3.1.1 Two traditional processes for discovering small-molecule candidates
The current research strategies for developing BSMs as drugs are quite limited. Figure 2 summarizes two traditional processes for developing small-molecule drugs in recent decades. One of them starts from the recognized BSMs, and then carries out experiments in different disease models and at different levels to explore the complex and variable regulatory networks of small molecules in order to find potential drug targets [86]. Usually, the BSMs in this research process are derived from natural products and have exact biological effects, which would reduce the research cycle and cost. However, it may also bring some negative effects. For example, the composition of natural products is complex, and it is difficult to directly determine which component plays a decisive role. At the same time, BSMs derived from natural products are often highly toxic and low bio-active with poor selectivity to targets. In addition, they might have poor stability and poor organic solubility during the preparation process. To solve this problem, various approaches combined with bioinformatics have been achieved with the development of computer-aided drug design (CADD) in recent years [10, 87, 88]. SwissTargetPrediction, a web server for target prediction of BSMs, was a good example [89]. It made full use of the natural concept that ‘similar bioactive molecules are more likely to share similar targets’ [90] to build a virtual platform, where the BSMs submitted would be compared with the BSMs existed in the sets to explore a potential target through molecular fingerprints and structural similarity measures with known ligands. And this method is widely used to choose the most sensitive BSMs among the homologous small molecules and it is more suitable for the optimization process in the later stages of drug development.
The other way to discover new drugs is based on the existing targets like kinases. With the development of high-throughput technology in recent years, this strategy has been combined with computers to virtually screen potential BSMs for the drug-target interactions (DTIs). And then the chosen BSMs will be subjected to subsequent pharmacological tests after being synthesized. This method is called structure-based drug design (SBDD). It demonstrates strong advantages on saving the time-consuming and cost of traditional biological experiments. However, the inevitable challenge it should confront is that the large amount of the high-dimensional data and background noise is always prepared to interfere the subsequent data analysis. In this field, different algorithms have been developed to solve this problem, such as molecular docking [91], structure-based virtual screening (SBVS) [92], and molecular dynamics (MD) [93] . AutoDock, a suite of integrated and interactive software was programmed to predict suitable small molecules binding with the known structure of the receptor based on the method of molecular docking, which saves both the time-consuming and high cost for traditional chemical library screening [94].