🔬 materials science

QSPR Analysis with Curvilinear Regression Modeling and Temperature-based Topological Indices

This study develops and compares linear, quadratic, and cubic curvilinear regression models using temperature-based topological indices to predict the thermodynamic properties of monocarboxylic acids within the QSPR framework.

Original authors: H. M. Nagesh

Published 2026-07-02

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: H. M. Nagesh

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you have a giant box of LEGO sets. Some are small, some are huge, and they all have different shapes. If you wanted to guess how much energy it would take to melt a specific LEGO set or how much heat it would release if you burned it, you wouldn't need to actually melt or burn every single one. You just need to look at the shape and the number of connections between the bricks.

This paper is essentially a recipe for doing exactly that, but with molecules instead of LEGOs.

Here is the breakdown of what the author, H. M. Nagesh, did, explained in plain English:

1. The Big Idea: Shape Determines Behavior

The paper starts with a simple rule: The way a molecule is built (its structure) dictates how it behaves (its properties).

In the world of chemistry, scientists use a tool called QSPR (Quantitative Structure-Property Relationship). Think of this as a translator. It translates the "language" of a molecule's shape into the "language" of its physical behavior, like how much heat it holds or how easily it turns into gas.

2. The Tools: "Temperature" Maps

To translate the shape, the author used a special type of map called a Topological Index. Usually, these maps just count connections. But this paper uses a newer, fancier kind of map called Temperature-based Topological Indices.

Imagine every atom in a molecule is a person at a party.

The "Temperature" of a person: This isn't about hot or cold. It's a mathematical score based on how many friends (bonds) they have compared to the total number of people at the party. If you have many friends, your "temperature" score is high.
The Four Maps: The author created four different ways to calculate the "temperature" of the whole party (the molecule):
1. Sum Connectivity: Adding up the temperatures of connected pairs.
2. Product Connectivity: Multiplying the temperatures of connected pairs.
3. F-Temperature: Squaring the temperatures and adding them up.
4. Symmetric Division: Comparing the temperatures of neighbors to see how balanced they are.

3. The Experiment: Testing 19 Molecules

The author took 19 different carboxylic acids (a family of chemicals that includes things like vinegar and fatty acids found in food). These molecules range from very small (2 carbons) to quite large (20 carbons).

For each molecule, the author:

Calculated the four "Temperature Maps" described above.
Looked up three real-world facts about them:
- Formation Enthalpy: How much energy was needed to build the molecule.
- Combustion Enthalpy: How much heat it releases when burned.
- Vaporization Enthalpy: How much heat is needed to turn it from liquid to gas.

4. The Math: Finding the Perfect Curve

Now, the author tried to draw a line connecting the "Temperature Maps" to the real-world heat facts. They tried three types of lines:

Straight Line (Linear): A simple, direct path.
Curved Line (Quadratic): A gentle hill or valley.
Wiggly Line (Cubic): A line that curves up, then down, then up again.

The Result:
The author found that the Wiggly Line (Cubic) was the best fit. It was the most accurate way to predict the heat properties just by looking at the shape.

5. The Winners and Losers

Not all the "Temperature Maps" were equally good at predicting the future:

The Champions: The Sum Connectivity and Product Connectivity indices were the stars of the show. They predicted the heat properties with incredible accuracy (over 99% correlation). If you used these two maps, you could guess the heat behavior of a new molecule almost perfectly.
The Runner-Up: The Symmetric Division index was okay, but not as good as the champions.
The Loser: The F-Temperature index was a poor predictor. It couldn't reliably guess the heat properties, no matter how complex the math got.

Summary

In short, this paper proves that if you want to know how much heat a specific type of acid will hold or release, you don't need to run a lab experiment. You just need to look at its molecular shape, calculate two specific "temperature" scores (Sum and Product connectivity), and plug them into a specific 3-step curve equation.

The paper concludes that this mathematical shortcut works very well for this family of 19 acids, offering a fast and accurate way to predict their thermodynamic behavior based purely on their structure.

Technical Summary: QSPR Analysis with Curvilinear Regression Modeling and Temperature-based Topological Indices

Problem Statement
The paper addresses the challenge of establishing quantitative correlations between the chemical structures of monocarboxylic acids and their thermodynamic properties. Specifically, the study focuses on predicting the enthalpies of formation ( $\Delta H^0_f$ liquid), enthalpies of combustion ( $\Delta H^0_C$ liquid), and enthalpies of vaporization ( $\Delta H^0_{vap}$ gas) for a series of 19 monocarboxylic acids ranging from $C_2H_4O_2$ to $C_{20}H_{40}O_2$ . The goal is to determine if temperature-based topological indices can serve as effective molecular descriptors within Quantitative Structure-Property Relationship (QSPR) models to accurately forecast these properties.

Methodology
The study employs chemical graph theory, representing molecules as graphs where vertices are atoms and edges are chemical bonds. The methodology proceeds through the following steps:

Definition of Indices: The paper utilizes four specific temperature-based topological indices derived from the "temperature" of a vertex, defined as $T(v) = \frac{deg(v)}{n - deg(v)}$ , where $deg(v)$ is the degree of vertex $v$ and $n$ is the total number of vertices. The indices calculated are:
- Sum Connectivity Temperature Index ($ST(G)$)
- Product Connectivity Temperature Index ($PT(G)$)
- F-Temperature Index ($FT(G)$)
- Symmetric Division Temperature Index ($SDT(G)$)
Data Collection: Thermodynamic property values for the 19 acids were sourced from existing literature (references [5, 14]).
Regression Modeling: The relationship between the calculated topological indices and the thermodynamic properties was analyzed using curvilinear regression models. Three types of equations were tested:
- Linear: $Y = a + b_1X_1$
- Quadratic: $Y = a + b_1X_1 + b_2X_1^2$
- Cubic: $Y = a + b_1X_1 + b_2X_1^2 + b_3X_1^3$
Statistical Evaluation: Model performance was assessed using the correlation coefficient ( $R^2$ ), the F-ratio test ( $F$ ), and the Root Mean Square Error (RMSE). The study posits that a good QSPR model requires $R^2 > 0.99$ , with the best predictive model identified by the minimum RMSE and maximum $R^2$ or $F$ values.

Key Contributions and Results
The study computed the four temperature-based indices for the 19 monocarboxylic acids and generated regression models for each thermodynamic property. The results indicate distinct variations in predictive capability among the indices:

Superior Indices: The Sum Connectivity Temperature Index ($ST(G)$) and Product Connectivity Temperature Index ($PT(G)$) demonstrated the highest efficacy. For all three thermodynamic properties ( $\Delta H^0_f$ , $\Delta H^0_C$ , and $\Delta H^0_{vap}$ ), models using these indices achieved $R^2$ values greater than 0.99.
- For $\Delta H^0_f$ , the best models were cubic (for $ST(G)$) and cubic (for $PT(G)$), with $R^2$ values of 0.9988 and 0.9986, respectively.
- For $\Delta H^0_C$ , the cubic model for $ST(G)$ yielded an $R^2$ of 0.9998, while the quadratic model for $PT(G)$ yielded 0.9941.
- For $\Delta H^0_{vap}$ , the cubic model for $ST(G)$ achieved an $R^2$ of 0.9978.
Inferior Indices: The F-Temperature Index ($FT(G)$) and Symmetric Division Temperature Index ($SDT(G)$) showed significantly lower predictive power.
- $FT(G)$ models resulted in low correlation coefficients (e.g., $R^2 = 0.2913$ for linear $\Delta H^0_f$ ), failing to meet the $R^2 > 0.99$ threshold even in cubic forms.
- $SDT(G)$ models showed moderate correlation (around $R^2 \approx 0.86$ ) but were deemed insufficient for high-accuracy prediction compared to $ST(G)$ and $PT(G)$.
Model Form: Across the dataset, the cubic equation consistently provided the best goodness of fit among the linear, quadratic, and curvilinear options, particularly for the $ST(G)$ and $PT(G)$ indices.

Significance and Claims
The paper claims that temperature-based topological indices, specifically the Sum Connectivity and Product Connectivity variants, are highly suitable descriptors for predicting the thermodynamic properties of monocarboxylic acids. The study concludes that these indices, when applied within cubic curvilinear regression models, offer a robust mathematical framework for QSPR analysis in this chemical series.

The authors modestly note that while the symmetric division temperature index shows some correlation, the F-temperature index is inadequate for making good predictions for these specific compounds. The work reinforces the utility of graph-theoretical approaches in linking molecular structure to thermodynamic behavior, providing specific regression equations that can be used to estimate enthalpies of formation, combustion, and vaporization for the studied range of acids without requiring experimental measurement.