This article explores the use of chromatographic retention measurements in both LC and SFC to develop quantitative retention activity relationship (QRAR) models, offering a predictive tool for skin permeability without animal testing.
Oral administration of drugs may present particular disadvantages, such as a pronounced first-pass effect or adverse side effects. To avoid these problems, transdermal administration may be a good alternative. Several chromatographic systems were investigated to estimate the skin permeability of various pharmaceutical and cosmetic compounds, including reversed-phase liquid chromatography (RPLC), micellar liquid chromatography (MLC), ultrahigh-pressure liquid chromatography (UHPLC), and supercritical fluid chromatography (SFC). Fast methods were searched as an alternative for the current methods to predict the skin permeability.
Texture of human skin © PixieMe - stock.adobe.com
Oral drug administration can present drawbacks, such as a pronounced first-pass effect. Transdermal application of active pharmaceutical ingredients (APIs) may provide a solution. Different formulations, for example, transdermal drug delivery systems, creams, ointments, gels, and lotions, can target the skin or use it to deliver compounds into the body. In the latter case, resorption through the skin is desired. On the other hand, topically applied products are intended to remain components in the superficial layers of the skin (for example, cosmetic products or sunscreens). Besides the dermal application of medical and cosmetic products, exposure to chemical compounds can also occur in an occupational, environmental, or consumer setting. Assessing the skin permeability of compounds is crucial not only in drug discovery and development but also in dermal exposure testing and risk assessment (1).
The permeation of substances after dermal application can be determined through in vivo methods on humans or laboratory animals. A clear advantage of these methods is the use of an intact system, both in terms of physiology and metabolism. However, the ethical issues of using animals is a drawback, as well as the limited throughput. Furthermore, extrapolation from animals to human skin is needed. As animal skin is often more permeable, human permeability is overestimated (2).
As an alternative for the in vivo methods, in vitro approaches were developed. The most frequently applied method to measure skin permeability uses a diffusion cell, for example, the Franz diffusion cell (3). This cell consists of a donor compartment, in which the tested compound is applied to a membrane (human or animal skin, artificial membranes or cultured skin). On the other side of the membrane, a receiver chamber is located and filled with receptor medium (4). However, when animal skin is used, the extrapolation to human permeation is not always straightforward (5). Furthermore, inter-laboratory variations and practical experience required to operate such diffusion cells are disadvantages (6).
Skin permeability measurements on humans or with animals are time-consuming, show high variability, and encounter ethical problems. Furthermore, cosmetic ingredients can no longer be tested on animals according to the European Cosmetics Regulation (1223/2009) (7), further necessitating alternative in vitro or in silico methods. Quantitative retention activity relationship (QRAR) models that employ chromatographic retention as a molecular descriptor could provide a solution, which could also be applicable for high-throughput screenings. Therefore, the general aim was to search for a fast and predictive chromatographic method, either using liquid chromatography (LC) or supercritical fluid chromatography (SFC), as an alternative for the current methods used to predict skin permeability of pharmaceutical or cosmetic compounds.
A test set of 58 pharmaceutical and cosmetic compounds was analyzed on all chromatographic systems. Components from different (pharmacological) classes, such as corticosteroids, anti-inflammatory drugs, hormones, preservatives, hair dye agents, and antiseptics, were selected. Their log P range varied between -1.13 and 4.45, and the logarithm of the skin permeability coefficient (log Kp) ranged from -5.52 to -0.24. Samples were prepared at a concentration of 0.1 mg/mL in methanol (MeOH). Retention times (tR) of the test set compounds were recorded on different chromatographic systems and the retention factors k were calculated.
Assessment of the correlation between the chromatographic descriptor and the log Kp was done using the correlation coefficient (r), which should approach 1 if there is a good correlation between both parameters. Next, quantitative structure activity relationship (QSAR) models were built that related log Kp to molecular descriptors. Models were built with only theoretical descriptors, which were either (i) a set of 22 descriptors that included the melting point and 21 calculated with the Vega ZZ software, or (ii) 408 E-Dragon-software descriptors.Models with theoretical descriptors were considered as references, as they require no experimental descriptors to be built. Subsequently, models that contained both theoretical and a chromatographic descriptor were built. Multiple linear regression (MLR) and partial least squares (PLS) models were constructed. However, since MLR provided the best results, only those models will be discussed further. Models were compared by means of their determination coefficient (r²), root mean squared error of cross validation (RMSECV) from a leave-one-out cross-validation, and the root mean squared error of calibration (RMSEC). r²-values should be close to 1, and RMSECV and RMSEC as low as possible. When the models with a chromatographic descriptor had an improved prediction ability compared to those containing only theoretical descriptors, the added value of the chromatographic descriptor was demonstrated.
An octadecyl silica (C18) column was used to establish the chromatographic descriptor(8). The mobile phase consisted of buffer–acetonitrile mixtures at two different pH levels. A phosphate buffer pH 7 was combined with 25 to 40% v/v acetonitrile (in steps of 5% v/v), and an acetate buffer at pH 5.5 (pH of the skin) was mixed with fractions between 25 and 45% v/v acetonitrile.
From the retention factors measured with mobile phases containing different fractions of modifier, kw, the retention factor in a purely aqueous mobile phase was estimated using the following equation (9):
log k = log kw − s φ
in which k and kw are the retention factors in regular mobile phases (containing organic modifier) and pure aqueous mobile phase, respectively, while φ represents the fraction organic modifier in the mobile phase.
First, the correlation between log k or log kw and the skin permeability coefficients, log Kp, was assessed for each mobile phase composition. Low correlations were observed, displaying a maximal value (r = 0.317) with 45% v/v acetonitrile at pH 5.5. The correlation between log Kp and the chromatographic data at pH 5.5 seemed slightly better than at pH 7, which is because pH 5.5 reflects the skin conditions better. Log k values from the mobile phases with a higher percentage of acetonitrile generally correlated better with the skin permeability. Conversely, the extrapolated log kw at pH 5.5 showed the lowest correlation with the skin permeability, while that at pH 7 was one of the highest. In summary, the retention factors demonstrated a low correlation with the log Kp values, which means that alone they were inadequate to predict skin permeability. Retention on the C18 column thus only represents part of the skin permeability process, and adding other molecular descriptors seems necessary to improve the modeling.
Figure 1: Predicted log Kp as a function of the experimental one, together with the regression line (solid line) and the bisector (dashed line) for the equation 2 stepwise MLR model. Adapted and reproduced with permission from reference 8.
Table I
Second, models were built using only theoretical descriptors; Table I shows the best models obtained. Equation 1 contains four Vega ZZ descriptors, while in equation 2, 10 E-Dragon descriptors were used. As seen from the quality parameters, r², RMSEC, and RMSECV, the best performing model was obtained with the E-Dragon descriptors. There is also a good correspondence between experimental and predicted log Kp with the latter model (Figure 1). The complexity of the latter model is large, which may be considered a drawback.
Table II
In the next step, the chromatographic descriptors were added to the models. For the Vega ZZ descriptor set (Table II), the best models were obtained including the extrapolated kw at pH 5.5 (equation 3) and the log k value with the mobile phase at pH 7 with 35% acetonitrile (equation 4).
When comparing these models (equations 3 and 4) with those containing only theoretical Vega ZZ descriptors (equation 1), the three models behaved similar in terms of RMSEC and RMSECV. In conclusion, the chromatographic descriptors on the C18 column provided no improvement to the model with only theoretical descriptors. Similar models with E-Dragon descriptors could not be created with regular stepwise MLR modeling because the retention factors were never selected by the algorithm to be added to the model.
MLC is a type of reversed-phase liquid chromatography (RPLC), where a surfactant is added to the mobile phase in a concentration above the critical micelle concentration. Micelles are formed in the mobile phase, resulting in a pseudo-phase with which compounds may also interact, alongside the stationary and mobile phases. MLC has frequently been applied for the prediction of different biopartitioning parameters, such as skin permeability and intestinal absorption (10,11). Incorporating a surfactant into the mobile phase offers the advantage of reducing—or even eliminating—the need for organic modifiers, rendering this method greener.
Sodium dodecyl sulphate (SDS, anionic) was used as surfactantin a recent study(12). To provide a fair comparison with the method from the RPLC study, an identical column to that used in reference 8 was applied, along with the same test set. From a two-factor central composite design, the optimal amounts of surfactant and organic modifier in the mobile phase were determined as 0.08 M SDS and 7.5% v/v 1-propanol. These were combined with a 0.05 M acetate buffer pH 5.5 in the mobile phase.
To further accelerate the MLC method, the test set was also analyzed on a monolithic column applying the same mobile phase. Monoliths consist of continuous porous material containing both macropores and mesopores, which allow an increasing flow rate without causing a high back pressure.
To investigate whether an increase in flow rate caused a loss of information on the monolithic column, the retention factors obtained at 1 and 8 mL/min were compared for the modeling. The correlations between log k and log Kp on both the particle-based and the monolithic columns were similar. The highest value was observed when using the monolithic column at a flow rate of 8 mL/min to measure log k. When comparing results between the two column types, retention factors were overall higher on the particle-based column. The correlation between log k on both columns was 0.789, indicating that for some compounds different elution mechanisms occurred. Analysis times on the monolithic column were, depending on the flow rate, two to seven times faster than on the particle-based column.
Table III
Table III shows the best models obtained using both descriptor types. For both descriptor sets, models of a better quality were obtained when the MLC chromatographic descriptor was included. The best models were generally obtained using E-Dragon descriptors, but they were again rather complex. Models with log kparticle or log km8 (equations 6 and 7) showed comparable results. Both models presented an added value of the chromatographic descriptor to the corresponding pure in silico models (equation 2), and were therefore suitable to apply in skin permeability applications. Since the monolithic column provided faster analysis, the MLC method on that stationary phase was preferred as a means to estimate skin permeability.
Biomimicking stationary phases can also be considered to predict skin permeability. These phases contain certain skin components, thus mimicking the skin structure. An immobilized artificial membrane (IAM) phase, which contains phospholipid analogues, and a cholesterol-bonded phase, which contains cholesteryl groups, were applied in a next step (13). Stationary phases with sub-2-µm particles, applied in UHPLC, allow faster analyses with an increased efficiency, and can therefore provide a high-throughput method to predict skin permeability. UHPLC on a C18 column with sub-2-µm particles was therefore also considered as a fast alternative method (13).
Different fractions of organic modifier were tested to estimate the extrapolated log kw values, which were used to model the skin permeability coefficient. An attempt was made to improve the modeling of skin permeability by incorporating theoretical molecular descriptors. Subsequently the methods were compared to select the most suitable. Furthermore, the high-throughput properties of the three methods were also compared.
The correlation between the retention (log kIAM) on the IAM column and the skin permeability coefficient, log Kp, was again relatively low. The correlation increased with higher fractions of organic modifier in the mobile phase, with a maximum at 40% acetonitrile (r = 0.483). On the cholesterol column, the highest correlation with log Kp was obtained with the mobile phase containing 50% acetonitrile (r = 0.461). In UHPLC, the log kBEH C18 values also showed a low correlation with the skin permeability, and again slightly better results were found with higher percentages of organic modifier. The highest correlation was seen with 40% acetonitrile, with r = 0.345.
Table IV
Table IV shows the best models obtained with each of the three methods. The best MLR model with Vega ZZ descriptors was obtained with the IAM chromatographic descriptor (equation 8). When comparing these results with a skin permeability model containing solely theoretical descriptors (equation 1), only a minor improvement was noticed in the performance parameters of the IAM model. Better, but again more complex, models were obtained from a stepwise MLR approach combining chromatographic and E-Dragon descriptors. Although these models are more complex and have a risk for overfitting (though RMSECV values are comparable to RMSEC), all performance parameters were again superior to the models with Vega ZZ descriptors. The best model was obtained with a UHPLC descriptor (equation 11) combined with 15 E-Dragon descriptors, and this was also the overall best model of this section. Figure 2 demonstrates that for this model, the predicted and experimental log Kp values show a very high correlation.
Figure 2: Predicted log Kp as a function of the experimental, together with the regression line (solid line) and the bisector (dashed line) for the best MLR model containing E-Dragon descriptors and a chromatographic UHPLC descriptor. Adapted and reproduced with permission from reference 13.
The three best MLR models comprising E-Dragon descriptors and a chromatographic one (equations 11–13) showed a clear improvement in performance parameters compared to the best model containing only E-Dragon descriptors. When comparing the three methods for their analysis time, the UHPLC method was three to four times faster than the IAM method and four to five times than the cholesterol-bonded method. The UHPLC method therefore presents an advantage regarding high-throughput, and is therefore the preferred method to generate a chromatographic descriptor for skin permeability predictions.
SFC is a separation technique in which a supercritical (or subcritical) fluid, obtained by applying conditions above the critical point of a substance, is used as the mobile phase. SFC forms an intermediate technique between liquid and gas chromatography. Most often, supercritical carbon dioxide (CO2) is used as the mobile phase basis because of its easily reachable critical temperature and pressure (31 °C and 74 bar). SFC offers many advantages, such as fast analyses, and it has green characteristics on top of highly efficient performances. Furthermore, a broad range of stationary phases can be applied. The aim of this last part was to explore whether SFC could provide a fast and equally good alternative to UHPLC to model skin permeability(14).
The column selection was based on a set of dissimilar columns proposed previously by Galea et al. (15,16). The stationary phases were composed of bare silica, along with 2-ethylpyridine-, cyanopropyl- (cyano, CN), pentafluorophenylpropyl- (pentafluorophenyl), aminopropyl-(amino), or phenyl-oxypropyl groups (phenyl) bonded to silica, crosslinked diol groups linked to silica (diol), or ethylene bridged hybrid silica (BEH). A cholesterol-bonded phase was added as the ninth column.
In the first step, a general gradient was applied on a small selection of compounds to determine a suitable organic modifier fraction in the mobile phase. For the cholesterol, cyanopropyl, and pentafluorophenyl columns, 5% (v/v) methanol was selected, while 10% (v/v) was used for the other stationary phases. The complete test set was then screened isocratically, and the retention factors on the different stationary phases were used to model the skin permeability.
Table V
Concerning the correlation between log k and log Kp (Table V), the lowest was obtained with the 2-ethylpyridine stationary phase, while the highest was observed with the cyanopropyl phase. An inversely proportional relationship between the log k and log Kp values was noticed, that is, compounds with higher skin permeability values showed lower retention and thus less affinity for the more polar (or intermediately polar) stationary phases applied in this study.
Table VI
When modeling was performed using the Vega ZZ descriptors, the overall best model was obtained including the cyanopropyl descriptor (Table VI, equation 14), as could be expected from the correlation between the retention factors and Kp values. When this MLR model was compared to the best models obtained in previous studies with the same test set, there was an improvement compared to the best MLR model containing only Vega ZZ descriptors, and also relative to the best models containing a chromatographic descriptor from MLC and IAM.
The best model with E-Dragon descriptors was obtained with the results from the phenyl column (Table VI, equation 15), showing very good fit and predictive properties. This model also performed better than the MLR model containing solely E-Dragon descriptors (equation 2). An enhancement of the performance parameters was also noticed compared to the best model with the MLC descriptor (equation 7), while similar performance parameters were obtained as the best model with a UHPLC descriptor (equation 11).
However, with 19 descriptors this was also a very complex model. A simplification of all models was therefore performed by assessing the RMSECV vs. the model complexity in the first instance. This was done by evaluating whether ΔRMSECV was equal to or larger than 0.02 for each descriptor added to a simpler model based on the results from both leave-one-out cross validation and k-fold cross validation. In addition, a RMSECVCrit value was calculated to obtain simplified models. The least complex model from both approaches was finally selected.
The best model with the phenyl descriptor could be simplified to a model with only nine descriptors (equation 16). However, this distinct reduction in number of descriptors also led to a decreased quality of the performance parameters as the RMSECV and RMSEC clearly increased. Finally, the best simplified model was obtained with the results on the cyanopropyl column (equation 17). Comparison with the model based solely on theoretical descriptors (equation 2) revealed an improvement in model quality.
Various chromatographic methods were employed to determine a chromatographic descriptor for modeling the skin permeability of pharmaceutical and cosmetic compounds. A rather large test set was used to explore different chromatographic systems in RPLC, MLC, UHPLC, and SFC to find those that offer a high correlation with the skin permeability, and also provide high-throughput potential.
The best correlation between the retention factors and the skin permeability coefficients was obtained with the cyanopropyl column in SFC. However, as for all other tested methods, this correlation was considered insufficient to properly predict the skin permeability, and the addition of theoretical descriptors was necessary to improve the models. Several modeling approaches were tested, but the overall best models were obtained with stepwise MLR on the E-Dragon descriptors. The MLR models with Vega ZZ descriptors provided simpler (too simple) models, but of an average quality. This was mainly due to the MLR approach included in the software. Moreover, there are also fewer possibilities for descriptor combinations because of the lower number of descriptors to build models with.
In terms of high-throughput potential, the MLC method on the monolithic column, as well as the UHPLC and SFC methods, provided favorable results. In addition, the overall best model of the entire project was obtained with the results from a phenyl column in SFC and E-Dragon descriptors. Similar results were observed for the equivalent model with a UHPLC descriptor. Both models had better fit and predictive abilities than the corresponding model with only theoretical descriptors, showing the benefit of these chromatographic descriptors. As a result of the “green” and high-throughput characteristics of SFC, this technique was finally preferred to model skin permeability.
Overall, it was seen that very good, but also rather complex, models (with a risk of overfitting) were obtained by combining theoretical and chromatographic descriptors. Their simplification also resulted in models that were similar to models containing only theoretical descriptors. Therefore, it still remains to be evaluated whether the time invested to generate a chromatographic descriptor is justified, that is, whether the model quality improves sufficiently by the addition of such a descriptor. However, such a conclusion can only be drawn when the models have been validated using an external test set.
This work was supported by the Research Foundation Flanders (FWO) [grant 11D3518N for Y.G.].
(1) Kielhorn, J.; Melching-Kollmuß, S.; Mangelsdorf, I. Environmental Health Criteria 235: Dermal Absorption; World Health Organization Press, 2006. https://inchem.org/documents/ehc/ehc/ehc235.pdf
(2) OECD, Test No. 427: Skin Absorption: In Vivo Method, in: OECD Guidelines for the Testing of Chemicals, Section 4 (Organisation for Economics Co-Operation and Development, 2004).
(3) Pulsoni, I.; Lubda, M.; Aiello, M.; et al. Comparison Between Franz Diffusion Cell and a novel Micro-physiological System for In Vitro Penetration Assay Using Different Skin Models. SLAS Technol. 2022, 27, 161–171. DOI: 10.1016/j.slast.2021.12.006
(4) Ruela, A. L. M.; Perissinato, A. G.; Lino, M. E. S.; Mudrik, P. S.; Pereira, G. R. Evaluation of Skin Absorption of Drugs from Topical and Transdermal Formulations. Braz. J. Pharm. Sci. 2014, 52, 527–544. DOI: 10.1590/S1984-82502016000300018
(5) Flaten, G. E.; Palac, Z.; Engesland, A.; et al. In Vitro Skin Models as a Tool in Optimization of Drug Formulation. Eur. J. Pharm. Sci. 2015, 75, 10–24. DOI: 10.1016/j.ejps.2015.02.018
(6) Supe, S.; Takudage, P. Methods for Evaluating Penetration of Drug into the Skin: A Review. Skin Res. Technol. 2021, 27, 299–308. DOI: 10.1111/srt.12968
(7) Buzek, J.; Ask, B. Regulation (EC) No 1223/2009 of the European Parliament and of the Council of 30 November 2009 on Cosmetic Products. Off. J. Eur. Union L. 2009, 342, 59–209.
(8) Grooten, Y.; Sych, A.; Mangelings, D.; Vander Heyden, Y. Comparison of In-silico Modelling and Reversed-phase Liquid Chromatographic Retention on an Octadecyl Silica Column to Predict Skin Permeability of Pharmaceutical and Cosmetic Compounds. J. Pharm. Biomed. An. 2021, 201, 114095. DOI: 10.1016/j.jpba.2021.114095
(9) Poole, C. F. The Essence of Chromatography; Elsevier, 2003.
(10) Waters, L. J.; Shahzad, Y.; Stephenson, J. Modelling Skin Permeability with Micellar Liquid Chromatography. Eur. J. Pharm. Sci. 2013, 50, 335–340. DOI: 10.1016/j.ejps.2013.08.002
(11) Martínez-Pla, J. J.; Martín-Biosca, Y.; Sagrado, S.; Villanueva-Camañas, R. M.; Medina-Hernández, M. J. Evaluation of the pH Effect of Formulations on the Skin Permeability of Drugs by Biopartitioning Micellar Chromatography. J. Chromatogr. A 2004, 1047, 255–262. DOI: 10.1016/j.chroma.2004.07.011
(12) Grooten, Y.; Marcelis, Q.; Mangelings, D.; Vander Heyden, Y. Evaluating Micellar Liquid Chromatographic Methods on Octadecyl Particle-based and Monolithic Columns to Predict the Skin Permeation of Drug and Cosmetic Molecules. J. Chromatogr. A 2022, 1663, 462753. DOI: 10.1016/j.chroma.2021.462753
(13) Grooten, Y.; Mangelings, D.; Vander Heyden, Y. Predicting Skin Permeability of Pharmaceutical and Cosmetic Compounds Using Retention on Octadecyl, Cholesterol-bonded and Immobilized Artificial Membrane Columns.J. Chromatogr. A 2022,1676, 463271. DOI: 10.1016/j.chroma.2022.463271
(14) Grooten, Y.; Mangelings, D.; Vander Heyden, Y. Comparison of Supercritical Fluid Chromatographic Methods to Predict the Skin Permeability of Pharmaceutical and Cosmetic Compounds. J. Chromatogr. A 2023, 1692, 463855. DOI: 10.1016/j.chroma.2023.463855
(15) Galea, C.; Mangelings, D.; Vander Heyden, Y. Method Development for Impurity Profiling in SFC: The Selection of a Dissimilar Set of Stationary Phases. J. Pharm. Biomed. Anal. 2015, 111, 333–343. DOI: 10.1016/j.jpba.2014.12.043
(16) Muscat Galea, C.; Slosse, A.; Mangelings, D.; Vander Heyden, Y. Investigation of the Effect of Column Temperature and Back-pressure in Achiral Supercritical Fluid Chromatography Within the Context of Drug Impurity Profiling.J. Chromatogr. A 2017, 1518, 78–88. DOI: 10.1016/j.chroma.2017.08.008
Debby Mangelings © Image courtesy of author
Debby Mangelings is a professor in the research group of Analytical Chemistry, Applied Chemometrics and Molecular Modelling at the Vrije Universiteit Brussel (VUB) in Belgium. Her research focuses on the use of liquid chromatography (LC), supercritical fluid chromatography (SFC), and miniaturized separation techniques such as capillary electrophoresis (CE) and capillary electrochromatography (CEC) for a variety of pharmaceutically oriented applications. In the field of chiral separations, enantiorecognition mechanisms on polysaccharide chiral phases are currently studied. Other investigated topics are the introduction of novel sample pretreatment techniques in metabolomics and method development for targeted and untargeted metabolomics using LC–MS and SFC–MS.
Yasmine Grooten © Image courtesy of author
Yasmine Grooten obtained her master’s degree in drug development at the Vrije Universiteit Brussel. Afterwards, she obtained a PhD in pharmaceutical sciences in the research group Analytical Chemistry, Applied Chemometrics and Molecular Modelling of the same university, under the supervision of Yvan Vander Heyden and Debby Mangelings. Her research focuses on the development of different analytical methods to model skin permeability. She is currently working as team lead of the chemistry laboratory at the Association of Pharmacists Belgium (APB), focusing on the quality control of (compounded) medicines on the Belgian market and applying chromatographic techniques such as (U)HPLC and GC.
Yvan Vander Heyden © Image courtesy of author
Yvan Vander Heyden is a full professor at the Vrije Universiteit Brussel (VUB), Belgium. He obtained the degrees of pharmacist and Ph.D. in pharmaceutical sciences at the same university. He heads a research group on chemometrics and separation science. The mission statement of his group is: “The rational use of new techniques in pharmaceutical analysis”, where analysis techniques refer both to separations and data-analysis. In separation science, research has been performed using different techniques such as CE, normal-phase liquid chromatography (NPLC), RPLC, polar organic solvent chromatography (POSC), SFC, UHPLC, and CEC. The data analysis aspects of his research are mainly conducted on separation science data. They comprise, for instance, multivariate calibration aspects and similarity analysis on fingerprint data (usually from herbal samples) or the construction of predictive quantitative structure-property relationships. As both analytical chemistry and applied chemometrics are logistic sciences, most research is performed as a cooperation with other researchers (often from other domains).
Navigating the Impact of Funding Cuts: Voices from the Research Community
July 9th 2025Shifts in U.S. federal science funding—driven by the newly established Department of Government Efficiency (DOGE)—are causing delays, uncertainty, and program changes across academic and regulatory research institutions.
Researchers Explore Comprehensive Modelling Approach for Liquid-Liquid Chromatography
June 30th 2025Scientists from the Technical University of Munich and Lund University explored the use of new models for liquid-liquid chromatography based off pre-existing procedures for liquid-liquid extraction.
