Machine learning tools to enable predictive approach in PEG-based ointments with textural & rheological outputs

Abstract

Polyethylene glycol (PEG) excipients, commonly used in ointment bases, significantly impact formulation properties based on their molecular weights and ratios. Traditional optimization methods are time-consuming and costly. This study employs machine learning to predict physicochemical properties of PEG-based ointments, enabling in-silico optimization. Predictive results align with experimental data, demonstrating the approach’s effectiveness in simulating formulation behavior. This method reduces development time and costs while improving accuracy for pharmaceutical formulation development.

Introduction

Polyethylene glycol (PEG) excipients are commonly used in pharmaceutical formulations, particularly in ointment bases, where their molecular weights and ratios significantly influence rheological and textural properties, as observed in a previous study conducted by this team (“Impact of Polyethylene Glycol molecular weights and ratios on Rheology and Textural Properties of Ointment Bases”). Traditional optimization of these formulations relies on time-consuming and costly trial-and-error laboratory experiments. In this scenario, a machine learning-based approach to analyze physicochemical data and predict the impact of PEG (molecular weights and ratios) on ointment characteristics could serve as an enabling tool to speed up pharmaceutical development, in a cost-effective way.

Materials and Methods

Results

Based on the model’s coefficients, it is noted that the ratio (%) between low and high molecular weight PEGs has a predominantly larger impact than the molecular weight (MW) of each of them.

Figure 2: Comparative chart regarding the difference in impact caused by the change in the % PEG and PEG MW on the model's firmness coefficient. — Figure 2: Comparative chart regarding the difference in impact caused by the change in the % PEG and PEG MW on the model’s firmness coefficient.

*Note: The Machine Learning models were developed based on the experimental data obtained from Study “Impact of Polyethylene Glycol molecular weights and ratios on Rheology and Textural Properties of Ointment Bases”.

To demonstrate this difference in impact between %PEG and PEG MW and verify model fit, the real vs. predicted firmness data are presented below.

Figure 3: Comparative charts of real values vs. predicted values demonstrating that the model coefficients are representative of reality.

Once predictive models follow the same trends and accurately represent real-world data, it is possible simulate, for example, hypothetical PEG MW behavior in a target formulation to evaluate its performance and commercial appeal—all before committing to experimental concept proofing. Below, the demonstration of this approach with the predicted parameter values for this case.

Figure 4: Qualitative chart with normalized data for evaluation of characteristics, considering the model prediction of PEG 5000 data compared to the real data of the other analyzed PEGs.

With the predicted models, it is also possible optimize formulations according to the desired delivery profile, minimize bench testing, and consequently save time and reduce costs.

Conclusion

Machine learning-based data prediction proved effective when comparing analyzed variables with bench test results. The predicted values closely matched real data trends and approximate values, reflecting the ointment formulation’s potential behavior.

See the full poster on Machine learning tools to enable predictive approach in PEG-based ointments with textural & rheological outputs here

(click the picture to download the poster)

Source: Indovinya, Rafael Caetano Jardim Pinto da Silva Salvato; Hemyle Rangel Rocha Chaves; Maurício Soares Júnior; Beatriz Rodrigues Pinto; Jordana Sakis Sonza (All for the same affiliation: Indorama Ventures – Indovinya), IPEC Americas Excipients World 2025

Interested in more information about Indovinya excipients here?

Name *

First

Last

Email *

Company *

Job Title

Address *

Address Line 1

Address Line 2

City

State / Province / Region

Postal Code

Country

I would like to get a sample of *

OXIPURITY® (please specify quantity in the comments)
EMCAPlus® (please specify quantity in the comments)
Information request (please specify in the comments)

Comment or Message *

I agree with the privacy policy as per https://www.pharmaexcipients.com/privacy/ . The contact details will be shared with the supplier who holds the request above. The supplier can contact you directly.

Comment