Drug discovery commences with careful selection of a target related to diseases. One of the golden standard schemes is library screening against the target of interest to find hit compound, elaboration to produce a lead compound, and lead optimization to give birth to the candidate. During the drug discovery and development process, off-target effects need to be taken into consideration because of the potential adverse effects at the later stages.

Target fishing, or reverse fishing, is a in silico strategy for the identification of the most likely targets of the molecule (ligand) of interest.¹⁾ In silico target fishing consists of target-to-ligand matching technologies and is applicable, by potential interaction prediction, for a variety of purposes:

• Target prediction
• Mechanism of action insights
• Off-target effects prediction²⁾
• Evaluation of polypharmacology³⁾
• Drug repositioning⁴⁾

In silico target fishing is classified into ligand-based and receptor-based approaches, which are further categorized like below.

• Ligand-based target fishing
  • Similarity searching
    • 2D
    • Mixed 2D and 3D
  • Machine Learning
• Receptor-based target fishing
  • Reverse docking (Inverse docking)
  • Pharmacophore-based approach
  • Machine Learning

Ligand-based target fishing is an approach to compare the structural information of the molecule of interest against the database like DrugBank, TTD, ChEMBL and PubChem. Ligand-based approach does not require the structural information of the target biomolecule.

Similarity searching is the assessment of the correlation between structural features and biological activities. 2D approach is based exclusively on the comparison of chemical substructures against compounds with known biological properties. 3D approach takes the spaces occupied by the similar portions of molecules as the representative comparison parameter. Electrostatic properties and pharmacophores can also be incorporated for evaluation.^5),6) 3D ligand-based strategy is feasible but the lack of available 3D structural data hampers its application in a broad range.

Machine learning (ML) is applicable for target fishing. Classification models have been widely used for target fishing and QSAR (quantitative structure-activity relationships), PCM (proteochemometrics) and model stacking are the major strategies in this field.^7),8)

Receptor-based target fishing makes full use of protein structural information for target-ligand matching. Receptor-based approach requires structural data of potential targets but does not need structural information of ligand molecules.

Reverse docking (Inverse docking) is the inversed way of virtual screening. It is basically a docking of a molecule of interest to a database of target structures to predict energetically favorable interaction with screening binding conformation of the ligand.⁹⁾

The tiresome step of reverse docking is data process like recognition of the target binding sites, removal of unnecessary water molecule, to construct a requisite database. The efficient use of automated process and tools like fpocket¹⁰⁾ and SiteMap¹¹⁾ is necessary for large sized screening. While most of the available reverse docking service dock rigid protein and flexible ligand, docking on semi-flexible protein is also possible.¹²⁾

Pharmacophore-based approach has similarity to what is often used for in structure-based drug design (SBDD). This approach utilizes defined pharmacophores of the molecule of interest to identify potential target biomolecules. Pharmacophoric patterns derive during SBDD for creating a lead or the candidate has high reliability and are suitable for target fishing.

Machine Learning is also applicable for receptor-based target fishing.¹³⁾ Interaction prediction of a molecule and targets is possible through ML in a 3D manner. Its application is still limited and it is a developing field of this approach.

In silico target fishing is widely available technology and web server-based systems are readily usable.¹⁴⁾ It is worth running both ligand- and receptor-based target fishing because the available datasets are different. Benchmark study has been done for famous services and it is relatively easy to select the right program to apply for each case.

In silico target fishing is profitable in order for developing a reliable candidate and making rational use of a present drug for repositioning. In drug discovery, adverse effects need to be avoided in an early stage so that the candidate can readily go through the clinical trial.

We are seriously thinking of the candidates’ properties. Our in silico capability is expanding and we would like to accommodate novel technologies to validate the candidate with its aids.

1. https://doi.org/10.3390/molecules26175124
2. https://doi.org/10.1038/nature11159
3. https://doi.org/10.1021/jm5006463
4. https://doi.org/10.1038/nrd1468
5. https://doi.org/10.1007/s10822-010-9374-0
6. https://doi.org/10.1021/acschembio.6b00253
7. https://doi.org/10.4018/jdwm.2007070101
8. https://doi.org/10.1021/acs.jcim.8b00553
9. https://doi.org/10.1080/17460441.2016.1190706
10. https://doi.org/10.1093/nar/gkq383
11. https://doi.org/10.1111/j.1747-0285.2007.00483.x
12. https://doi.org/10.4137/GEI.S29821
13. https://doi.org/10.1186/s13321-023-00689-w
14. https://doi.org/10.3390/ijms20051023