Machine Learning-Assisted Prediction of Szeged and Mostar Indices for Class of Dendrimers

In the microscopic world of chemistry, molecules are not just random collections of atoms; they are intricate structures with specific shapes, much like tiny architectural marvels. For decades, scientists have sought to understand how the shape of a molecule dictates its behavior—how it dissolves, how it reacts, or how it interacts with living cells. To do this, they often turn to a branch of mathematics called chemical graph theory. In this field, a molecule is treated as a map: the atoms are the points on the map, and the bonds connecting them are the lines. By studying the distances between these points, researchers can assign a single number to a molecule that captures its overall geometry. These numbers, known as topological descriptors, act as a shorthand for a molecule's physical personality, allowing scientists to predict properties like boiling points or how well a drug might travel through the body without having to build and test every single version in a laboratory.

One particularly promising class of molecules for these applications is the dendrimer. Imagine a tree that grows in perfect, symmetrical layers, starting from a central trunk and branching out into a precise, repeating pattern. Dendrimers are synthetic versions of this concept. They are highly branched, tree-like structures that scientists can build layer by layer, with each new layer adding a specific amount of size and complexity. Because their shape is so precise, they are incredibly useful in medicine, particularly for delivering drugs to specific targets in the body or for use in advanced materials. However, as these molecular trees grow larger, predicting their exact behavior becomes a challenge. The question researchers face is whether the mathematical numbers that describe their shape can accurately forecast how they will perform in the real world.

A team of researchers from Pakistan and Saudi Arabia set out to answer this question by focusing on three distinct types of dendrimers: those containing phosphorus, those built around a core called perylenediimide, and those made from aliphatic polyamides. Their goal was to calculate a specific set of mathematical numbers for each of these molecular structures. These numbers, which the researchers refer to as Szeged and Mostar indices, measure different aspects of the molecule's internal distances and symmetry. The Szeged indices look at how the atoms are distributed relative to the bonds, while the Mostar indices measure how uneven or symmetrical the molecule is. By calculating these values for the first three generations of growth for each dendrimer type, the team created a detailed mathematical profile for each structure.

The researchers then took these calculated numbers and compared them against real-world data that had been gathered in previous experiments. They looked at three key physical properties: the molecular weight (how heavy the molecule is), the way the molecule absorbs infrared light (which reveals details about its chemical bonds), and its glass transition temperature (the point at which the material shifts from being hard and brittle to soft and rubbery). By plotting the mathematical numbers against these physical measurements, they used statistical tools to see if a pattern emerged. The results showed a clear and strong connection. For the phosphorus-containing dendrimers, the mathematical number that best predicted the physical properties was the edge Mostar index. For the perylenediimide-cored dendrimers, the total Mostar index provided the strongest link. And for the aliphatic polyamide dendrimers, a combination of the edge and total Mostar indices offered the best predictions.

The study found that as these molecular trees grew larger, their mathematical indices grew in predictable ways. The total Szeged index, which measures the overall complexity of the structure, showed a rapid, exponential increase as the generations advanced, while the Mostar indices grew more slowly. This allowed the researchers to build simple linear equations that could estimate a molecule's physical traits based solely on its mathematical description. For instance, they determined that for the phosphorus-based trees, the edge Mostar index was the most reliable predictor of molecular weight and glass transition temperature. Similarly, for the perylenediimide structures, the total Mostar index was the key to understanding how the molecule would behave as it grew.

These findings suggest that scientists can rely on these mathematical descriptors to forecast the behavior of new dendrimers before they are even synthesized. The phosphorus-containing dendrimers, which showed the highest values for these indices, appear to have a highly branched structure capable of holding a large number of drug molecules, though their asymmetry might make them less soluble. In contrast, the aliphatic polyamide dendrimers had the simplest structure and the lowest indices, suggesting they are more symmetrical and likely more biocompatible, making them ideal for predictable drug delivery. The perylenediimide-cored dendrimers fell in the middle, offering a balanced structure that could efficiently load drugs while maintaining stability.

By proving that these topological numbers correlate so strongly with physical reality, the study offers a powerful tool for drug design and material science. Instead of spending time and resources building and testing countless variations of a molecule in a lab, researchers can now use these mathematical models to screen candidates and predict their properties with high accuracy. This approach does not replace the need for experimental validation, but it significantly narrows the field, allowing scientists to focus their efforts on the most promising molecular designs. The work demonstrates that the language of mathematics can effectively describe the physical world of molecules, turning complex structural problems into solvable equations and paving the way for more efficient development of life-saving drugs and advanced materials.