Presentation Information
[4DSP-05]Machine Learning-Driven Prediction of Liposome Particle Size by In-line Process Analytical Technology
○Junghu Lee1, Nozomi Watanabe Morishita1, Noriko Yoshimoto2, Seonghyun Eom3, Moonkyu Kwak4, Hosup Jung3,5, Hiroshi Umakoshi1 (1. The University of Osaka (Japan), 2. Yamaguchi University (Japan), 3. Nbiocell Inc. (Korea), 4. Kyungpook National University (Korea), 5. Seoul National University (Korea))
Keywords:
Liposome,Particle size,Machine learning,Process analytical technology,Quality control
[Purpose]
In Liposomal drug delivery systems, particle size is a critical quality attribute. Microfluidic systems have emerged as yielding liposomes with a narrow size distribution compared to conventional methods. Meanwhile, the final particle size is the result of a complex interplay between process parameters. Our previous studies underscore that membrane physicochemical characteristics must be balanced with synthesis conditions to achieve precise size control.This study introduces an approach towards membrane characteristic based machine learning (ML) prediction in microfluidic liposome synthesis. We propose a framework integrating in-line process analytical technology (PAT) within a closed-loop ML prediction.
[Method]
The process analytical technology (PAT) system was constructed to directly couple a microfluidic device with an optical detection module capable of acquiring continuous fluorescence spectra during liposome preparation. Four different ratios of 1,2-dioleoyl-sn-glycero-3-phosphocholine (DOPC), cholesterol and egg sphingomyelin were chosen.
[Results]
In the development of ML to predict particle size, physicochemical parameters were employed as input features with five different algorithms.Particle size prediction performance were evaluated using actual versus predicted value plots, demonstrating generalization capability in ML prediction; it achieved the highest predictive accuracy (RMSE = 7.18 nm, 7.6% relative to the mean particle size). To evaluate the predictive capability beyond the training dataset, an external validation was conducted using eight independent experimental conditions that were completely excluded from the model training phase. in-line PAT-ML model successfully predicted the particle sizes of external validation, achieving an RMSE of 7.53 nm.
[Consideration]
This achievement represents a significant reduction in the data, time, and resources required for high predictive accuracy ML model development.
[Conclusion]
This study developed a framework integrating in-line PAT with ML for the prediction and control of liposome particle size. The model’s RMSE was 7.18 nm, external validation on unseen experimental conditions—including novel process parameters and lipid composition—confirmed robust generalization (RMSE 7.53 nm), proving the model capability.
In Liposomal drug delivery systems, particle size is a critical quality attribute. Microfluidic systems have emerged as yielding liposomes with a narrow size distribution compared to conventional methods. Meanwhile, the final particle size is the result of a complex interplay between process parameters. Our previous studies underscore that membrane physicochemical characteristics must be balanced with synthesis conditions to achieve precise size control.This study introduces an approach towards membrane characteristic based machine learning (ML) prediction in microfluidic liposome synthesis. We propose a framework integrating in-line process analytical technology (PAT) within a closed-loop ML prediction.
[Method]
The process analytical technology (PAT) system was constructed to directly couple a microfluidic device with an optical detection module capable of acquiring continuous fluorescence spectra during liposome preparation. Four different ratios of 1,2-dioleoyl-sn-glycero-3-phosphocholine (DOPC), cholesterol and egg sphingomyelin were chosen.
[Results]
In the development of ML to predict particle size, physicochemical parameters were employed as input features with five different algorithms.Particle size prediction performance were evaluated using actual versus predicted value plots, demonstrating generalization capability in ML prediction; it achieved the highest predictive accuracy (RMSE = 7.18 nm, 7.6% relative to the mean particle size). To evaluate the predictive capability beyond the training dataset, an external validation was conducted using eight independent experimental conditions that were completely excluded from the model training phase. in-line PAT-ML model successfully predicted the particle sizes of external validation, achieving an RMSE of 7.53 nm.
[Consideration]
This achievement represents a significant reduction in the data, time, and resources required for high predictive accuracy ML model development.
[Conclusion]
This study developed a framework integrating in-line PAT with ML for the prediction and control of liposome particle size. The model’s RMSE was 7.18 nm, external validation on unseen experimental conditions—including novel process parameters and lipid composition—confirmed robust generalization (RMSE 7.53 nm), proving the model capability.
Comment
To browse or post comments, you must log in.Log in