Comprehensive modeling of cell culture profile using Raman spectroscopy and machine learning.

Chinese hamster ovary (CHO) cells are widely utilized in the production of antibody drugs. To ensure the production of large quantities of antibodies that meet the required specifications, it is crucial to monitor and control the levels of metabolites comprehensively during CHO cell culture. In recent years, continuous analysis methods employing on-line/in-line techniques using Raman spectroscopy have attracted attention. While these analytical methods can nondestructively monitor culture data, constructing a highly accurate measurement model for numerous components is time-consuming, making it challenging to implement in the rapid research and development of pharmaceutical manufacturing processes. In this study, we developed a comprehensive, simple, and automated method for constructing a Raman model of various components measured by LC-MS and other techniques using machine learning with Python. Preprocessing and spectral-range optimization of data for model construction (partial least square (PLS) regression) were automated and accelerated using Bayes optimization. Subsequently, models were constructed for each component using various model construction techniques, including linear regression, ridge regression, XGBoost, and neural network. This enabled the model accuracy to be improved compared with PLS regression. This automated approach allows continuous monitoring of various parameters for over 100 components, facilitating process optimization and process monitoring of CHO cells.

Development of a retention prediction model in ion-pair reversed-phase HPLC for nucleoside triphosphates used as mRNA vaccine raw materials.

The International Conference on Harmonization guidelines for quality on pharmaceutical development recommends a systematic development approach including robustness studies which assure performance of manufacturing and analytical method development of drug product. It was demonstrated that the retention prediction model for nucleoside triphosphates (NTPs) on ion-pair reversed-phase HPLC was developed by a highly accurate Kawabe's model which supports the development of robust HPLC methods. As NTPs and its derivatives are typically used for Messenger ribonucleic acid (mRNA) vaccine production, adenosine-5'-triphosphate (ATP), guanosine-5'-triphosphate (GTP), cytidine-5'-triphosphate (CTP), 5-methylcytidine-5'-triphosphate (m5-CTP), uridine-5'-triphosphate (UTP), 5-methyluridine-5'-triphosphate (m5-UTP), pseudouridine-5'-triphosphate (Ψ-UTP) and N1-methylpseudouridine-5'-triphosphate (m1Ψ-UTP) were applied for prediction model development. By a comparison of the predicted retention factor in eight studied samples with the retention factor measured under six isocratic conditions, the absolute prediction error was 0.075 and also the prediction error (%) was 2.70%. In practical examples, analytical method for residual ATP, GTP, CTP, and m1Ψ-UTP in the commercial mRNA-based drugs and purity method for UTP derivatives were optimized by QbD approach. The design space for the minimum resolution between adjacent peaks was simulated with the models developed to evaluate the robustness of peak separation, and the optimal mobile phase condition was also simulated. As a conclusion, the desired peak was successfully separated under the optimized condition, and we thought that these retention models could optimize the mobile phase condition of the NTP analysis method for applying to various quality tests, such as quantity, purity and identity test for NTPs and its derivates in the mRNA-based drugs.

The development of retention time prediction model using multilinear gradient profiles of seven pharmaceuticals.

The ICH guidance on pharmaceutical development recommends a systematic development approach including robustness studies which assure performance of manufacturing and analytical method development of drug product. The retention model by T. Kawabe et al have an excellent correlation between observed and predicted retention time in various kinds of pharmaceutical compounds during isocratic elution by the multiple regression modeling of solvent strength parameters. However, it cannot be successfully applied to the predictability of the retention time during multilinear gradient elution and also it does not consider the instrument dependent parameters such as dwell volume. The current study demonstrated that the solution of the fundamental gradient elution equation was applied to T. Kawabe's retention time prediction model to predict the retention time using a multilinear gradient profile with taking the delay volume of HPLC system into account. Seven pharmaceutical compounds were used for evaluation of prediction models for retention time. The predicted retention time was compared with the measured retention time obtained by several multilinear gradient using two HPLC systems with different dwell volume. The evaluated prediction error (%) was 1.10 % and 1.54 % with H-Class and Nexera XR HPLC systems, respectively. In order to evaluate the robustness of the analytical method and to set the system suitability test (SST) for proper method performance, the design space for the ACN/MeOH mixture ratio in the total organic solvent and the full width at half maximum (FWHM) relationship to the minimum resolution was simulated by the developed retention time prediction. The optimized condition of the ACN/MeOH mixture ratio, the acceptance criterion of the SST for achieving the robust separation was estimated based on the simulated design space. As a conclusion, the developed retention time prediction will be useful during analytical method transfer among different manufacturing/analytical sites of the pharmaceuticals with different HPLC systems.

Analysis of available surface area can predict the long-term dissolution profile of tablets using short-term stability studies.

Generally, since at least 6 months are usually needed for accelerated testing of tablet at 40 °C/75% relative humidity (RH), it would be crucial important to predict the dissolution profiles during long-term storage period by using samples stored with shorter periods such as 3 months. In this study, we developed a new method for predicting changes in dissolution from tablets during long-term storage-based changes in the available surface area [S (t)]. In addition, we discussed the dissolution behavior and mechanisms using S (t). The results revealed drastic delays in dissolution in samples stored at 40 °C/75% RH for 7 weeks. Considering changes of S (t) patterns, this delay was derived from changes of the tablet surface. New parameters, namely T22.1 and T63.2, calculated from the S (t) profile tended to increase with an increased duration of testing. Concerning the long-term prediction model using short-term data, a nonlinear model was deemed appropriate because good agreement was observed between the value predicted using the model and the measured value for samples stored at 40 °C/75% RH for 6 months. Therefore, using the new evaluation method based on S (t), we can predict changes in dissolution during long-term storage using short-term methods.

Optical properties of drying wood studied by time-resolved near-infrared spectroscopy.

We measured the optical properties of drying wood with the moisture contents ranging from 10% to 200%. By using time-resolved near-infrared spectroscopy, the reduced scattering coefficient µs' and absorption coefficient µa were determined independent of each other, providing information on the chemical and structural changes, respectively, of wood on the nanometer scale. Scattering from dry pores dominated, which allowed us to determine the drying process of large pores during the period of constant drying rate, and the drying process of smaller pores during the period of decreasing drying rate. The surface layer and interior of the wood exhibit different moisture states, which affect the scattering properties of the wood.

Determination of true optical absorption and scattering coefficient of wooden cell wall substance by time-of-flight near infrared spectroscopy.

The true absorption coefficient (µa) and reduced scattering coefficient (µ´s) of the cell wall substance in Douglas fir were determined using time-of-flight near infrared spectroscopy. Samples were saturated with hexane, toluene or quinolone to minimize the multiple reflections of light on the boundary between pore-cell wall substance in wood. µ´s exhibited its minimum value when the wood was saturated with toluene because the refractive index of toluene is close to that of the wood cell wall substance. The optical parameters of the wood cell wall substance calculated were µa = 0.030 mm(-1) and µ´s= 18.4 mm(-1). Monte Carlo simulations using these values were in good agreement with the measured time-resolved transmittance profiles.