Computational ranking puts Peyronie’s disease drugs through a physicochemical filter
A new computational analysis reframes how drug candidates for Peyronie’s disease can be compared. Instead of asking which agent works best at the bedside, the study asks which molecules present the most favorable structural and physicochemical profile when judged through graph theory, quantitative structure–property relationship modeling, and multi-criteria decision-making methods. The distinction matters because Peyronie’s disease is a fibrotic disorder of the penile tunica albuginea, and the pharmacology of any proposed treatment ultimately depends on how the compound behaves as a molecule before it ever reaches a patient.
The investigators examined twenty drugs used in the treatment of Peyronie’s disease. Their framework combined chemical graph theory with QSPR and MCDM techniques, a layered approach designed to move from atomic connectivity to property prediction and then to ranked comparison. In practical terms, the workflow treated each drug as a molecular structure first, then as a bundle of measurable properties, and finally as a decision object that could be ordered against its peers.
One sentence captures the scope of the work plainly: the output is a ranking of molecules, not a clinical hierarchy of therapies.
Graph indices linked structure to drug-like properties
Five degree-based topological descriptors formed the structural core of the analysis: the second Zagreb index, Atom Bond Connectivity, Geometric Arithmetic, Sombor, and Nirmala indices. These descriptors arise from molecular graph representations, where atoms become vertices and chemical bonds become edges. In a disease area such as Peyronie’s, where multiple therapies compete without a single universally accepted comparative standard, this sort of mathematical profiling can help separate compounds with similar indications but different structural behavior.
The researchers then correlated those descriptors with nine physicochemical properties. Among the tested traits, molar refractivity, polarizability, and molecular weight gave the most consistent QSPR performance. Those properties are not abstract numbers. They reflect how a molecule occupies space, distorts under an electric field, and navigates the balance between bulk and reactivity. In drug development, such attributes often influence permeability, binding, and how readily a compound can be formulated or delivered.
The models were not left to stand on a single fit statistic. Multiple linear regression was developed and validated with Leave-One-Out Cross Validation, ridge regression with nested LOOCV, and Y-randomization testing. The reported model performance was strong across the three leading properties: R² values reached 0.9733 for molar refractivity, 0.9829 for polarizability, and 0.9768 for molecular weight, while the corresponding LOOCV Q² values were 0.9435, 0.9547, and 0.9500. Those figures suggest the descriptors carried stable predictive signal rather than a fragile fit that collapses under resampling.
Decision weights were built from entropy and intercriteria structure
After the property models were built, the analysis moved into ranking. Entropy and CRITIC methods were used to assign criterion weights, and the two were then merged through a hybrid weighting approach. This step matters because a ranking system can only be as credible as the importance assigned to each criterion. A compound that looks attractive under one property may lose ground once other dimensions are introduced, especially when the method is built to balance competing advantages rather than reward a single favorable metric.
Five MCDM approaches were then applied: MARCOS, VIKOR, WASPAS, TOPSIS, and EDAS. The use of multiple ranking systems gave the authors a way to test whether the order of the twenty drugs held up across different decision rules. Spearman rank correlation and consensus aggregation through average ranking and Borda count then served as the final consistency checks. The result was a repeatable comparative framework rather than a one-off ordering generated by a single algorithm.
One compact conclusion emerges from the data: the rank list is computationally stable, but it remains a physicochemical ordering, not a claim of therapeutic superiority.
Why the results matter for fibrotic drug comparison
Peyronie’s disease is driven by fibrotic plaque formation in the penile tunica albuginea, and that biology makes any treatment discussion especially sensitive to drug behavior. Molecules with more favorable refractivity, polarizability, and mass-related properties may interact differently with biological systems, but the current analysis stops short of proving clinical benefit. Instead, it offers a structured way to prioritize candidate drugs before the more expensive and clinically demanding stages of comparative evaluation.
This is where the study fits into translational work. QSPR can identify relationships between structure and likely property performance, while MCDM can compress several property dimensions into a decision framework. Used together, the two methods create a screening tool for comparing compounds that already have a role in treatment discussions. For a connective tissue disorder with persistent deformity, pain, and erectile dysfunction, such screening may help researchers focus subsequent experimental work on compounds that appear most balanced on paper.
The findings also make clear what the approach cannot do. It does not measure symptom relief, plaque regression, curvature improvement, pain reduction, or erectile function. It does not compare patient outcomes. It does not substitute for pharmacodynamic or clinical trial data. What it does provide is a mathematically explicit path from molecular structure to ranked comparison, with validation steps that reduce the chance that the pattern is driven by chance alone.
Validation strengthened the model, but the clinical question remains open
The authors also applied Y-randomization testing, a safeguard against spurious descriptor-property relationships. In a computational study of this type, that step is important because a model can appear persuasive simply by exploiting noise in the data. The combination of LOOCV, ridge regression, and randomization testing makes the framework more robust than a simple correlation screen.
Even so, the central limitation is unavoidable: physicochemical prioritization is not the same as clinical efficacy. A compound can look favorable in QSPR and MCDM terms while still failing in real-world disease modification. That gap is particularly important in Peyronie’s disease, where treatment decisions depend on symptom stage, plaque characteristics, functional impairment, and patient goals, none of which are captured by a molecular ranking exercise.
The study therefore lands in a narrow but useful place. It offers a reproducible computational method for comparing twenty drugs used in Peyronie’s disease, identifies molar refractivity, polarizability, and molecular weight as the most reliable modeled properties, and shows that several ranking systems can converge on a stable order. The unanswered question is not whether the mathematics works, but whether this kind of ranking can be connected to future laboratory and clinical testing in a way that changes how therapies are selected at the bedside.