on held-out test tasks from 30.0% to 55.0% with DeepSeek-V4.1-Flash and from 35.0% to 46.0% with Qwen3.8-Flash.

    Tentative
    Evidence Score
    3.0±0.2
    Reliability
    No concerns

How the rubric works →