Purpose

Anti-inflammatory drugs are in the spotlight of pharmaceutical investigations due to the involvement of this condition in different cancers and diseases. Non-steroidal anti-inflammatory drugs (NSAIDs) are considered to be the most widely used anti-inflammatory drugs. However, the inability of these molecules to distinguish between COX-1 and COX-2 results in concomitant side effects. In the present study, a new algorithmic procedure is applied to build a multi-linear regression model capable of circumventing this problem.

Methods

In this regard, the structures of FDA-approved NSAIDs and COX-1 and COX-2 molecules were prepared. The top 10 COX-2 specific molecules were selected based on their interaction energies and exploited for similarity searches. The resulting 2000 molecules were subjected to various screening processes. Several dependable bioactivities of these compounds along with partial coefficients of all possible affecting descriptors such as molecular weight, H-acceptor/donor, and polar surface area were calculated. The best multi-linear regression approach was used to analyze the descriptors with the highest impact.

Results

Ultimately, a highly reliable model was designed based on 17 screened molecules with higher than 90% anti-inflammatory activity. The attained model was endowed with higher than 84% accuracy according to 5 descriptors, including Log P, Log D, molar refractivity, polarity number, and aromaticity ratio. Our results demonstrated that the bipolarity of drugs is more important than the number of hydrogen bonds to achieve the better anti-inflammatory activity. Moreover, in contrast to prior reports, it is assumable that some Lipinski elements could play a less critical role in drug discovery and improvement efforts.

Conclusions

The quantitative structure and activity relationship (QSAR) model formulated in this study demonstrates an accurate prediction of anti-inflammatory activity of NSAID-like structures. Newly suggested structures are highly resembled to third-generation NSAIDs. This property endorses the trustworthiness and reliability of the obtained equation to design a new generation of NSAIDs.

中文翻译：

---

拟议的用于识别环氧合酶2选择性活性药物成分的多元线性回归模型

目的

由于这种疾病与不同的癌症和疾病有关，因此抗炎药成为药物研究的焦点。非甾体抗炎药（NSAIDs）被认为是使用最广泛的抗炎药。但是，这些分子无法区分COX-1和COX-2会导致伴随的副作用。在本研究中，一种新的算法程序被应用来建立能够避免该问题的多线性回归模型。

方法

在这方面，制备了FDA批准的NSAID和COX-1和COX-2分子的结构。根据其相互作用能选择排名前10的COX-2特定分子，并将其用于相似性搜索。对所得的2000个分子进行各种筛选过程。计算了这些化合物的几种可靠的生物活性，以及​​所有可能的影响描述子的部分系数，例如分子量，H受体/供体和极性表面积。最佳多线性回归方法用于分析影响最大的描述符。

结果

最终，基于17种具有90％以上抗炎活性的筛选分子设计出了高度可靠的模型。根据Log P，Log D，摩尔折射率，极性数和芳族比等5个描述词，获得的模型具有高于84％的准确度。我们的结果表明，药物的双极性比氢键的数量更重要，以实现更好的抗炎活性。此外，与先前的报道相比，可以认为某些Lipinski元素在药物发现和改进工作中的作用可能不太重要。

结论

这项研究中建立的定量结构和活性关系（QSAR）模型证明了NSAID样结构的抗炎活性的准确预测。最新提出的结构与第三代NSAID非常相似。此属性支持所获得方程式的可信度和可靠性，以设计新一代的NSAID。