1 Introduction

Cinnamic acid is one of chemical substances found in plants or fruit. Its derivatives have been identified as interesting compounds with multifunctional activities (Peperidou et al., 2017) such as inhibitor α-glucosidase (Adisakwattana et al., 2009), antimicrobial (Narasimhan et al., 2004), antifungal (Bisogno, 2007), antioxidant (Sova, 2012), and anticancer (J. A. do Vale, , 2022). As an anticancer, cinnamic acid derivatives are effectives to inhibit the growth of cancer in various type of cancers. One of their derivatives, cinnamaldyhde, inhibit and destroy the cancer cell through apoptosis mechanism. The other potential derivative of cinnamic acid is 4-alkoxy cinnamic group.

4-alkoxy cinnamic groups are a derivative of cinnamic acid that have been studied as anticancer through P388 leukemia cell line, which began when p-coumaramide successfully isolated from Kleinhovia Hospita L. and has IC₅₀ value of 44 µg/mL (Firdaus, 2009). Modifications through the structure of 4-alkoxy cinnamic groups show the increasing and decreasing activity based on the type of functional group as the substituents. Despite as anticancer, this analogues were also reported as inhibitors of M. Tuberculosis (Adeniji et al., 2020) and α-glucosidase inhibitor (Ernawati et al., 2020). However, development of anticancer agent was in line with the increasing of inhibitor’s toxicity such cardiovascular toxicity that found as a side effect when threating by small molecule tyrosine kinase inhibitor. Perceived side effect can be hypertension, congestive heart failure, heart disease, etc (Brown et al., 2016). This side effect making a researcher should develop and found the new cancer inhibitors. One way to find a new potential drug molecule is by using computational drug design such as Quantitative Structure Activity Relationship (QSAR). This tool could understand the relationship between the structure compound and their activity (Iman et al., 2014).

QSAR analysis has been a common tool in novel drug design based on ligand approach. This computational tool helps the researchers to minimize the cost and time-consuming during synthesis stage. Qsar analysis of 4-alkoxy cinnamic derivatives has been performed as M. Tuberculosis inhibitors (Adeniji et al., 2020). However, QSAR study of 4-alkoxy cinnamic analogues as anticancer through P388 cell line has never been modelled. This research focused on the finding of the new 4-alokxy cinnamic derivatives as anticancer through QSAR analysis by using 24 compounds that we had synthesized since 2019 until now. In this 3D-QSAR analysis, Genetic Algorithm (GA) (Labjar et al., 2022) parallelized with Multi Linear Regression (MLR) were attempted (Jawarkar, 2022). GA is chosen to develop the QSAR equation since this algorithm has an effective and easy procedure to attempt in QSAR modelling (Serra et al., 2020). This algorithm is combined with MLR due to the transparency, effectiveness and the easy step to conduct this method in a good accuracy (Arthur et al., 2016). This study aims to develop the relationship between 4-alkoxy cinnamic structure and their activity to the P388 cell line, then predict a newly potent analogues to be synthesized in our next research.

2 Materials and method

2.1

2.1 Data collection

Twenty-three molecules of 4-alkoxy cinnamic analogues used in this research as potent anticancer through P388 cell line. All molecules were collected from some of our reported articles since 2012 until 2021 (Firdaus, 2009; Firdaus et al., 2017, Firdaus et al., 2018, Firdaus et al., 2019, Firdaus et al., 2020a, Firdaus et al., 2020b, Firdaus et al., 2021a, Firdaus et al., 2021b, Firdaus et al., 2021c). Bioactivity of all analogues was determined as inhibition concentration 50 % (IC₅₀) then converted as logarithmic scale (pIC₅₀) in order to obtain normal distribution. The structures and bioactivity values of analogues are shown in Table 1.

Table 1 Structure and bioactivity of 4-alkoxy cinnamic analogues as anticancer .

No.	R1	R2	R3	IC50	pIC50
1	–OCH3	–OH	-H	16.15	−1.21
2	–OCH3	–OCH3	-H	21.18	−1.33
3	–NH2	–OH	-H	44.00	−1.64
4	N-diethyl	–OH	-H	23.50	−1.37
5	N-propyl	–OH	-H	53.56	−1.73
6	N-butyl	–OH	–OH	0.609	0.22
7	N-benzyl	–OH	-H	16.15	−1.21
8	N-benzyl	–OH	–OH	674.38	−2.83
9	N-benzyl	–OH	–OCH3	179.56	−2.25
10	piperidinil	–OH	-H	5.34	−0.73
11	piperidinil	–OH	–OCH3	46.67	−1.67
12	piperidinil	–OH	–OH	0.861	0.06
13	pyrolidinil	–OH	-H	53.46	−1.73
14	pyrolidinil	–OH	–OH	11.35	−1.05
15	morpholine	–OH	-H	19.35	−1.29
16	morpholine	–OH	–OH	1.48	−0.17
17		–OH	–OCH3	10.79	−1.03
18		–OH	-H	1.0	0.00
19		–OH	–OCH3	29.14	−1.46
20		-OAc	-OAc	0.5	0.30
21		–OH	-H	5.89	−0.77
22		–OH	–OH	0.91	0.04
23		–OH	-H	16.97	−1.23

3 Result and discussion

This study focused on development of QSAR model between 4-alokoxy cinnamic analogues and their bioactivity as anticancer in P388 cell lines. Kennard-Stone’s algorithm was chosen to divide the 3D descriptor data into training and test set data. To make sure that the resulted division data had a good quality, univariate analysis was attempted. Table 2 present the result of univariate analysis data and it shows that test set value (-1.6400 to 0.2200) was in line with training set value (-2.8300 to 0.3000). Standard deviation of training data is close to test data value. Thereby, Kennard-Stone’s algorithm applied has produced the test set data that reflects training set data.

BuildQSAR software was used to obtain some of qsar models by using training set data. There were 4 qsar models produced and validated by some statistical parameters. Table 3 shows validation parameter resulted in each qsar models that generated from Genetic Algortihm (GA). All qsar models are eligible for the parameter of coefficient determination, but the selected qsar model is model 1 because it has the best value for some statistical parameters specifically in coefficient determination of test set data which about 0.9812, indicating the linear relationship in predicting bioactivity value.

The selected qsar model (Model 1) was conducted to calculate residual of the inhibition data between experimental and predicted data. Table 4 present the residual value of inhibition data, range of the difference value of experiment and predicted data is only about −0.89 until 0.69. This low residual value indicate that Model 1 is a good choice in predicted the inhibition value.

The standardized residual was also evaluated through the experimental value that is shown in Fig. 1. General rule to figure out the standardized value is that it should has the value close to zero. If the residual is less than −2, the observed data is less than the predicted data (Jalali-Heravi and Kyani, 2004). In our result, there is only one point that close to −2. However, the other values show the range between 0 and $\pm$ 1, and the spread of values is close to 0.

Close

Fig. 1

Graphical plot of standardized residual and experimental activity.

Export to PPT

Despite of internal and external validation explained before, Y-randomization test was performed to check the accuracy of Model 1 in prediction of inhibition activity. Table 5 shows the result of Y-randomization test for Model 1. Ten random models were resulted and then compared its R, R², and Q² value to the original model (Model 1). The selected qsar model would be conclude as inaccurate if there is a strong correlation from the random R² value (R² > 0.5). Result of Y-randomization test shows that all of random models have a weak correlation to the bioactivities, thus the Model 1 could be inferred to have a responsive in predicted the bioactivities. Applicability domain is the last validation applied to Model 1. The critical value (h*) of this model is 0.67 and the cut off is $\pm$ 3. The William plot in Fig. 2 presents that there is no outlier in the training and test set data indicating the validity of the model. The QSAR equation of Model 1 resulted from Genetic Algorithm (GA) is.

Close

Fig. 2

The William Plot of the standardized residuals vs the leverage value.

Export to PPT

pIC50 = 2.7350 (±1.0653) $\times$ RDF145u − 2.2276 (±0.9655) $\times$ RDF120m + 1.8369 (±1.1295) $\times$ E2m − 1.5611 (±0.4518).

where there are three descriptors involved in the equation, namely RDF145u, RDF120m, and E2m. RDF descriptor is one of geometrical descriptor that describe the molecular structure in space. The alphabet u in RDF145u means unweighted method, while the notation 145 related to the distance between two atoms by distance about 14.5 Å. RDF120m belongs to the radii which about 12 Å from the center atom and alphabet m means the relative mass used as weighted method (Arief et al., 2021). E2m is referred to one of Weighted Holistic Invariant Molecular (WHIM) descriptor, positive sign for this descriptor means a comprehensive distribution of the mass and polarizability to the second principal axis and it influences the anticancer activity by the high pIC₅₀ (or lower IC₅₀ value).

Comparison of the resulted qsar model presents in Table 6. Based on our literature study, qsar analysis that use 4-alkoxy cinnamic acid derivative as anticancer has never been conducted. Arthur et al. (2016) build a qsar model in the same bioactivity as anticancer in P388 cell line but different dataset (Arthur et al., 2016). The other qsar study used cinnamic acid derivatives but have a different bioactivity such as anti-mycobacterium tuberculosis and acetylcholinesterase inhibitor. This comparison shows the novelty of this research and the great impact for the development anticancer agent in the future. To improve the anticancer activity of 4-alkoxy cinnamic analogues, we have to find functional groups that can improve the number of RDF145u and E2m but lower RDF120m. The list of new designed compounds is presented in Table 7 and it shows that compound Q8 is the most active predicted compound which is derived from –NO₂ substituent in R₃ position. Even after we changed the amine in R₁ position, it still shows that derivation of –NO₂ substituent still give a better predicted activity. This result was in line by the previous research that showed a potential of –NO₂ moiety in cinnamic acid derivatives could increase the bioactivity of the compounds (Ernawati et al., 2020). Evaluation of drug likeness property in Table 7 supports the ability of the newly design compounds to be use as a drug molecule due to the absence of violation in the Lipinski rule. Retrosynthetic analysis which is shown in Fig. 3 explain that compound Q8 is possible to synthesize by using p-methoxy cinnamic acid as the starting material. The synthesis pathway consists of three steps of reaction, namely chlorination, amidation (Castral, 2011) and nitration (Giles, 2004). However, further research still requires to make sure that these design compounds are safe in patient such as in vitro and in vivo test.

Table 7 List of newly designed compounds obtained from QSAR analysis and evaluation of drug likeness property .

Compound code	R1	R2	R3	IC50 (prediction)	Mr (g/mol)	Log P	Acceptor H-bond	Donor H-bond	Violation
Q1	–NH2	–OH	-CF3	7.9544	231.17	1.88	5	2	0
Q2	N-dietyl	–OH	-CF3	14.7049	259.22	2.43	5	2	0
Q3	N-propyl	–OH	-CF3	24.1760	273.25	2.69	5	2	0
Q4	N-propyl	–OCH3	-CF3	0.0209	287.28	2.94	5	1	0
Q5	N-butyl	–OCH3	-CF3	12.5831	301.30	3.19	5	1	0
Q6	N-benzyl	–OCH3	-CF3	0.5172	335.32	3.65	5	1	0
Q7	N-propyl	–OCH3	–NO2	4.9926	264.28	0.92	4	1	0
Q8	N-butyl	–OCH3	–NO2	0.0019	278.30	1.19	4	1	0
Q9	–NH2	–OCH3	–NO2	6.4998	222.20	0.08	4	1	0
Q10	Morfolil	–OCH3	-H	23.7601	247.29	1.05	3	0	0

Close

Fig. 3

Retrosynthesis analysis of compound Q8.

Export to PPT

4 Conclusion

A QSAR model between 3D descriptors and pIC50 data of 4-alkoxy cinnamic analogues has been developed using Genetic Algorithm (GA) and employing Multi Linear Regression (MLR) method. The contributed descriptors are RDF145u, RDF120m, and E2m. The model has satisfied some of statistical parameters in the internal and external validation. The selected model (Model 1) is applied to design new analogues and resulted that compound Q8 as the most active compound with IC₅₀ prediction value about 0.0019 (µg/mL) and has no violation in the drug likeness property based on the Lipinski rule. This result contributes on the drug design field specially on the finding the new potential anticancer compound. The newly design compounds can be synthesized and tested through in vivo and in vitro.