Next Article in Journal
Query Optimization for Hybrid Plans in Row–Column Dual Store HTAP Databases
Previous Article in Journal
Reasoning-Centric Framework for Open-Set Wild Plant Recognition
Previous Article in Special Issue
A Rapid and Low-Organic Consumption Analytical Method for Doxycycline with Application to Dissolution and Permeability Studies
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Applied Sciences
Volume 16
Issue 11
10.3390/app16115295
applsci-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Application of Biomimetic IAM Chromatography and QSAR Modeling for Predicting Selected Properties of Potential Drugs and Plant Protection Products

by
Małgorzata Janicka
1,
Małgorzata Sztanke
2,
Anna Pachuta-Stec
3 and
Krzysztof Sztanke
4,*
1
Department of Physical Chemistry, Faculty of Chemistry, Institute of Chemical Science, Maria Curie-Skłodowska University, Maria Curie-Skłodowska Sq. 2, 20-031 Lublin, Poland
2
Department of Medical Chemistry, Medical University of Lublin, 4A Chodźki Street, 20-093 Lublin, Poland
3
Independent Radiopharmacy Unit, Faculty of Pharmacy, Medical University of Lublin, 4A Chodźki Street, 20-093 Lublin, Poland
4
Laboratory of Bioorganic Compounds Synthesis and Analysis, Medical University of Lublin, 4A Chodźki Street, 20-093 Lublin, Poland
*
Author to whom correspondence should be addressed.
Appl. Sci. 2026, 16(11), 5295; https://doi.org/10.3390/app16115295
Submission received: 30 April 2026 / Revised: 18 May 2026 / Accepted: 19 May 2026 / Published: 25 May 2026
(This article belongs to the Special Issue Research on Organic and Medicinal Chemistry, Second Edition)
Browse Figures
Versions Notes

Abstract

A hybrid method combining biomimetic liquid chromatography with immobilized artificial membrane (IAM) and quantitative structure–activity relationships (QSARs) was used to derive helpful models for predicting selected properties related to distribution (binding to human serum albumin (log Pw/HSA)) and absorption (skin permeation (log Kw/sp), plant cuticle permeation (log Pw/pc), and human intestinal permeability (Caco-2)), and therefore influencing the effectiveness or unwanted effects of 199 synthetic compounds that are regarded as potential drugs or plant protection products. The molecules under investigation—derivatives of 5H-6,7-dihydroimidazo [2,1-c][1,2,4]triazole, 7,8-dihydroimidazo[2,1-c][1,2,4]triazin-4(6H)-one, 2,6,7,8-tetrahydroimidazo[2,1-c][1,2,4]triazine-3,4-dione, 1H-1,2,4-triazole, carbamic and phenoxyacetic acid—differ in their properties but all meet the requirements for xenobiotics to be considered as medicinal products. Reliable high-concept models were developed, indicating lipophilicity, molecular size, electronic properties, and the number of rotatable bonds as descriptors that determine the biological properties of these compounds. These models have been optimized and cross-validated, confirming their reliability and high predictivity.

1. Introduction

The search for new bioactive substances, whether as potential drugs or plant protection products, is not merely a test of the skills of organic chemists. A more significant challenge is obtaining molecules with optimal properties that produce a desired effect with no or minimal side effects, and thus reliably predict the potential effectiveness of such substances. An experimental assessment of the LADME-Tox profile of new substances [1], particularly in screening studies, is unreasonable: it requires a huge amount of work, funding, and unethical animal testing. In this context, it is sensible to turn to alternative methods offered by modern science. In the search for new bioactive molecules, biomimetic liquid chromatography techniques are particularly useful. When combined with quantitative structure–activity relationship (QSAR) methodologies, they allow important properties of medicinal substances and/or plant protection products to be predicted with a high degree of probability. This combination of in vitro and in silico techniques accelerates research and reduces costs, eliminating or significantly reducing the need for animal testing [2,3,4,5]. In preclinical screening studies, it is essential to predict the key characteristics and potential properties of new substances.
One of the most important pharmacokinetic properties of a bioactive compound is its ability to cross the blood–brain barrier [6,7]. This information is essential and must be assessed for new bioactive substances, regardless of their intended use. For drugs intended to act peripherally, as well as plant protection products, good penetration of the blood–brain barrier is undesirable, as it can cause harmful side effects. However, for potential drugs intended to act on the central nervous system, good penetration into the brain is desirable and even necessary. Therefore, regardless of the objective in searching for new chemical entities, predicting their ability to cross the blood–brain barrier should be undertaken at the earliest possible stage of research. One of the key properties determining the distribution of a xenobiotic in a living animal organism is its ability to bind to human serum albumin. Molecules bound to albumin are distributed throughout the body. Binding to albumin also affects the concentration of the unbound, free fraction of the endogenous substance, and thus its ability to interact with a pharmacoreceptor [8,9]. The rate of skin permeability [10,11] is an important parameter not only for potential medicinal products, such as creams, gels, or ointments, but also for potential plant protection products, as it indicates a potential risk to both animals and humans in the event of non-compliance with application procedures. Penetration through biological membranes, such as the plant cuticle [12,13] or the rate of absorption in the intestines [14,15], helps to assess both the potential for effective action and the risk of side effects of drug candidates or plant protection products. Our previous works [16] focused on predicting the penetration of the blood–brain barrier by new organic entities considered as potential drugs. In the present study, we aimed to predict the following properties of the structurally diversified bioactive molecules from various classes: the log PHSA parameter, which characterizes the binding of a compound to human serum albumin; the log Kw/sp parameter, which describes the penetration of a substance through the skin; the log Pw/pc parameter, which characterizes the penetration of a molecule through the plant cuticle; and Caco-2, which describes the rate of absorption of a given compound in the gastrointestinal tract.
This study focuses on 199 synthetic compounds with fully established structures and confirmed purity/homogeneity [16,17,18,19,20,21] (see Table 1 and Table 2) that are structurally related within each set. These small molecules are still under investigation for their potential applications as drugs or plant protection products. The majority of them are derivatives of 5H-6,7-dihydroimidazo[2,1-c][1,2,4]triazole (1–14), 7,8-dihydroimidazo[2,1-c][1,2,4]triazin-4(6H)-one (15–126), 2,6,7,8-tetrahydroimidazo[2,1-c][1,2,4]triazine-3,4-dione (127–134), and 1H-1,2,4-triazole (135–155), and the remaining ones—derivatives of carbamic (156–164) and phenoxyacetic acid (165–199). The molecules 1–155 are of pharmaceutical interest for potential anticancer, antibacterial, antiviral, antioxidant, or analgesic agents [16,18,19]. A large number of pharmacologically active compounds subjected to present investigations may be regarded as promising candidate drugs, because they are safe for embryos and larvae of zebrafish (Danio rerio), red blood cells, and non-tumor cells [16]. Among them is structure 53, which is low in toxicity in normal cells as well as in mice, and was identified as an antagonist of A2A adenosine receptors [20] and patented as a possible pharmaceutical against hepatic cirrhosis. However, compounds 156–199 are regarded as potential plant protection products, along with some molecules already approved for the control of weeds and plant diseases [17,21].

2. Materials and Methods

2.1. Chromatographic Measurements

All retention parameters, i.e., the log kw,IAM values, were extrapolated (compounds 1–170) or measured (compounds 171–199) using buffer (0.01 M disodium hydrogen phosphate and 0.02 M citric acid; pH = 7.4) as the mobile phase, and an IAM column as the stationary phase, in our earlier investigations. These values are given in Table 3. For details see [16,17,18,19,21].
In the case of compounds 171–199, due to low lipophilicity, the log kw,IAM values were measured using buffer as the mobile phase [17]. In all other cases, these values were calculated using the following Soczewiński–Wachtmeister’s equation [22]:
log k = log k w − s φ
where φ is the volume fraction of organic modifier (acetonitrile) in the mobile phase. The log kw values were evaluated as linear regression coefficients.

2.2. In Silico Calculations

Molecular weight (MW), topological polar surface area (TPSA), polarizability (α), parachor (Ƥ), the numbers of hydrogen bond donors and acceptors (HBD and HBA), and rotatable bonds (NRB) of the tested compounds, as well as biological parameters, such as log PHSA, log Kw/sp, log Pw/pc, and Caco-2 (Table 3), were calculated according to their 2D molecular structures (Table 1 and Table 2) by using ACD/Percepta software, version 1994–2012 (ACD/Labs, Advanced Chemistry Development, Inc., Toronto, ON, Canada).

2.3. Statistical Analysis

Linear regression (LR), multiple linear regression (MLR), and cross-validation were performed employing statistical software Minitab 16 (Minitab Inc., State College, PA, USA).

2.4. QSAR Modeling

The methodology of quantitative structure–activity relationships (QSARs) is based on the assumption that a compound’s biological activity depends on its lipophilic, electronic, and structural properties, which can be expressed as follows:
log SP = aA + bB + cC + ⋯ + const = ƒ (lipophilic, electronic, steric properties)
where log SP is a parameter describing a given property, e.g., log K, log P, etc. Lipophilic, electronic, or steric properties can be assessed using various parameters, whether measured (experimentally) or calculated (in silico). The lipophilicity of a substance allows the n-octanol/water partition coefficient log Po/w or the chromatographic retention factor log kw to be assessed. The number of hydrogen bond donors and acceptors (HBD and HBA) and the topological polar surface area (TPSA) are typical electronic descriptors when molar mass, polarizability, and parachor characterize steric properties (molecule size). In turn, the number of rotatable bonds allows the flexibility of the molecule to be assessed.
Most often, the model presented in Equation (4) takes the form of a polynomial derived using multiple linear regression (MLR) or partial least squares (PLS). Deriving the model, one should use reliable and accurate data, select descriptors (independent variables), and validate the resulting equation [23,24]. The independent variables used in the model must allow for the differentiation of the compounds under investigation and must not be correlated with one another. It is also important to minimize the number of independent variables so that the model is not overloaded.
Inter-dependencies between independent variables should be checked using statistical methods, e.g., cluster analysis, principal component analysis (PCA), or the variance inflation factor (VIF). The derived model undergoes a preliminary statistical assessment. The most important parameters of this assessment are as follows: the coefficient of determination R2, which allows the fit of the model to the input data to be assessed; the adjusted R2 (R2adj) and the predicted R-squared value (R2pred), which allow for a comparison of the fit of regression models containing different numbers of independent variables (R2adj) and an assessment of the reliability of the forecasts generated by the resulting model (R2pred) [25]. An important factor is the predicted residual error sum of squares (PRESS) because it is a good estimate of the real prediction error of the model. It assesses a model’s predictive ability; in general, the smaller the PRESS value, the better the model’s predictive ability [26]. The calculated global PRESS value must be lower than the sum of the squares of the response values of the total observations (SS). This proves that the developed models predict better than chance [27]. The derived QSAR models should be verified. In our research, we typically use leave-out cross-validation, which allows us to assess the fit of the derived model by excluding one (usually more) variable (s) from the dataset and comparing the statistical parameters obtained for the relationships derived from the complete and incomplete datasets.
The application domain (AD) [28,29] indicates the range within which the model allows predictions to be made with a specified probability and exclusion of outliers. In the present work, we used the leverage approach (Williams plot), where the warning leverage h* was calculated according to
h ∗ = 3 ( p + 1 ) n
where n is the total number of samples in, and p is the number of descriptors involved in the correlation [30].
In recent years, there have been increasing reports of the use of artificial neural networks (ANNs) in the development of QSAR models. Such models are exceptionally reliable and highly predictive; however, they require more sophisticated computational tools [31,32,33,34].

3. Results and Discussion

3.1. Molecular In Silico Properties

The molecular properties of the tested small molecules are diverse and vary across fairly wide ranges (Table 3). However, all of them meet the recommendations for drug-like compounds provided by the fairly general rule of five (Ro5) by Lipiński [35], as well as more detailed guidelines provided by Ghose et al. [36], Clark [37], van de Waterbeemd et al. [38], and Kelder et al. [39] regarding molar masses and TPSA values. Molecular weights (MW) range from 166.17 to 430.11 g mol−1, with an average molecular weight of 302.94 g mol−1. The values of the topological polar surface area (TPSA) range from 29.1 to 117.62 Å2, with an average value of 63.16 Å2. The values of polarizability (α) and parachor (Ƥ), parameters characterizing the size of the molecule, range from 15.51 to 41.7 Å3 (the average value of α is 31.45 Å3) and from 322.5 to 792.07 m3mol−1 (the average value is Ƥ = 584.74 m3mol−1). The number of hydrogen bond donors and acceptors (HBD and HBA) falls within the ranges of 0–3 (HBD) and 2–9 (HBA). The sum of the hydrogen bond donors and acceptors (HBD + HBA) ranges from 3 to 10, except for molecules 51–60 bearing hydrazide functionalities, for which the sum of HBD and HBA is 11 or 12. The number of hydrogen bond donors (HBD) and hydrogen bond acceptors (HBA) also follows the guidelines of Ro5 [35], Krajl et al. [40], van de Waterbeemd [41], and Ajay et al. [42] regarding compounds that activate the central nervous system. The presence of hydrogen bond acceptors (HBA) and donors (HBD) in the molecules of the tested substances results in their basic or acidic nature (HBA > 0 and HBD > 0). Compounds whose molecules contain hydrogen bond acceptors (HBA > 0) but lack hydrogen bond donors (HBD = 0) are basic in nature (Table 3). The structural parameters described above are relevant for xenobiotics considered as potential drugs. However, in the case of potential plant protection products, they indicate a real risk of possible or undesirable side effects on the environment, including humans.

3.2. Biological Properties

In our previous study [16], we developed a model to predict blood–brain barrier penetration, combining experimental chromatographic data obtained using the IAM column with QSAR (a quantitative structure–activity relationship) methodology. This model was developed for 126 molecules, all of which are included in the studies described in this article. The log BB values (where BB = the solute concentration in the brain/solute concentration in the blood), calculated for the compounds currently under study according to the aforementioned model, are presented in Table 3. Analysis of these data indicates that 59 small molecules (29.6%) have poor penetration (log BB < 0), while 140 (70.4%) penetrate the brain fairly well (log BB > 0) (Figure 1).
Based on the molecular structures (Table 1 and Table 2), selected biological parameters of the compounds were calculated as follows (Table 3): the log PHSA value, characterizing the distribution of the molecule between water and human serum albumin; the log Kw/sp parameter, describing the permeation (p) of the substance from water (w) through the skin (s); the log Pw/pc parameter, describing the permeation of the molecule from water through the plant cuticle (pc); and the Caco-2 parameter, characterizing the rate of intestinal absorption of the compound.
Most of the tested small molecules (76.9%) bind well to human serum albumin, exhibiting log PHSA > 0. A smaller group (23.1%) binds poorly to albumin (Figure 2). Good binding to albumin is essential for potential drugs but undesirable for plant protection products. The substances bound to albumin can circulate within the blood in a living organism and ultimately reach the target site. Excessively strong binding of pharmaceutical substances to albumin may hinder binding to the pharmacoreceptor and prevent the desired therapeutic effect from occurring. The log PHSA values of the tested compounds range from −1.796 to 2.697.
Substance permeation through the skin is characterized by log Kw/sp parameters, ranging from −10.1 to −2.679. Most compounds investigated (76.9%) exhibit log Kw/sp values in the range from −7 to −5, while 9.5% have lower and 13.6% have higher values (Figure 3).
Plant cuticle permeability, as described by the log Pw/pc parameter, varies for the tested substances in a similar manner to skin permeation (log Kw/sp); for most of them (84.4%), the log Pw/pc values fall within the range of 0–5. Only 10% of the compounds studied have log Pw/pc values less than 0, and 5.6% have a Pw/pc greater than 5 (Figure 4).
The rate of intestinal absorption is described by the Caco-2 parameter, with values ranging from 0.2 to 245 × 10−6 cm/s. Most compounds (86.4%) exhibit high permeability, while only 1.5% have low and 12.1% moderate permeability (Figure 5) [43].
The analysis of the presented graphs and numerical data reveals that most substances bind well to human serum albumin, cross the blood–brain barrier, penetrate the skin and plant cuticle well, and are well absorbed in the intestines.

3.3. Chromatographic Parameters

Retention parameters, determined (measured or calculated) using reversed-phase liquid chromatography with pure water (buffer) as the mobile phase, are accepted and recognized as descriptors of the lipophilic properties of substances, serving as an alternative to the n-octanol/water partition coefficient (log Po/w). The theoretical basis here is the definition of the retention coefficient (k), according to the following equation:
  k = n s n m
where n is the amount of substance (number of moles, n) in the stationary phase (s) and in the mobile phase (m).
If the concentration of the solute in the stationary phase is cs and in the mobile phase is cm, and the volume of the stationary phase is Vs and that of the mobile phase is Vm, then the following relationship holds:
  k = c s c m V s V m
where the value Vs/Vm is the phase ratio φ. For a non-polar stationary phase and water (buffer) as the mobile phase, the expression cs/cm corresponds to the definition of the partition coefficient (P) of a chromatographic substance in two immiscible solvents (e.g., n-octanol and water), as given by Nernst’s law.
The verification of retention parameters as descriptors of lipophilicity involves comparing them with the values of the n-octanol/water partition coefficients. For the compounds studied, we compared the chromatographic lipophilicity parameters log kw, determined for the IAM column (log kw,IAM), and buffer as the mobile phase with the log Po/w parameters calculated based on the molecular structures of these substances (Table 3, Figure 6). The results of this comparison are not impressive. The reasons may lie both in the chromatographic data and in the in silico calculated log Po/w values. In screening studies, when searching for new biologically active substances, determining the log Po/w values experimentally is not justified. It is an expensive, time-consuming, and very demanding technique. A sensible approach is to predict partitioning lipophilicity log Po/w values based on the structures of the substances under investigation. Estimates of partition coefficients can be made using a variety of methods, including fragment-based, atom-based, or knowledge-based methods. Various platforms and software packages offering such services are currently available on the market (https://www.acdlabs.com/products/percepta-platform/physchem-suite/logp/ (accessed on 20 April 2026)).
However, log Po/w values predicted in silico are not always entirely reliable, particularly for molecules with a more complex structure or for isomers. There may also be doubts regarding experimentally determined retention factors in buffer, particularly when an extrapolation method has been used for this purpose, as this depends on the choice of extrapolation function, the concentration range, and the type of organic modifier in the mobile phase.
Of the 199 substances studied, the majority are moderately lipophilic with log kw,IAM ≥ 1 and log Po/w ≥ 1; the remaining compounds are weakly lipophilic and water-soluble with log kw,IAM and log Po/w < l (Figure 6B). As shown in the data presented in Figure 6B, experimental chromatographic parameters (log kw,IAM) classify fewer substances as ‘lipophilic’, or in other words, they consider them to be less lipophilic than the in silico log Po/w parameters suggest.
It should be noted that due to the presence of a phosphate group in the phosphatidylcholine molecule bound to the surface of silica gel, the chromatographic system using an IAM column is not identical to the n-octanol/water partition system. In this case, the retention of the substance is not solely the result of hydrophobic interactions but also of possible specific interactions. As a result, chromatographic systems equipped with an IAM column do not so much predict the lipophilic properties of substances as their behavior in biological separation systems. The similarity between the artificial membrane and biological membranes is a major advantage in this case [44,45].

3.4. Derived QSAR Models

In the present study, helpful QSAR models for predicting the biological properties of the test substances (i.e., log PHSA, log Kw/sp, log Pw/pc, and Caco-2 values) were derived using the MLR method. For this purpose, it was necessary to identify the independent variables and examine the correlations between them. In this case, we applied similarity analysis, and the results obtained are presented as a dendrogram in Figure 7. This figure shows clusters grouping individual structural descriptors as follows: cluster I combines molecular weight (MW), polarizability (α), and parachor (Ƥ), characterizes the size of the molecule, and has a similarity of 95.6%; cluster II combines electronic descriptors (TPSA and HBD + HBA) and has a similarity of 93.9%. In Figure 7, two parameters, NRB and chromatographic lipophilicity log kw,IAM, form two single-element clusters.
In the developed models, one parameter from each cluster was included to ensure that the model did not contain more than one descriptor characterizing the same property. Using this procedure, we examined all possible combinations of independent variables (for details see [16,17]. Ultimately, analyzing the statistical parameters of all models under consideration allowed us to select the best models—those with the most favorable statistical parameters. These models are presented in Table 4 and Table 5, which show the statistical parameters that were calculated for them. From among the most favorable MI-MVIII models, optimal models were selected, i.e., those with the smallest number of independent variables and, at the same time, with the most favorable values for the statistical parameters: the highest R2, R2adj, and R2pred values and the lowest Δ values, representing the smallest difference between the R2 and R2pred parameters (Figure 8). A low Δ value indicates model stability and high predictive power of the model. As mentioned earlier, a lower PRESS value is also an indicator. In all cases, the PRESS values are lower than the SS, demonstrating that the developed models perform better than random chance [26]. Analysis of the results obtained allowed the MIII, MV, and MVII models to be assessed as being very good and highly predictive. The MVIII model, developed to predict the intestinal absorption of the tested compounds (Caco-2 values), is significantly poorer; that is, it is only moderately fitted to the input data (R2 < 0.8) and has no predictive power (R2pred < 0.8).
The selected models were subjected to leave-30–out cross-validation, as illustrated by the graphs shown in Figure 9, Figure 10, Figure 11 and Figure 12.
The MIII model allows the binding affinity of the tested substances to human serum albumin to be predicted, i.e., the log PHSA values. As illustrated by the graph shown in Figure 9B, the binding of the tested compounds to albumin is affected by electronic properties, i.e., the sum of the hydrogen bond donors and acceptors (HBD + HBD), the molecular size described by the parachor value (Ƥ), and the chromatographic lipophilicity (log kw,IAM). However, an increase in molecular size strongly increases the log PHSA values, while an increase in the sum of HBD + HBA equally strongly reduces this value. In this case, lipophilicity has a significantly smaller and negative effect on the log PHSA values. As mentioned earlier, it should be noted that the log kw,IAM parameters characterize not so much the lipophilic properties of the tested substances as their affinity for the IAM stationary phase, which mimics biological membranes. Furthermore, the conclusions drawn from the results are limited to molecules whose properties fall within the specified AD range. It can, therefore, be concluded that larger molecules with fewer hydrogen bond donors and/or acceptors bind more effectively to human serum albumins.
The permeation through the skin, i.e., the log Kw/sp values, is predicted by the MV model (Figure 10). According to this model, log Kw/sp values depend on the lipophilicity of the substance, the particle size, the number of hydrogen bond donors and acceptors, and the molecular flexibility described by the number of rotatable bonds (NRB). As illustrated in Figure 10B, lipophilicity, molecular size, and molecular flexibility increase the rate of substance permeation through the skin, while polarity, expressed as the sum of hydrogen bond donors and acceptors, reduces it. In this case, too, chromatographic lipophilicity is not the dominant factor.
The permeability through the plant cuticle, described by the log Pw/pc parameter, depends in a very similar way on the properties of these compounds: an increase in lipophilicity, molecular size, and flexibility increases the log Pw/pc values, while polarity, dependent on the number of hydrogen bond donors and acceptors, reduces them. This means that the permeability of substances through the plant cuticle is favored by a larger molecule size, flexibility, and lipophilicity, while polarity is unfavorable. As before, lipophilicity is not the dominant factor.
In the MIII, MV, and MVII models presented, particle size is described by the parachor value. Statistical parameters (Table 5) indicate that replacing this descriptor with polarizability slightly but noticeably degrades the quality of the models.
As mentioned earlier, the MVIII model, developed to predict the intestinal absorption of the tested compounds, is statistically weak and lacks predictive power. However, by analyzing the graph shown in Figure 12B, it is possible to identify those properties of compounds that are likely to influence the Caco-2 values. Analysis of the graph indicates that an increase in lipophilicity and molecular size, as expressed by the parachor value, increases the Caco-2 values, i.e., the rate of intestinal absorption. Conversely, high molecular flexibility and polarity do not favor this process. As before, lipophilicity is not the dominant factor.
For all the models evaluated, we observe very good agreement between the in silico values (Table 3) and those calculated using the derived models, including for the validated models (Figure 9A, Figure 10A, Figure 11A and Figure 12A). In all cases, the applicability domains AD (Figure 9C, Figure 10C, Figure 11C and Figure 12C), limited by the values ± SD and the warning leverage h*, indicate that all the models are valid within the domain for which they were developed. In Figure 10C, Figure 11C and Figure 12C, one carbamic acid derivative (158) behaves as an outlier. It is likely that the presence of a single –F atom in the molecule promotes specific interactions with the stationary phase, abnormally increasing affinity to the stationary phase and the log kw/IAM values.

4. Conclusions

In the present study, a biomimetic liquid chromatography technique using the IAM column and quantitative structure–activity relationship (QSAR) method was employed to predict key parameters related to absorption and distribution, namely log Kw/sp, log Pw/pc, Caco-2, and log PHSA, for 199 compounds belonging to various bioactive sets. All compounds, although differing in chemical structure and physicochemical properties, meet the requirements for drug-like molecules. Statistical parameters and cross-validation allow the selected models to be considered robust and highly predictive. The derived models enable the prediction of a compound’s binding to human serum albumin and its penetration through biological membranes (skin and plant cuticle), and identify the physicochemical properties of the substance that dominate these processes. Permeation of the analyzed molecules through the skin and plant cuticle, as well as their binding to human serum albumin, enhances significantly with increasing molecular size, and decreases significantly with increasing polarity. The flexibility of molecules, described by NRB, has no effect on binding to albumin but significantly increases their permeation through biological membranes. The chromatographic lipophilicity of compounds, expressed by the log kw,IAM parameters, does not have a dominant influence on the biological parameters under investigation—the conclusions drawn from the conducted studies concern substances whose structural properties fall within the defined applicability domain. The combination of experimental data obtained from a biomimetic chromatographic system with QSAR methodology enables the prediction of key biological parameters of organic compounds influencing their efficacy or side effects. As such, it is highly useful in screening studies, guiding the search for new compounds with desired properties.

Author Contributions

Conceptualization, M.J.; Methodology, M.J.; Software, M.J.; Validation, M.J.; Formal analysis, M.J.; Investigation, M.J.; Resources, M.J., M.S., A.P.-S. and K.S.; Writing—Original Draft, M.J., M.S. and K.S.; Writing—Review and Editing, M.J., M.S. and K.S.; Funding acquisition, K.S. All authors have read and agreed to the published version of the manuscript.

Funding

This research received no external funding.

Institutional Review Board Statement

Not applicable.

Informed Consent Statement

Not applicable.

Data Availability Statement

Data are contained within the article.

Conflicts of Interest

The authors declare no conflicts of interest.

References

  1. Chackalamannil, S.; Rotella, D.; Ward, S.E. (Eds.) Comprehensive Medicinal Chemistry III; Elsevier: Amsterdam, The Netherlands, 2017. [Google Scholar]
  2. Jiang, Z.; Reilly, J. Chromatography approaches for early screening of the phospholipidosis-inducing potential of pharmaceuticals. J. Pharm. Biomed. Anal. 2012, 6, 184–190. [Google Scholar] [CrossRef]
  3. Valko, K.; Teague, S.; Pidgeon, C. In vitro membrane binding and protein binding (IAM MB/PB technology) to estimate in vivo distribution: Applications in early drug discovery. ADMET DMPK 2017, 5, 14–38. [Google Scholar] [CrossRef]
  4. Tsopelas, F.; Giaginis, C.; Tsantili-Kakoulidou, A. Lipophilicity and biomimetic properties to support drug discovery. Expert Opin. Drug Discov. 2017, 12, 885–896. [Google Scholar] [CrossRef]
  5. Grumetto, L.; Russo, G.; Barbato, F. Immobilized artificial membrane HPLC derived parameters vs. PAMPA-BBB data in estimating in situ measured blood-brain barrier permeation of drugs. Mol. Pharm. 2016, 13, 2808–2816. [Google Scholar] [CrossRef]
  6. De Vrieze, M.; Verzele, D.; Szucs, R.; Sandra, P.; Lynen, F. Evaluation of sphingomyelin, cholester, and phosphatidylcholine-based immobilized artificial membrane liquid chromatography to predict drug penetration across the blood-brain barrier. Anal. Bioanal. Chem. 2014, 406, 6179–6188. [Google Scholar] [CrossRef] [PubMed]
  7. Di, L. An update on the importance of plasma protein binding in drug discovery and development. Expert Opin. Drug Discov. 2021, 16, 1453–1465. [Google Scholar] [CrossRef]
  8. Talevi, A.; Bellera, C.L. Drug binding to plasma proteins. In The ADME Encyclopedia; Springer International Publishing: Cham, Switzerland, 2022; pp. 315–326. [Google Scholar]
  9. Tesseromatis, C.; Alevizou, A. The role of the protein-binding on the mode of drug action as well the interactions with 490 other drugs. Eur. J. Drug Metab. Pharmacokinet. 2008, 33, 225–230. [Google Scholar] [CrossRef]
  10. Potts, R.O.; Guy, R.H. Predicting skin permeability. Pharm. Res. 1992, 9, 663–669. [Google Scholar] [CrossRef] [PubMed]
  11. Sobańska, A.W.; Brzezińska, E. IAM chromatographic models of skin permeation. Molecules 2022, 27, 1893. [Google Scholar] [CrossRef] [PubMed]
  12. Schreiber, L.; Schönherr, J. Water Permeability. In Water and Solute Permeability of Plant Cuticles: Measurement and Data Analysis; Springer: Berlin/Heidelberg, Germany, 2009. [Google Scholar]
  13. Valeska Zeisler-Diehl, V.; Migdal, B.; Schreiber, L. Quantitative characterization of cuticular barrier properties: Methods, requirements, and problems. J. Exp. Bot. 2017, 68, 5281–5291. [Google Scholar] [CrossRef]
  14. Falcón-Cano, G.; Molina, C.; Cabrera-Pérez, M.Á. Reliable prediction of Caco-2 permeability by supervised recursive machine learning approaches. Pharmaceutics 2022, 14, 1998. [Google Scholar] [CrossRef]
  15. Al-Sulaiti, F.K.; Dogan, A.; Kir, F.; Sahin, S. An isocratic RP-HPLC method for simultaneous quantification of nateglinide and furosemide: Application to in vitro Caco-2 intestinal permeability studies of nanodrugs. Biomed. Chromatogr. 2025, 11, e70229. [Google Scholar] [CrossRef] [PubMed]
  16. Janicka, M.; Sztanke, M.; Sztanke, K. Modeling the blood-brain barrier permeability of potential heterocyclic drugs via biomimetic IAM chromatography technique combined with QSAR methodology. Molecules 2024, 29, 287. [Google Scholar] [CrossRef] [PubMed]
  17. Janicka, M.; Sztanke, M.; Sztanke, K. Biomimetic chromatography/QSAR investigations in modeling properties influencing the biological efficacy of phenoxyacetic acid-derived congeners. Molecules 2025, 30, 688. [Google Scholar] [CrossRef]
  18. Janicka, M.; Sztanke, M.; Sztanke, K. Reversed-phase liquid chromatography with octadecylsilyl, immobilized artificial membrane and cholesterol columns in correlation studies with in silico biological descriptors of newly synthesized antiproliferative and analgesic active compounds. J. Chromatogr. A 2013, 1318, 92–101. [Google Scholar] [CrossRef] [PubMed]
  19. Janicka, M.; Pachuta-Stec, A. Retention-property relationships of 1,2,4-triazoles by micellar and reversed-phase liquid chromatography. J. Sep. Sci. 2014, 37, 1419–1428. [Google Scholar] [CrossRef]
  20. Szuster-Ciesielska, A.; Sztanke, K.; Kandefer-Szerszeń, M. A novel fused 1,2,4-triazine aryl derivative as antioxidant and nonselective antagonist of adenosine A2A receptors in ethanol-activated liver stellate cells. Chem. Biol. Interact. 2012, 195, 18–24. [Google Scholar] [CrossRef]
  21. Janicka, M. Correlations between chromatographic parameters and bioactivity predictors of potential herbicides. J. Chromatogr. Sci. 2014, 52, 676–684. [Google Scholar] [CrossRef][Green Version]
  22. Soczewiński, E.; Wachtmeister, C.A. The relation between the composition of certain ternary two-phase solvent systems and RM values. J. Chromatogr. 1962, 7, 311–320. [Google Scholar] [CrossRef]
  23. Eriksson, L.; Jaworska, J.; Worth, A.P.; Cronin, M.T.; McDowell, R.M.; Gramatica, P. Methods for reliability and uncertainty assessment and for applicability evaluations of classification- and regression-based QSARs. Environ. Health Perspect. 2003, 111, 1361–1375. [Google Scholar] [CrossRef]
  24. Organization for Economic Co-Operation and Development. Guidance Document on the Validation of (Quantitative) Structure-Activity Relationship [(Q)SAR] Models; OECD Series on Testing and Assessment, No. 69; OECD Publishing: Paris, France, 2014. [Google Scholar]
  25. Frost, J. Regression Analysis: An Intuitive Guide for Using and Interpreting Linear Models; Statistics By Jim Publishing: Columbia, MD, USA, 2020. [Google Scholar]
  26. Chen, J.W.; Li, X.H.; Yu, H.Y.; Wang, Y.N.; Qiao, X.L. Progress and perspectives of quantitative structure–activity relationships used for ecological risk assessment of toxic organic compounds. Sci. China Ser. B-Chem. 2008, 51, 593–606. [Google Scholar] [CrossRef]
  27. Clementi, M.; Clementi, S.; Fornaciari, M.; Orlandi, F.; Romano, B. The GOLPE procedure for predicting olive crop production from climatic parameters. J. Chemom. 2001, 15, 397–404. [Google Scholar] [CrossRef]
  28. Roy, K.; Kar, S.; Ambure, P. On a simple approach for determining applicability domain of QSAR models. Chemom. Intell. Lab. Syst. 2015, 145, 22–29. [Google Scholar] [CrossRef]
  29. Sahigara, F.; Mansouri, K.; Ballabio, D.; Mauri, A.; Consonni, V.; Todeschini, R. Comparison of different approaches to define the applicability domain of QSAR models. Molecules 2012, 17, 4791–4810. [Google Scholar] [CrossRef]
  30. Abbott, N.J. Prediction of blood–brain permeation in drug discovery from in vivo, in vitro and in silico models. Drug Discov. Today Technol. 2004, 1, 407–416. [Google Scholar] [CrossRef]
  31. He, R.; Ma, H.; Zhao, W.; Qu, W.; Zhao, J.; Luo, L.; Zhu, W. Modeling the QSAR of ACE-inhibitory peptides with ANN and its applied illustration. Int. J. Pept. 2012, 2012, 620609. [Google Scholar] [CrossRef] [PubMed]
  32. Deng, B.; Ni, X.; Zhai, Z.; Tang, T.; Tan, C.; Yan, Y.; Deng, J.; Yin, Y. New quantitative structure-activity relationship model for angiotensin-converting enzyme inhibitory dipeptides based on integrated descriptors. J. Agric. Food Chem. 2017, 65, 9774–9781. [Google Scholar] [CrossRef]
  33. Singh, Y.R.; Nisterenko, W.; Fedorowicz, J.; Sączewski, J.; Szulczyk, D.; Greber, K.E.; Sawicki, W.; Ciura, K. Machine learning-driven QSRR modeling of albumin binding in fluoroquinolones: An SVR approach supported by HSA chromatography. Int. J. Mol. Sci. 2026, 27, 3700. [Google Scholar] [CrossRef]
  34. Mendenhall, J.; Meiler, J. Improving quantitative structure-activity relationship models using artificial neural networks trained with dropout. J. Comput. Aided Mol. Des. 2016, 30, 177–189. [Google Scholar] [CrossRef] [PubMed]
  35. Lipinski, C.A.; Lombardo, F.; Dominy, B.W.; Feeney, P.J. Experimental and computational approaches to estimate solubility and permeability in drug discovery and development settings. Adv. Drug Deliv. Rev. 1997, 23, 3–25. [Google Scholar] [CrossRef]
  36. Ghose, A.K.; Viswanadhan, V.N.; Wendoloski, J.J. A knowledge-based approach in designing combinatorial or medicinal chemistry libraries for drug discovery. 1. A qualitative and quantitative characterization of known drug databases. J. Comb. Chem. 1999, 1, 55–68. [Google Scholar] [CrossRef]
  37. Clark, D.E. In silico prediction of blood–brain barrier permeation. Drug Discov. Today 2003, 15, 927–933. [Google Scholar] [CrossRef]
  38. van de Waterbeemd, H.; Camenish, G.; Folkers, G.; Chretien, J.R.; Raevsky, O.A. Estimation of blood-brain barrier crossing of drugs using molecular size and shape, and H-bonding descriptors. J. Drugs Target. 1998, 6, 151–165. [Google Scholar] [CrossRef]
  39. Kelder, J.; Grootenhuis, P.D.; Bayada, D.M.; Delbressine, L.P.; Ploemen, J.P. Polar molecular surface as dominating determinant for oral absorption and brain penetration of drugs. Pharm. Res. 1999, 16, 1514–1519. [Google Scholar] [CrossRef]
  40. Kralj, S.; Jukic, M.; Bren, U. Molecular filters in medicinal chemistry. Encyclopedia 2023, 3, 501–511. [Google Scholar] [CrossRef]
  41. van de Waterbeemd, H.; Smith, D.A.; Beaumont, K.; Walker, D.K. Property-based design: Optimization of drug absorption and pharmacokinetics. J. Med. Chem. 2001, 44, 1313–1333. [Google Scholar] [CrossRef] [PubMed]
  42. Ajay; Bemis, G.W.; Murcko, M.A. Designing libraries with CNS activity. J. Med. Chem. 1999, 42, 4942–4951. [Google Scholar] [CrossRef]
  43. Yamashita, S.; Furubayashi, T.; Kataoka, M.; Sakane, T.; Sezaki, H.; Tokuda, H. Optimized conditions for prediction of intestinal drug permeability using Caco-2 cells. Eur. J. Pharm. Sci. 2000, 10, 195–204. [Google Scholar] [CrossRef] [PubMed]
  44. Sobańska, A.W.; Brzezińska, E. Phospholipid-based immobilized artificial membrane (IAM) chromatography: A powerful tool to model drug distribution processes. Curr. Pharm. Des. 2017, 23, 6784–6794. [Google Scholar] [CrossRef]
  45. Ermondi, G.; Vallaro, M.; Caron, G. Learning how to use IAM chromatography for predicting permeability. Eur. J. Pharm. Sci. 2018, 114, 385–390. [Google Scholar] [CrossRef]
Applsci 16 05295 g001
Figure 1. Calculated log BB values (Table 3). A: log BB < 0; B: log BB > 0.
Figure 1. Calculated log BB values (Table 3). A: log BB < 0; B: log BB > 0.
Applsci 16 05295 g001
Applsci 16 05295 g002
Figure 2. Calculated log PHSA values (Table 3). A: log PHSA < 0; B: log PHSA ϵ <0–1>; C: log PHSA > 1.
Figure 2. Calculated log PHSA values (Table 3). A: log PHSA < 0; B: log PHSA ϵ <0–1>; C: log PHSA > 1.
Applsci 16 05295 g002
Applsci 16 05295 g003
Figure 3. Calculated log Kw/sp values (Table 3). A: log Kw/sp < −7; B: log Kw/sp ϵ <−7–−5>; C: log Kw/sp > −5.
Figure 3. Calculated log Kw/sp values (Table 3). A: log Kw/sp < −7; B: log Kw/sp ϵ <−7–−5>; C: log Kw/sp > −5.
Applsci 16 05295 g003
Applsci 16 05295 g004
Figure 4. Calculated log Pw/pc values (Table 3). A: log Pw/pc < 0; B: log Pw/pc ϵ <0–5>; C: log Pw/pc > 5.
Figure 4. Calculated log Pw/pc values (Table 3). A: log Pw/pc < 0; B: log Pw/pc ϵ <0–5>; C: log Pw/pc > 5.
Applsci 16 05295 g004
Applsci 16 05295 g005
Figure 5. Calculated Caco-2 values (Table 3). A: Caco-2 < 0.4; B: Caco-2 ϵ <0.4–7>; C: Caco-2 > 7.
Figure 5. Calculated Caco-2 values (Table 3). A: Caco-2 < 0.4; B: Caco-2 ϵ <0.4–7>; C: Caco-2 > 7.
Applsci 16 05295 g005
Applsci 16 05295 g006
Figure 6. (A) log Po/w vs. log kw,IAM relationship, and (B) measured log kw,IAM values: (log kw,IAM and log Po/w < 1 (A), and log kw,IAM and log Po/w ≥ 1 (B).
Figure 6. (A) log Po/w vs. log kw,IAM relationship, and (B) measured log kw,IAM values: (log kw,IAM and log Po/w < 1 (A), and log kw,IAM and log Po/w ≥ 1 (B).
Applsci 16 05295 g006
Applsci 16 05295 g007
Figure 7. Dendrogram: similarities between molecular descriptors.
Figure 7. Dendrogram: similarities between molecular descriptors.
Applsci 16 05295 g007
Applsci 16 05295 g008
Figure 8. Comparison of the coefficients of determination (R2, R2adj, and R2pred) calculated for the derived QSAR models (MI-MVIII).
Figure 8. Comparison of the coefficients of determination (R2, R2adj, and R2pred) calculated for the derived QSAR models (MI-MVIII).
Applsci 16 05295 g008
Applsci 16 05295 g009
Figure 9. Observed vs. predicted log PHSA values (A), standardized coefficients (B), and Williams plot for applicability domain assessment (C) obtained for the MIII model.
Figure 9. Observed vs. predicted log PHSA values (A), standardized coefficients (B), and Williams plot for applicability domain assessment (C) obtained for the MIII model.
Applsci 16 05295 g009
Applsci 16 05295 g010
Figure 10. Observed vs. predicted log Kw/sp values (A), standardized coefficients (B), and Williams plot for applicability domain assessment (C) obtained for the MV model.
Figure 10. Observed vs. predicted log Kw/sp values (A), standardized coefficients (B), and Williams plot for applicability domain assessment (C) obtained for the MV model.
Applsci 16 05295 g010
Applsci 16 05295 g011
Figure 11. Observed vs. predicted log Pw/pc values (A), standardized coefficients (B), and Williams plot for applicability domain assessment (C) obtained for the MVII model.
Figure 11. Observed vs. predicted log Pw/pc values (A), standardized coefficients (B), and Williams plot for applicability domain assessment (C) obtained for the MVII model.
Applsci 16 05295 g011
Applsci 16 05295 g012
Figure 12. Observed vs. predicted Caco-2 values (A), standardized coefficients (B), and Williams plot for applicability domain assessment (C) obtained for the MVIII model.
Figure 12. Observed vs. predicted Caco-2 values (A), standardized coefficients (B), and Williams plot for applicability domain assessment (C) obtained for the MVIII model.
Applsci 16 05295 g012
Table 1. Structures of potential drugs (1–155): derivatives of 5H-6,7-dihydroimidazo[2,1-c][1,2,4]triazole (1–14), 7,8-dihydroimidazo[2,1-c][1,2,4]triazin-4(6H)-one (15–126), 2,6,7,8-tetrahydroimidazo[2,1-c][1,2,4]triazine-3,4-dione (127–134), and 1H-1,2,4-triazole (135–155).
Table 1. Structures of potential drugs (1–155): derivatives of 5H-6,7-dihydroimidazo[2,1-c][1,2,4]triazole (1–14), 7,8-dihydroimidazo[2,1-c][1,2,4]triazin-4(6H)-one (15–126), 2,6,7,8-tetrahydroimidazo[2,1-c][1,2,4]triazine-3,4-dione (127–134), and 1H-1,2,4-triazole (135–155).
Applsci 16 05295 i001
NoR1R2NoR1R2
1HH82-Me,4-ClC6H3OCH2H
2H4-Me92-Me,4-ClC6H3OCH24-Me
3H4-OMe104-ClC6H4OCH24-Cl
4H3-Cl112-Me,4-ClC6H3OCH24-Cl
5H3,4-Cl2122,4-Cl2C6H3OCH24-Cl
6PhOCH2H132,4-Cl2C6H3OCH23,4-Cl2
74-ClC6H4OCH24-OMe142,4,5-Cl3C6H2OCH24-Cl
Applsci 16 05295 i002
NoR1R2NoR1R2
15CF3H71furan-2-yl2,6-Cl2
16CF32-Me72thien-2-ylH
17CF34-Me73thien-2-yl2-Me
18CF32-OMe74thien-2-yl4-Me
19CF32-Cl75thien-2-yl2,3-diMe
20CF33-Cl76thien-2-yl2-OMe
21CF34-Cl77thien-2-yl2-Cl
22CF33,4-Cl278thien-2-yl3-Cl
23EtH79thien-2-yl4-Cl
24Et4-Me80thien-2-yl3,4-Cl2
25Et2-Cl81PhH
26Et3-Cl82Ph2-Me
27Et4-Cl83Ph3-Me
28Et3,4-Cl284Ph4-Me
29i-PrH85Ph2-OMe
30i-Pr4-Me86Ph4-OMe
31i-Pr2-Cl87Ph4-OEt
32i-Pr3-Cl88Ph2,3-diMe
33i-Pr4-Cl89Ph2-Cl
34i-Pr3,4-Cl290Ph3-Cl
35COOEtH91Ph4-Cl
36COOEt4-Me92Ph3,4-Cl2
37COOEt3-Cl934-NO2C6H4H
38COOEt4-Cl944-NO2C6H42-Me
39COOEt3,4-Cl2954-NO2C6H44-Me
40CH2COOMeH964-NO2C6H42-OMe
41CH2COOMe4-Me974-NO2C6H42,3-diMe
42CH2COOMe4-OMe984-NO2C6H42-Cl
43CH2COOMe4-OEt994-NO2C6H43-Cl
44CH2COOMe4-Cl1004-NO2C6H44-Cl
45CH2COOEtH1014-NO2C6H43,4-Cl2
46CH2COOEt4-Me102PhCH2H
47CH2COOEt2-OMe1032-ClC6H4CH2H
48CH2COOEt3-Cl1043-ClC6H4CH2H
49CH2COOEt4-Cl1054-ClC6H4CH2H
50CH2COOEt3,4-Cl2106PhCH24-Me
51CONHNH2H1074-MeC6H4CH24-Me
52CONHNH24-Me1083-MeC6H4CH24-Me
53CONHNH24-OMe1092-ClC6H4CH24-Me
54CONHNH23-Cl1103-ClC6H4CH24-Me
55CONHNH23,4-Cl21114-ClC6H4CH24-Me
56CH2CONHNH2H112PhCH24-OEt
57CH2CONHNH24-Me1134-MeC6H4CH24-OEt
58CH2CONHNH24-OMe1142-ClC6H4CH24-OEt
59CH2CONHNH24-OEt1153-ClC6H4CH24-OEt
60CH2CONHNH24-Cl1164-ClC6H4CH24-OEt
61furan-2-ylH1172-ClC6H4CH22-Me
62furan-2-yl2-Me118PhCH24-Cl
63furan-2-yl4-Me1192-ClC6H4CH24-Cl
64furan-2-yl2,3-diMe1203-ClC6H4CH24-Cl
65furan-2-yl2-OMe1214-ClC6H4CH24-Cl
66furan-2-yl4-OMe122PhCH2CH2H
67furan-2-yl2-Cl123PhCH2CH24-Me
68furan-2-yl3-Cl124PhCH2CH22-Cl
69furan-2-yl4-Cl125PhCH2CH24-Cl
70furan-2-yl3,4-Cl2126PhCH2CH23,4-Cl2
Applsci 16 05295 i003
NoRNoRNoRNoR
127H1294-Me1314-OMe1334-Cl
1282-Me1302-OMe1323-Cl1343,4-Cl2
Applsci 16 05295 i004
NoRNoRNoR
135Pr139Ph1432-ClC6H4
136CH(CH3)CH2CH31404-MeOC6H41444-BrC6H4
137Bu141C6H11
138PhCH2142PhCH2CH2
Applsci 16 05295 i005
NoRNoRNoR
145Pr149Ph1532-ClC6H4
146CH(CH3)CH2CH31504-MeOC6H41544-BrC6H4
147Bu151C6H111553-MeC6H4
148PhCH2152PhCH2CH2
Table 2. Structures of potential plant protection products (156–199): derivatives of carbamic acid (156–164) and phenoxyacetic acid (165–199).
Table 2. Structures of potential plant protection products (156–199): derivatives of carbamic acid (156–164) and phenoxyacetic acid (165–199).
Derivatives ofStructure/No
carbamic acidApplsci 16 05295 i006
156		Applsci 16 05295 i007
157		Applsci 16 05295 i008
158
Applsci 16 05295 i009
159		Applsci 16 05295 i010
160		Applsci 16 05295 i011
161
Applsci 16 05295 i012
162		Applsci 16 05295 i013
163		Applsci 16 05295 i014
164
phenoxyacetic acidApplsci 16 05295 i015
165		Applsci 16 05295 i016
166		Applsci 16 05295 i017
167
Applsci 16 05295 i018
168		Applsci 16 05295 i019
169		Applsci 16 05295 i020
170
Applsci 16 05295 i021
171		Applsci 16 05295 i022
172		Applsci 16 05295 i023
173
Applsci 16 05295 i024
174		Applsci 16 05295 i025
175		Applsci 16 05295 i026
176
Applsci 16 05295 i027
177		Applsci 16 05295 i028
178		Applsci 16 05295 i029
179
Applsci 16 05295 i030
180		Applsci 16 05295 i031
181		Applsci 16 05295 i032
182
Applsci 16 05295 i033
183		Applsci 16 05295 i034
184		Applsci 16 05295 i035
185
Applsci 16 05295 i036
186		Applsci 16 05295 i037
187		Applsci 16 05295 i038
188
Applsci 16 05295 i039
189		Applsci 16 05295 i040
190		Applsci 16 05295 i041
191
Applsci 16 05295 i042
192		Applsci 16 05295 i043
193		Applsci 16 05295 i044
194
Applsci 16 05295 i045
195		Applsci 16 05295 i046
196		Applsci 16 05295 i047
197
Applsci 16 05295 i048
198		Applsci 16 05295 i049
199
Table 3. Characteristics of the investigated compounds: physicochemical (MW, TPSA, HBD, HBA, NRB, α, Ƥ, and log Po/w), biological (log Kw/sp, log Pw/pc, log PHSA, and Caco-2 calculated from molecular structures using ACD/Percepta software version 1994–2012; the log BB values modeled according to Equation (4) in [16]), and chromatographic (log kw,IAM) parameters.
Table 3. Characteristics of the investigated compounds: physicochemical (MW, TPSA, HBD, HBA, NRB, α, Ƥ, and log Po/w), biological (log Kw/sp, log Pw/pc, log PHSA, and Caco-2 calculated from molecular structures using ACD/Percepta software version 1994–2012; the log BB values modeled according to Equation (4) in [16]), and chromatographic (log kw,IAM) parameters.
NoMW
[g mol−1]		TPSA
[Å2]		HBDHBANRBα
[Å3]		Ƥ
[m3 mol−1]		log Po/wlog BBlog Kw/splog Pw/pclog PHSACaco-2
[10−6 cm s−1]		log kw,IAM
1186.2133.9504121.80383.821.6960.004−5.6262.1400.2431900.10
2200.2433.9504123.56414.922.3140.068−5.2772.7330.5152140.41
3216.2443.1805224.11434.081.678−0.067−5.8322.0930.1861930.10
4220.6633.9504123.63412.082.3810.160−5.2072.9170.6442280.64
5255.1033.9504125.45441.522.9180.316−4.9203.5050.9402401.18
6292.3443.1805434.13618.473.4730.268−5.0054.2171.2832410.11
7356.8152.4106538.26697.583.9140.348−4.9724.7291.5242430.06
8340.8143.1805437.71678.414.5890.481−4.4125.3671.8512430.09
9354.8343.1805439.47709.515.2060.558−4.0635.9602.1252411.95
10361.2243.1805437.78676.174.4470.585−4.5185.3431.8862421.86
11375.2543.1805439.54707.265.0640.649−4.1705.9362.1602352.14
12395.6743.1805439.61705.025.1440.740−4.0726.1452.2872282.25
13430.1143.1805441.43733.875.8350.897−3.6256.9412.6801862.91
14430.1143.1805441.43733.875.8440.894−3.6106.9752.6971942.55
15282.2248.2705225.85482.131.6610.230−6.0211.596−0.2021960.94
16296.2548.2705227.61513.222.2730.291−5.6722.1830.0642200.76
17296.2548.2705227.61513.222.2730.295−5.6722.1830.0642201.25
18312.2557.5006328.16532.391.6470.157−6.2451.498−0.2961950.67
19316.6748.2705227.68510.982.1510.381−5.7622.1600.0942200.88
20316.6748.2705227.68510.982.3450.388−5.6032.3730.1992211.66
21316.6748.2705227.68510.982.1310.387−5.7782.1590.0992181.58
22351.1148.2705229.50539.832.8280.544−5.3322.9610.5012352.29
23242.2848.2705227.55497.811.6390.051−6.2981.593−0.0621670.55
24256.3048.2705229.30528.902.2510.117−5.9482.1800.2051931.00
25276.7248.2705229.37526.662.1640.205−6.0382.1570.2342000.80
26276.7248.2705229.37526.662.3130.205−5.8842.3660.3362020.84
27276.7248.2705229.37526.662.1090.209−6.0552.1560.2401961.24
28311.1748.2705231.20555.512.8410.365−5.6082.9580.6412211.85
29256.3048.2705229.30528.902.0890.115−6.0612.0310.1191870.76
30270.3348.2705231.06560.002.7010.179−5.7112.6180.3862171.05
31290.7548.2705231.13557.752.6080.266−5.8002.5890.4072150.65
32290.7548.2705231.13557.752.7570.272−5.6452.7980.5092211.49
33290.7548.2705231.13557.752.5590.272−5.8172.5940.4212081.39
34325.1948.2705232.95586.603.2500.425−5.3703.3900.8142301.67
35286.2974.5707430.27561.260.931−0.161−6.9890.953−0.4581340.48
36300.3174.5707432.02592.351.548−0.088−6.6401.547−0.1831691.93
37320.7374.5707432.09590.111.6030.000−6.5751.726−0.0601791.73
38320.7374.5707432.09590.111.401−0.005−6.7461.517−0.1561701.12
39355.1874.5707433.91618.962.1320.165−6.3002.3180.2452023.36
40286.2974.5707430.27561.260.931−0.159−6.9890.953−0.4581390.81
41300.3174.5707432.02592.351.548−0.093−6.6401.547−0.1831451.33
42316.3183.8008532.57611.520.868−0.230−7.2300.861−0.5381030.81
43330.3483.8008634.40650.131.405−0.163−6.9061.412−0.3091371.42
44320.7374.5707432.09590.111.4010.001−6.7461.517−0.1561421.82
45300.3174.5707532.09599.871.458−0.094−6.7001.459−0.2531571.21
46314.3474.5707533.85630.962.070−0.028−6.3502.0460.0131881.70
47330.3483.8008634.40650.131.444−0.167−6.8931.402−0.3271460.91
48334.7674.5707533.92628.722.1420.065−6.2812.2360.1492002.05
49334.7674.5707533.92628.721.9280.065−6.4272.0630.0681832.11
50369.2074.5707535.74657.572.6250.219−5.9812.8690.4692202.42
51272.26103.3938228.10499.31−2.270−1.040−9.860−2.475−1.7090.3−0.14
52286.31103.3938229.86530.41−1.658−0.976−9.511−1.888−1.4420.50.16
53302.29112.6239330.41549.57−2.304−1.110−10.100−2.573−1.7960.2−0.10
54306.71103.3938229.93528.16−1.601−0.884−9.475−1.750−1.3390.60.41
55341.15103.3938231.75557.01−1.074−0.730−9.170−1.116−1.0140.80.69
56288.31103.3938329.93537.92−1.651−0.965−9.580−2.020−1.52110.40
57302.33103.3938331.69569.02−1.039−0.906−9.231−1.423−1.25530.13
58318.33112.6239432.24588.18−1.679−1.041−9.791−2.071−1.5810.9−0.27
59332.36117.6239534.06626.79−1.177−0.977−9.498−1.561−1.37220.06
60322.75103.3938331.76566.77−1.181−0.814−9.337−1.456−1.21920.32
61280.2861.4106230.82546.921.5900.021−6.5221.7460.1111831.29
62294.3161.4106232.57578.012.2020.081−6.1722.3330.3772110.96
63294.3161.4106232.57578.012.2020.087−6.1722.3330.3772111.69
64308.3361.4106234.33609.112.8540.146−5.8232.9260.6522251.42
65310.3170.6407333.12597.181.576−0.051−6.7151.6890.0371761.11
66310.3170.6407333.12597.111.572−0.050−6.7281.6990.0551661.19
67314.7361.4106232.64575.772.1090.173−6.2612.3040.3992111.27
68314.7361.4106232.64575.772.2690.180−6.1022.5170.5042122.08
69314.7361.4106232.64575.772.0600.180−6.2792.3090.4122082.02
70349.1761.4106234.47604.622.7920.341−5.8323.1110.8142283.23
71349.1761.4106234.47604.622.8120.328−5.8163.1110.8082341.70
72296.3576.5105233.36577.252.3840.298−6.0472.6030.5752151.64
73310.3776.5105235.12608.352.9850.359−5.7023.1860.8382291.54
74310.3776.5105235.12608.352.9850.362−5.7023.1860.8382291.94
75324.4076.5105236.87639.443.6370.418−5.3533.7791.1122381.17
76326.3785.7406335.67627.512.3700.226−6.2402.5460.5012071.50
77330.7976.5105235.19606.102.8980.451−5.7623.2030.8872311.75
78330.7976.5105235.19606.103.0570.455−5.6333.3760.9732322.26
79330.7976.5105235.19606.102.8530.449−5.7743.2070.8962281.53
80365.2476.5105237.01634.953.5850.610−5.3284.0091.2972402.68
81290.3248.2705233.92604.972.4130.274−6.0682.6950.6192261.91
82304.3548.2705235.68636.073.0590.333−5.7493.2410.8652341.53
83304.3548.2705235.68636.073.0590.338−5.7493.2410.8652342.26
84304.3548.2705235.68636.073.0590.338−5.7493.2410.8652342.22
85320.3557.5006336.23655.232.3990.199−6.2922.5970.5252221.47
86320.3557.5006336.23655.232.3790.201−6.3082.5970.5312151.71
87334.3757.5006438.05693.842.8810.267−5.9843.1480.7592272.14
88318.3748.2705237.43667.163.6710.397−5.3993.8291.1322401.85
89324.7648.2705235.75633.822.9320.424−5.8083.2530.9062361.73
90324.7648.2705235.75633.823.0910.431−5.6783.4250.9922372.64
91324.7648.2705235.75633.822.9170.431−5.8253.2580.9202332.57
92359.2148.2705237.57662.673.6080.588−5.3784.0541.3132413.26
93335.3297.1008336.17650.432.311−0.137−6.2652.8860.7122231.92
94349.3497.1008337.92681.532.982−0.070−5.9173.4790.9872322.52
95349.3497.1008337.92681.532.982−0.076−5.9173.4790.9872321.83
96365.34106.3309438.47700.702.303−0.209−6.4602.8350.6462181.78
97363.3797.1008339.67712.623.540−0.015−5.5674.0661.2532381.78
98369.7697.1008337.99679.282.8010.016−6.0063.4501.0082362.16
99369.7697.1008337.99679.282.9950.015−5.8473.6631.1142362.02
100369.7697.1008337.99679.282.7860.019−6.0233.4551.0222302.52
101404.2197.1008339.81708.133.4780.177−5.5764.2511.4152413.22
102304.3548.2705335.75643.582.9440.347−5.7743.1990.8192283.26
103338.7948.2705337.57672.433.4110.491−5.5163.7741.1312392.34
104338.7948.2705337.57672.433.4110.492−5.5163.7741.1312392.48
105338.7948.2705337.57672.433.4110.491−5.5163.7741.1312392.39
106318.3748.2705337.50674.683.5510.399−5.4303.7881.0902382.15
107332.4048.2705339.26705.774.1630.464−5.0804.3751.3562422.51
108332.4048.2705339.26705.774.1630.465−5.0804.3751.3562422.59
109352.8248.2705339.33703.534.0530.555−5.2024.3231.3832422.65
110352.8248.2705339.33703.534.0530.556−5.2024.3231.3832422.78
111352.8248.2705339.33703.534.0530.556−5.2024.3231.3832422.81
112348.4057.5006539.88732.463.4190.328−5.6613.6980.9872362.11
113362.4357.5006541.63763.554.0200.392−5.3164.2811.2502412.43
114382.8457.5006541.70761.313.8750.483−5.4384.2291.2762412.51
115382.8457.5006541.70761.313.8750.485−5.4384.2291.2762412.73
116382.8457.5006541.70761.313.8750.485−5.4384.2291.2762412.72
117352.8248.2705339.33703.534.0530.550−5.2024.3231.3832422.02
118338.7948.2705337.57672.433.4140.492−5.5013.8031.1402392.52
119373.2448.2705339.40701.283.9160.649−5.2734.3381.4332433.22
120373.2448.2705339.40701.283.9160.649−5.2734.3381.4332433.15
121373.2448.2705339.40701.283.9160.649−5.2734.3381.4332433.17
122318.3748.2705437.58682.193.4710.399−5.4553.7451.0442342.07
123332.4048.2705439.33713.294.0890.463−5.1064.3391.3182412.41
124352.8248.2705439.40711.043.9610.548−5.1964.3091.3402421.81
125352.8248.2705439.40711.043.9410.555−5.2124.3091.3452422.68
126387.2648.2705441.22739.894.6380.712−4.7665.1101.7462443.40
127230.2265.0116124.31433.77−1.551−0.408−8.616−1.375−1.4060.50.52
128244.2565.0116126.06464.87−0.934−0.348−8.267−0.781−1.1310.60.33
129244.2565.0116126.06464.87−0.934−0.343−8.267−0.781−1.1310.60.93
130260.2574.2417226.61484.03−1.565−0.479−8.809−1.431−1.4800.50.48
131260.2574.2417226.61483.03−1.614−0.479−8.827−1.426−1.4660.40.52
132264.6765.0116126.13462.62−0.878−0.250−8.202−0.602−1.00811.26
133264.6765.0116126.13462.62−1.082−0.254−8.373−0.811−1.1050.80.79
134299.1165.0116127.95491.47−0.358−0.098−7.927−0.010−0.70321.33
135282.4195.9725530.48647.323.049−0.179−5.6342.8750.6331570.46
136296.4395.9725532.31685.353.465−0.114−5.3973.3130.8141640.88
137296.4395.9725632.32687.403.552−0.112−5.3403.3850.8411921.02
138330.4595.9725536.61741.033.8330.040−5.4303.9401.2971971.37
139316.4295.9725434.77700.953.361−0.020−5.7213.4331.1682011.59
140346.45105.2026537.30759.573.298−0.091−5.9623.3401.0081881.54
141322.4795.9725435.17736.024.2170.006−5.0344.1191.2542091.44
142344.4895.9725638.45781.104.3600.104−5.1114.4871.5222031.65
143350.8795.9725436.69738.093.8750.134−5.4673.9921.4602121.82
144395.3295.9725437.84752.014.0450.335−5.4494.2471.6892162.49
145292.4095.9725531.46641.432.689−0.135−6.0712.5940.6841140.48
146306.4395.9725533.28679.463.139−0.071−5.8333.0320.8651380.76
147306.4395.9725633.29681.513.216−0.070−5.7523.1410.9091610.87
148340.4495.9725535.75736.143.4560.083−5.8403.6901.3571731.22
149326.4295.9725435.75695.062.9900.022−6.1633.1481.2161581.32
150356.44105.2026538.27792.073.500−0.049−6.0533.6091.3581881.29
151332.4695.9725436.15730.133.8410.048−5.4443.8691.3131751.27
152354.4795.9725639.42775.213.9830.147−5.5514.1961.5621861.53
153374.8995.9725439.49732.203.5150.237−5.9043.7111.5121881.56
154405.3195.9725438.81746.123.6790.376−5.8843.9601.7322132.18
155340.4495.9725437.58733.343.6020.086−5.8133.7351.4821881.66
156238.5029.1012221.82421.432.2850.453−5.0952.5130.2802342.06
157252.5229.1012223.73459.082.9370.518−4.7473.1060.5542362.40
158256.4929.1012221.82428.562.3090.543−5.0752.5040.2532343.41
159272.9429.1012223.76457.302.9710.610−4.7103.2540.6942312.72
160272.9429.1012223.76457.302.8100.606−4.8363.0770.5762372.28
161307.3929.1012225.70493.173.7550.760−4.2444.0921.1322193.06
162307.3929.1012225.70493.173.4880.761−4.4343.8470.9932322.67
163317.3929.1012224.87471.932.9800.806−4.8183.3310.8042342.79
164232.4929.1003118.77405.601.3450.417−5.5381.522−0.4922170.87
165235.0635.5303421.31439.982.9170.439−4.4313.1320.3132422.17
166283.5335.5303425.07513.024.0670.657−3.7314.4000.9042452.78
167398.5938.3313534.27667.834.9360.961−3.7415.5201.642813.87
168412.6238.3313536.09705.005.3861.032−3.5045.8581.823504.03
169347.2450.8015634.28707.432.6000.320−5.7612.7650.2002131.77
170292.7644.1204331.59601.413.8070.497−4.5404.2401.0172443.05
171166.1746.5313317.42363.171.853−0.088−5.3991.559−0.17715−0.36
172180.2046.5313419.30402.122.390−0.022−5.0752.1100.051230.16
173194.2346.5313421.14440.572.8050.043−4.8382.5480.232390.49
174180.2046.5313319.34400.812.470−0.022−5.0502.1530.097210.18
175180.2046.5313319.34400.812.470−0.023−5.0502.1530.097210.12
176180.2046.5313319.34400.812.470−0.022−5.0502.1530.097210.13
177182.1755.7614418.16382.190.684−0.222−6.2840.563−0.7355−0.53
178182.1755.7614418.16382.121.387−0.222−5.8121.087−0.4317−0.51
179182.1755.7614418.16382.121.183−0.222−5.9530.919−0.5087−0.51
180236.1455.7614518.34405.811.7420.024−5.3831.432−0.440200.47
181208.2546.5313422.80473.563.3790.106−4.5183.0600.434480.70
182292.4146.5313733.76701.176.5350.486−2.6796.2011.741111.85
183231.0446.5313318.56376.021.9280.203−5.4411.8150.062210.30
184309.9446.5313321.61426.512.8430.554−4.9312.8970.667400.83
185186.5946.5313317.45361.391.7600.004−5.4881.530−0.15617−0.14
186186.5946.5313317.45361.391.9200.005−5.3281.743−0.050200.02
187200.6246.5313319.37399.042.3230.072−5.1562.1230.124320.60
188221.0446.5313319.39397.262.4570.162−5.0422.3310.245410.62
189221.0446.5313319.39397.262.4130.162−5.0532.3350.254430.61
190255.4846.5313321.33433.133.1180.317−4.5923.1710.673661.09
191255.4846.5313321.33433.133.0070.314−4.7043.0220.609630.72
192170.1446.5313315.51332.651.274−0.072−5.6651.038−0.4577−0.55
193278.0446.5313320.63401.132.2870.404−5.3722.2180.39027−0.38
194168.1566.7624316.26340.540.424−0.493−6.768−0.080−0.6681−1.30
195182.1766.7624418.07379.060.582−0.432−6.7580.000−0.7260.7−1.39
196197.1495.3616418.11381.011.121−0.562−5.9421.173−0.3386−0.50
197180.1663.6014418.19370.580.459−0.235−6.5170.343−0.7334−1.08
198194.1863.6014419.49410.711.014−0.169−6.2230.824−0.5385−0.53
199258.2755.7614627.87554.343.0200.125−5.1123.0770.606451.00
Table 4. QSAR models.
Table 4. QSAR models.
Model No.Model
MIlog PHSA = −1.081(±0.142) − 0.017(±0.042) log kw,IAM + 0.101(±0.007) α − 0.339(±0.019)(HBD + HBA) + 0.128(±0.022) NRB
MIIlog PHSA = −1.407(±0.136) − 0.051(±0.038) log kw,IAM + 0.007(±0.000) Ƥ − 0.340(±0.017)(HBD + HBA) − 0.008(±0.023) NRB
MIIIlog PHSA = −1.403(±0.135) − 0.046(±0.03)5 log kw,IAM + 0.007(±0.000) Ƥ − 0.340(±0.017)(HBD + HBA)
MIVlog Kw/sp = −4.719(±0.173) + 0.066(±0.051) log kw,IAM + 0.045(±0.008) α − 0.630(±0.023)(HBD + HBA) + 0.314(±0.027) NRB
MVlog Kw/sp = −4.951(±0.176) + 0.022(±0.049) log kw,IAM + 0.003(±0.000) Ƥ − 0.640(±0.022)(HBD + HBA) + 0.241(±0.030) NRB
MVIlog Pw/pc = 0.367(±0.245) + 0.061(±0.073) log kw,IAM + 0.175(±0.012) α − 0.758(±0.033) (HBD + HBA) + 0.349(±0.039) NRB
MVIIlog Pw/pc = −0.151(±0.242) + 0.019(±0.067) log kw,IAM + 0.012(±0.012) Ƥ − 0.754(±0.001) (HBD + HBA) + 0.120(±0.041) NRB
MVIIICaco-2 = −11.612(±19.247) + 8.325(±5.34490) log kw,IAM + 0.634(±0.054) Ƥ − 21.703(±2.381) (HBD + HBA) − 24.858(±3.283) NRB
Table 5. Statistics of the QSAR models.
Table 5. Statistics of the QSAR models.
Model No.R2R2adjR2predΔPRESSVIF *SSMSEQ2cvPRESScv
MI0.81670.81290.80830.008429.1840<3.0124.3570.14390.816728.5276
MII0.85040.84730.84320.007223.8750<3.4129.9290.11740.850423.3562
MIII0.85030.84800.84460.005723.6661<2.4129.4800.11690.850323.1894
MIV0.86150.85870.85470.006843.5660<3.0258.3400.21410.861542.6058
MV0.87160.86900.86520.006440.4121<3.4261.3690.19800.871639.5166
MVI0.86500.86220.85860.006487.8799<2.7537.4000.43200.865085.6700
MVII0.88350.88110.87770.005875.9593<3.4548.8750.37300.883574.0710
MVIII0.71060.70470.68910.0059492000<3.4112427523610.7106469095
The coefficient of determination (R2, Q2), the determination coefficient adjusted (R2adj), the determination coefficient predicted (R2pred), Δ = R2 − R2pred, the predicted residual error sum of squares (PRESS), the variance inflation factor (VIF), the sum of squared differences from the mean (SS), the mean squared error (MSE), and * the highest value; cv—cross-validated.
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Janicka, M.; Sztanke, M.; Pachuta-Stec, A.; Sztanke, K. Application of Biomimetic IAM Chromatography and QSAR Modeling for Predicting Selected Properties of Potential Drugs and Plant Protection Products. Appl. Sci. 2026, 16, 5295. https://doi.org/10.3390/app16115295

AMA Style

Janicka M, Sztanke M, Pachuta-Stec A, Sztanke K. Application of Biomimetic IAM Chromatography and QSAR Modeling for Predicting Selected Properties of Potential Drugs and Plant Protection Products. Applied Sciences. 2026; 16(11):5295. https://doi.org/10.3390/app16115295

Chicago/Turabian Style

Janicka, Małgorzata, Małgorzata Sztanke, Anna Pachuta-Stec, and Krzysztof Sztanke. 2026. "Application of Biomimetic IAM Chromatography and QSAR Modeling for Predicting Selected Properties of Potential Drugs and Plant Protection Products" Applied Sciences 16, no. 11: 5295. https://doi.org/10.3390/app16115295

APA Style

Janicka, M., Sztanke, M., Pachuta-Stec, A., & Sztanke, K. (2026). Application of Biomimetic IAM Chromatography and QSAR Modeling for Predicting Selected Properties of Potential Drugs and Plant Protection Products. Applied Sciences, 16(11), 5295. https://doi.org/10.3390/app16115295

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop