Pyridopyrazine derivatives as highly selective histamine H4 receptor antagonist for the treatment of atopic dermatitis: QSAR modeling and molecular docking studies

1.   Introduction

Atopic dermatitis (AD) is a common, long-term allergic skin condition affecting individuals of all ages. Characterized by pruritic and inflamed skin, AD develops through complex interactions between genetic predispositions, immune dysregulation, and environmental factors. The primary causes include abnormalities in skin cell development and an overactive immune response, which activate specific cytokine pathways [1], such as TH22, TH17/IL-23, and TH1, varying across AD subtypes [2]–[4]. Recent advancements in understanding AD's pathogenesis have shifted the focus from identifying causes to developing effective treatments [3]. Biological therapies targeting specific immune pathways have shown promise by addressing mechanisms such as skin barrier defects, skin dysbiosis [5], and abnormal immune responses [6],[7]. Due to AD's complex nature and its potential impact on overall health [8], accurate diagnosis and targeted therapeutic interventions are essential [9].

Pyridopyrazine derivatives display a diverse range of pharmacological activities. Studies have shown these compounds exhibit anti-inflammatory effects by modulating specific pathways [10], although the mechanisms are complex and can involve both anti-inflammatory and pro-inflammatory processes depending on the context [11]–[13]. Additionally, their antioxidant properties protect cells from oxidative stress, showcasing their therapeutic potential. Pyridopyrazines also have antimicrobial activity, neurological implications through neurotransmitter system modulation, and even anti-cancer effects, suggesting applications in oncology [14]–[16]. However, the exact pathways through which they influence cancer cells remain under investigation. With cardiovascular effects noted in several studies, pyridopyrazine compounds represent a pharmacologically versatile class with broad therapeutic implications [17]–[21].

This study investigates pyridopyrazine derivatives as potential antagonists of the histamine H4 receptor (H4R) for AD treatment. We applied quantitative structure–activity relationship (QSAR) analysis and molecular docking to assess their suitability [22]–[24]. In QSAR analysis, IC50 values measure the inhibitory activity of each compound, with values converted to pIC50 to facilitate correlation with molecular descriptors. Through statistical models—principal component regression (PCR), multiple linear regression (MLR), and partial least squares (PLS)—we analyzed descriptors to understand how structural factors influence activity. QSAR aims to streamline structural data analysis and pinpoint molecular features critical to activity, though accurately predicting the behavior of new compounds remains a significant challenge [25]–[28].

Molecular docking complements QSAR by modeling compound–receptor interactions, specifically examining binding modes and affinities with H4R at the active site of protein 7YFD [29]–[31]. Through this dual approach, combining QSAR's predictive capabilities with docking's structural insights, we aim to identify derivatives with enhanced efficacy as potential AD treatments [32],[33]. Compound C11 demonstrated a particularly high binding affinity to H4R, highlighting it as a promising candidate for anti-inflammatory drug development and advancing targeted AD therapies [34].

2.   Materials and methods

2.1.   Experimental data

Ko et al. [35] designed a series of pyridopyrazine compounds to illustrate their potential as a treatment for atopic dermatitis (H4R) (Figure 1). The compounds were also evaluated for their effectiveness as a histamine H4 receptor antagonist.

2.2.   Molecular descriptors

Achieving a statistically robust model depends on the efficacy of the descriptors, which are based on a systematic and mathematical methodology [36] to describe the variation of activity with respect to the molecular structure. The kind of chemical representation and the technique designed to compute and predict the correlation within the 33 chemical derivatives and their inhibiting activity usually determine the information contained in the descriptors.

In order to establish a correlation between the inhibitory activity and the chemical structures of the investigated compounds, we utilized Python to compute additional molecular descriptors, the detailed results of which are presented in Table S1. In order to provide an accurate representation of the QSAR models, statistical techniques including PCR, MLR, and PLS were applied in the development of the QSAR relationship.

Figure 1.  2D molecular representations of the investigated pyridopyrazine derivatives.

DownLoad: Full-Size Img PowerPoint

2.3.   Statistical analysis

In the pursuit of constructing a QSAR model, a subset of 33 chemicals renowned for their potent inhibitory activity, as reported in previous studies, was chosen. The XLSTAT program was employed to randomly partition the entire set into a test set (7 compounds) for model validation and a training set (26 compounds) for model construction. The proposed methodology incorporates PCA to assess redundancy and collinearity among the studied descriptors [37]–[40]. To find a relationship between activity and molecular structure, three statistical models (PLS, PCR, and MLR) were compared statistically. Once the computations were completed using the 2014 version of the XLSTAT program, the selected descriptors were used to create statistical models that relate the activity of the chemicals to their chemical structures. The dataset was split into two subsets: a test set that was used to evaluate the performance of the derived models and a training set that was used to establish the three models.

2.4.   External validation

The core step in a QSAR investigation is model computation; however, this alone cannot guarantee the model's validity. An external validation procedure is needed to ascertain the model's capacity to predict new compounds. Confirming that the model was not created at random is also crucial. For this test, we used the QSAR models that were created to estimate the activities of the chemicals in the test set. Although it was not involved in the creation of the QSAR models, the latter incorporates compounds from the series of molecules examined in this study. The external ability of the QSAR models to forecast the activity of the test set molecules was assessed by computing the R² coefficient between the predicted and observed pIC50 values after the test set was included. Globarikh and Tropsha [41] stated that determining the result of the R² test is useful for external validation of QSAR models. This explanation states that when the R² test result is higher than 0.5, the model is statistically adequate for prediction and may be applied to additional external data [42].

2.5.   Molecular docking methodology

The scoring results from the molecular docking procedure were easier to calculate and publish due to the modeling software. ChemDraw (18.2) was used to create molecular structures. The Protein Data Bank provided pertinent crystal structures for atopic dermatitis, including the Cryo-EM structure of the histamine H4 receptor and the Gq complex that is attached to it (PDB code: 7YFD) (https://www.rcsb.org/structure/7YFD). Water-binding ligands and cofactors were eliminated from the protein structure in order to improve the structure, and hydrogen atoms were subsequently introduced to stabilize it. Active sites were selectively excluded to create fictitious atoms. The MMFF94x force field was then used to assign the parameters and charges. The docking module made it easier to connect molecular models to the protein surface, and the alpha site search module was used to create alpha site spheres. Dock scoring within the program used the London dG scoring function, followed by refinement using two independent techniques. Self-turning docks were applied to the dock poses, which were selected based on high scores. Using the database browser, the docking postures were aligned with the co-crystallized structure's ligand, and the docking pose's root mean square deviation (RMSD) was calculated.

To assess the binding affinity, binding free energy, and formation of hydrogen bonds, calculations were performed for the synthesized molecules and the amino acid residues of the receptor. The default-docking model identified several forms of interactions, similar to the receptor structure's (native) ligand's RMSD [43].

3.   Results and discussion

3.1.   Principal component analysis

PCA, a qualitative statistical technique, is utilized to reduce dimensionality and de-correlate extensive datasets. The resulting variables, referred to as principal components, enable the reduction of variables and reduce information redundancy. In this study, PCA was utilized on the 13 descriptors that characterize the molecular composition of 33 compounds. The resulting 13 principal components are shown in Figure 2.

Figure 2.  Principal components with their variations.

DownLoad: Full-Size Img PowerPoint

Table 1 summarizes each descriptor's contribution to the principal components F1, F2, and F3. Consequently, the descriptors that contribute most to F1 are weinerPol, vsurf_CW4, vsurf_HL1, and vsurf_HL2. On the other hand, descriptors like vsurf_CP, VAdjEq, zagreb, and weinerPol contribute the most to F2. The three descriptors that contribute most to F3 are rsynth, ASA_P, and ast_violation.

When the variables are projected onto the plane defined by the first three principal components (F1, F2, and F3) and their percentage contributions are analyzed in the two correlation circles depicted in Figure 3, these axes collectively explain 73.96% of the variance. This percentage is considered sufficient to capture the information conveyed by the dataset.

Table 1.  Descriptors' contributions to the primary three principal components F1, F2, and F3.

	F1		F2		F3
	correlation	contribution	correlation	contribution	correlation	contribution
PEOE_VSA_FPNEG	0.692	8.298	−0.037	0.052	0.270	5.836
weinerPol	−0.780	10.556	0.537	11.074	0.127	1.294
zagreb	−0.643	7.176	0.671	17.329	0.100	0.811
VAdjEq	0.740	9.485	−0.595	13.631	−0.191	2.928
vsurf_CP	−0.633	6.936	−0.585	13.156	−0.075	0.447
vsurf_CW4	0.812	11.429	0.471	8.537	0.177	2.508
vsurf_HL1	0.856	12.708	0.477	8.746	0.041	0.133
vsurf_HL2	0.773	10.349	0.452	7.846	0.042	0.143
rsynth	−0.651	7.346	−0.178	1.223	0.498	19.958
ASA_P	0.060	0.063	−0.337	4.359	0.684	37.583
ast_violation	−0.604	6.315	0.360	4.973	−0.521	21.789
ASA+	−0.642	7.137	−0.145	0.813	0.123	1.212
ASA−	−0.356	2.202	0.464	8.262	0.258	5.358

---

 | Show Table

DownLoad: CSV

Figure 3.  Correlation circles between F1–F2 and F1–F3, the principal compounds.

DownLoad: Full-Size Img PowerPoint

After the PCA process, we identified two descriptors, including vsurf_CP, and ASA_P, for developing QSAR models. These descriptors were selected from a pool of 13 based on their correlation coefficients. Additionally, descriptors with the lowest correlation coefficients, as shown in Table 2 correlation matrix, were included. The database was then randomly divided into training and test sets using the XLSTAT program.

Table 2.  Matrix depicting the correlation among various descriptors acquired.

Variables	PEOE_VSA_FPNEG	weinerPol	zagreb	VAdjEq	vsurf_CP	vsurf_CW4	vsurf_HL1	vsurf_HL2	rsynth	ASA_P	ast_violation	ASA+	ASA−
PEOE_VSA_FPNEG	1
weinerPol	−0.559	1
zagreb	−0.390	0.866	1
VAdjEq	0.517	−0.958	−0.896	1
vsurf_CP	−0.356	0.199	0.071	−0.121	1
vsurf_CW4	0.534	−0.328	−0.191	0.264	−0.766	1
vsurf_HL1	0.529	−0.401	−0.223	0.320	−0.802	0.940	1
vsurf_HL2	0.410	−0.354	−0.189	0.278	−0.663	0.851	0.942	1
rsynth	−0.300	0.504	0.276	−0.474	0.497	−0.476	−0.614	−0.518	1
ASA_P	0.061	−0.127	−0.161	0.094	0.107	0.021	−0.056	−0.039	0.149	1
ast_violation	−0.528	0.579	0.560	−0.544	0.237	−0.403	−0.364	−0.313	0.037	−0.311	1
ASA+	−0.447	0.345	0.343	−0.408	0.359	−0.589	−0.546	−0.462	0.398	0.055	0.204	1
ASA−	0.011	0.428	0.493	−0.451	-0.090	−0.137	−0.163	−0.167	0.221	−0.132	0.261	0.142	1

---

 | Show Table

DownLoad: CSV

3.2.   Partial least squares

The quantitative evaluation of the chemicals to enhance the structure–activity relationship was conducted through partial least squares analysis. The correlation coefficient (R²), MSE, and RMSE were employed for the assessment and validation of the established model.

pIC50 = −11.12 + 0.27*weinerPol + 0.10*Zagreb − 7.91*VAdjEq + 20.01*vsurf_HL1 − 8.74*rsynth − 3.38*ast_violationn.

N = 26; R² = 0.744; MSE = 2.318; RMSE = 1.523.

The significance of the chosen descriptors in the developed model is crucial, and their representation is visually depicted in Figure 4.

Figure 4.  Standardized coefficients in relation to variables within the proposed PLS model.

DownLoad: Full-Size Img PowerPoint

The statistical metrics obtained for PLS affirm the reliability and predictive capability of the developed model, as delineated in Table 3. Furthermore, the symmetrical distribution of residuals around zero, illustrated in Figure 5, indicates the absence of systematic errors in the model. The calculated R² for PLS signifies a close alignment between predicted and observed pIC50 values, validating the robustness of the developed model (Figure 6), Table 4.

3.3.   Principal components regression

The selected descriptors were used in a PCR to enhance the accuracy of forecasting the relationship between activity and molecular structure. The resulting equation, along with the statistical parameters obtained from the ANOVA table, is presented as follows:

pIC50 = 11.97 − 705.61*PEOE_VSA_FPNEG − 0.65*weinerPol − 0.27*Zagreb + 88.48*vsurf_CP − 15.93*vsurf_CW4 − 21.85*vsurf_HL2 − 12.87*rsynth + 4.89*ast_violation.

N = 26; R² = 0.897; R²_adj = 0.826; MSE = 1.033; RMSE = 1.017.

The results of the analysis of variance are concisely presented in Table 3 of the ANOVA.

The statistical outcomes derived from the PCR demonstrate high quality and enhanced predictive capabilities comparable to PLS. This is substantiated by significantly high values of the R² (=0.897) and R²adj (=0.826), coupled with a minimal MSE of 1.033. The expected pIC50 values correlate well with the experimental pIC50, as illustrated in Figure 7.

Table 3.  ANOVA for PCR model.

Source	DF	Sum of squares	Mean squares	F	Pr > F
Model	9	117.657	13.073	12.650	0.0001
Error	13	13.435	1.033
Total	22	131.092

---

 | Show Table

DownLoad: CSV

Figure 5.  Residuals vs. observed pIC50.

DownLoad: Full-Size Img PowerPoint

Figure 6.  Experimental against computed pIC50 obtained by the PLS model.

DownLoad: Full-Size Img PowerPoint

Table 4.  Observed and expected pIC50 values from PLS, PCR, and MLR models.

Compounds	pIC50	PLS model		PCR model		MLR model
		Pred (pIC50)	Residual	Pred (pIC50)	Residual	Pred (pIC50)	Residual
1	3	3.725	−0.725	1.836	1.164	3.084	−0.084
2	3	−0.491	0.568	3.047	−0.047	−21.382	24.382
3	0.077	0.581	−0.514	0.419	−0.342	1.457	−1.380
4	0.067	2.234	−1.724	1.062	−0.995	2.827	−2.760
5	0.034	−2.420	2.700	0.628	−0.594	−0.282	0.316
6	0.51	−0.639	0.879	0.000	3.800	1.446	−0.936
7	0.28	−0.513	0.693	−0.549	0.829	−2.001	2.281
8	0.24	2.656	−2.136	1.364	−1.124	0.744	−0.504
9	0.18	1.863	−1.623	0.000	0.445	1.341	−1.161
10	0.52	−0.425	0.725	−0.522	1.042	1.416	−0.896
11	0.24	−0.457	0.587	0.378	−0.138	1.321	−1.081
12	0.27	8.698	1.302	0.781	−0.511	−0.636	0.906
13	0.3	1.013	−0.633	−0.155	0.455	−0.533	0.833
14	0.13	3.480	−0.580	0.000	−1.055	0.203	−0.073
15	10	1.579	3.221	0.000	3.549	8.270	1.730
16	0.38	5.957	0.343	1.035	−0.655	1.756	−1.376
17	0.76	−0.174	0.239	0.000	−4.264	2.560	−1.800
18	2.9	1.663	−1.233	3.797	−0.897	1.557	1.343
19	10	1.218	−0.728	9.008	0.992	9.588	0.412
20	4.8	2.919	−2.867	3.925	0.875	2.593	2.207
21	4.1	7.113	2.887	0.000	2.597	3.372	0.728
22	6.3	0.967	−0.905	0.000	8.742	2.410	3.890
23	1.2	0.530	−0.477	1.517	−0.317	−0.511	1.711
24	0.065	−40.557	43.557	−0.087	0.152	0.989	−0.924
25	5.2	−1.733	1.767	5.438	−0.238	4.228	0.972
26	0.43	−0.584	0.854	0.427	0.003	−0.059	0.489
27	0.49	−1.420	2.180	0.000	−10.243	0.608	−0.118
28	0.052	−11.404	21.404	−0.612	0.664	4.535	−4.483
29	10	1.796	2.304	0.000	14.530	2.228	7.772
30	0.062	−0.935	2.135	0.000	1.448	1.526	−1.464
31	0.028	2.740	2.460	0.781	−0.753	1.895	−1.867
32	0.053	0.325	−0.297	−1.352	1.405	0.606	−0.553
33	0.027	−6.308	6.335	1.000	−0.973	−0.204	0.231

---

 | Show Table

DownLoad: CSV

Figure 7.  Comparison of experimental and computed pIC50 values acquired through the PCR model.

DownLoad: Full-Size Img PowerPoint

A novel QSAR model was created utilizing the MLR technique in order to improve the correlation between the set of 8 molecular descriptors and the anticipated activities generated through QSAR models created through PLS and PCR techniques.

3.4.   Multiple linear regression

Figure 8.  Modeling characterization by the normalized coefficients.

DownLoad: Full-Size Img PowerPoint

Figure 9.  Correlation between the observed and the predicted activities.

DownLoad: Full-Size Img PowerPoint

Four criteria are considered when applying the MLR method: the Fisher ratio value (F), the root RMSE, the R², and the R²adj. The MLR results, including the correlation between observed and predicted activities and the standardized coefficients of the descriptors, are shown in Figures 8 and 9. Furthermore, the following equation represents the formulated model and the values of the statistical parameters:

pIC50 = 9.33 − 38.48*vsurf_CW4 + 167.69*vsurf_HL1 − 153.98*vsurf_HL2 + 9.40*rsynth − 1.79E − 7*ASA+ + 1.368E − 7*ASA− − 2.48E − 7*ASA_P.

N = 26; R² = 0.816; R²adj = 0.735; RMSE = 1.493; F = 10.124.

When analyzing the coefficient normalization diagram, it becomes clear that the constructed model highlights three outstanding descriptors (vsurf_CW4, vsurf_HL1, and vsurf_HL2) that correlate strongly with pIC50. The high R² (0.816), the reduced MSE (0.100), and the high statistical confidence (F = 10.124) confirm the statistical acceptability of the QSAR model. In addition, the achieved p-value of less than 0.05 (Pr < 0.0001) emphasizes the statistical significance of the QSAR model equation at a confidence level of over 95%. Figure 9 shows a significant correlation between observed and predicted pIC50 values, a consequence of the low MSE value achieved. This emphasizes the clear correlation that exists between the values found through experimentation and the values that the QSAR model predicts.

3.5.   External validation

An external validation is performed to verify the validity of the prediction power of the obtained QSAR models. The purpose of the external validation test is to evaluate the QSAR models' ability to predict the pIC50 activity levels of the molecules in the test set. A critical criterion for assessing how well externally verified models perform in forecasting the activities of molecules unrelated to the model's creation is the R² test coefficient calculation, which is part of this evaluation. The R² test values achieved for the PLS, PCR, and MLR models are 0.636, 0.802, and 0.722 respectively, all above 0.5. These results verify the robust predictive ability of the QSAR models in estimating pIC50 values during external validation.

3.6.   Molecular docking

This study utilized molecular docking to predict interaction types, binding sites, and binding affinities of pyridopyrazine derivatives, particularly compound C11, against the histamine H4 receptor (H4R), which is implicated in inflammatory responses. Molecular docking provided valuable insights into binding affinities, with compound C11 showing the lowest binding energy at −5.80 kcal/mol, indicating a strong interaction with the H4R protein (7YFD). This binding affinity, among the highest in the compound library, supports the potential of C11 as a selective H4R antagonist for anti-inflammatory applications in conditions such as atopic dermatitis.

The results in Table 5 and Figure 10 detail these interactions, showing C11's significant hydrogen bonding with key residues (e.g., Thr 251 and Lys 27) and π-stacking interactions with Thr 219. These bonds are particularly relevant as they contribute to stabilizing the receptor–ligand complex. However, while the binding energy and interaction specifics suggest a strong affinity, it is essential to interpret these findings with caution. Docking studies are predictive, and factors like solubility, bioavailability, and off-target interactions in a physiological context need further validation through experimental assays.

Table 5.  Interaction table between the compounds and 7YFD protein of H4R inflammatory.

compounds	Binding energy (Kcal/mol)	Ligand	receptor	interaction	distance	E (kcal/mol)
1	−5.38	6-ring	THR 219	pi-H	4.4	−0.7
2	−5.21	5-ring	THR 219	pi-H	3.31	−0.7
		6-ring	THR 219	pi-H	4.38	−0.7
3	−5.38	5-ring	THR 219	pi-H	3.29	−0.8
		6-ring	THR 219	pi-H	4.31	−0.8
4	−5.59	N 24	THR 251	H-acceptor	3.36	−0.6
5	−5.61	I 35	ARG 31	H-donor	4.19	−0.5
		N 24	THR 251	H-acceptor	3.27	−0.9
6	−5.42	5-ring	THR 219	pi-H	3.29	−0.8
		6-ring	THR 219	pi-H	4.37	−0.9
7	−5.74	F 38	ARG 35	H-acceptor	3.23	−0.7
8	−5.53	N 36	LYS 27	H-acceptor	2.97	−1.3
		6-ring	THR 219	pi-H	4.62	−1.2
		5-ring	THR 219	pi-H	3.34	−1.3
9	−5.45	5-ring	THR 219	pi-H	3.29	−0.8
		6-ring	THR 219	pi-H	4.3	−0.8
10	−5.76	5-ring	THR 219	pi-H	3.28	−0.8
		6-ring	THR 219	pi-H	4.29	−0.9
11	−5.80	CL 34	THR 251	H-donor	3.2	−0.6
		N 23	LYS 27	H-acceptor	3.39	−0.7
		5-ring	THR 219	pi-H	4.15	−0.8
		6-ring	THR 219	pi-H	4.82	−0.8
12	−5.75	N 36	SER 253	H-acceptor	3.81	−0.8
13	−5.64	6-ring	THR 219	pi-H	3.98	−0.9
14	−5.44	6-ring	THR 219	pi-H	3.98	−0.9
15	−5.55	6-ring	THR 219	pi-H	4.19	−0.6
16	−5.67	N 24	THR 251	H-acceptor	3.33	−0.6
17	−5.56	N 24	ARG 35	H-acceptor	3.26	−4.4
		5-ring	THR 219	pi-H	3.8	−1
18	−5.46	5-ring	THR 219	pi-H	3.3	−1
		6-ring	THR 219	pi-H	4.43	−1
19	−5.59	N 28	ARG 35	H-acceptor	3.09	−5.2
20	−5.6	N 30	ARG 35	H-acceptor	2.97	−5.4
21	−5.28	5-ring	THR 219	pi-H	3.77	−0.6
22	−5.26	6-ring	THR 219	pi-H	4.43	−1.2
		5-ring	THR 219	pi-H	3.29	−1.1
23	−5.59	5-ring	THR 219	pi-H	3.3	−0.6
		6-ring	THR 219	pi-H	4.3	−0.7
24	−5.23	N 27	ARG 35	H-acceptor	3.08	−4.8
		5-ring	THR 219	pi-H	3.77	−1
25	−5.48	N 27	THR 33	H-donor	2.94	−1.2
26	−5.66	6-ring	THR 219	pi-H	4.47	−1
		5-ring	THR 219	pi-H	3.31	−1
27	−5.53	6-ring	THR 219	pi-H	4.27	−0.7
28	−5.46	N 17	THR 219	H-acceptor	2.93	24.3
29	−5.79	N 29	ARG 35	H-acceptor	2.98	−3.9
30	−5.38	N 24	THR 219	H-acceptor	2.9	−2.7
31	−5.3	BR 17	THR 251	H-donor	3.55	−0.6
		N 24	LYS 27	H-acceptor	3.34	−1
		5-ring	THR 219	pi-H	3.4	−0.6
		6-ring	THR 219	pi-H	4.45	−0.6
32	−5.4	N 26	THR 219	H-acceptor	3.02	−0.7
33	−5.42	N 26	ARG 35	H-acceptor	3.21	−1

---

 | Show Table

DownLoad: CSV

Figure 10.  3D docking and 2D of compound C11 and 7YFD protein of H4R inflammatory.

DownLoad: Full-Size Img PowerPoint

In terms of structural interaction, the hydrogen bond distances (3.20–4.82 Å) and stabilizing energies (−0.6 to −0.8 kcal/mol) reveal that these interactions are consistent with moderate affinity, but further in vitro and in vivo testing is needed to confirm the stability and efficacy of C11 in a biological system. Additionally, the alignment of docking data with experimental LD10 values supports the model's reliability, though this correlation should be expanded by testing other promising compounds in the pyridopyrazine series to ensure broader applicability.

Simultaneously, QSAR modeling served as a predictive tool to understand structural determinants of activity. By identifying molecular descriptors with significant predictive power, such as vsurf_CW4 and weinerPol, the QSAR model allowed for efficient virtual screening of potential H4R antagonists, accelerating the identification of compounds with optimal pharmacological properties. However, one limitation observed is that QSAR models rely on known chemical libraries, and their predictions are limited when applied to novel structural frameworks. Future work should aim to refine QSAR models to include diverse compound libraries, allowing for a more robust selection process.

Overall, while molecular docking and QSAR provided strong preliminary data on compound C11 and other pyridopyrazine derivatives, these computational methods should be complemented with experimental validation. The observed binding affinities and structural insights are promising for advancing AD treatments, yet translating these findings into clinical applications will require further studies to confirm efficacy, selectivity, and safety profiles in biological systems.

4.   Conclusion

In this study, we evaluated the inhibitory activity of pyridopyrazine derivatives against the histamine H4 receptor (H4R) as a potential treatment for atopic dermatitis. Through QSAR modeling and molecular docking, compound C11 emerged as a promising candidate, demonstrating strong binding affinity and favorable interactions with H4R. These findings highlight the potential of pyridopyrazine derivatives in anti-inflammatory therapy development for AD, with future studies needed to confirm their therapeutic viability.

Use of AI tools declaration

The authors declare they have not used Artificial Intelligence (AI) tools in the creation of this article.