From Small Molecules to Target Proteins: Identifying Protein Targets with SwissTargetPrediction

SwissTargetPrediction is a web-based tool, available online since 2014, designed to predict the most probable protein targets of small molecules. The predictions are based on the principle of similarity through reverse screening. The 2019 version introduced a major update in terms of underlying data, backend, and web interface. Bioactivity data have been updated, the model retrained, and similarity thresholds redefined. In this new version, predictions are made by searching for 2D and 3D similarities with experimentally validated active molecules within an expanded compound library of 376,342 compounds and a broadened set of 3,068 macromolecular targets (1).

The efficient backend implementation allows results for a drug-like molecule against human proteins to be returned within 15–20 seconds. The redesigned web interface enhances user experience with easy input and new features for improved analysis. Although exploration of broader biological and chemical spaces remains necessary, the tool maintains a high level of predictive performance. For instance, over 70% of compounds have at least one correct human target within the top 15 predictions. The SwissTargetPrediction web server is accessible at http://www.swisstargetprediction.ch/ (1).

In this example application, two molecules—one being Triclosan (the example molecule available on the SwissTargetPrediction web server) and the other a natural product-based molecule, limonene—will be used to predict the potential protein targets associated with these compounds.

Step 1: Access the SwissTargetPrediction web server using the link provided above.

Figure 1. SwissTargetPrediction Web Server Main Page.

Step 2: First, select Homo sapiens under the "Select a species" section. Then, choose the Triclosan molecule from the Examples tab and click the "Predict targets" button.

Figure 2. SwissTargetPrediction Web Server Example Molecule Input.

Step 3: The predicted protein targets are examined by reviewing the results for the Triclosan molecule.

Figure 3. Results Page for the Triclosan Molecule.

In the results, the more filled the green bar under the Probability column, the higher the likelihood that the corresponding protein is a potential target for that molecule. For example, in the case of the Triclosan molecule, it can be concluded that Carbonyl reductase [NADPH] 1 is the most probable target protein.

Step 4: The same steps are repeated for the natural product-based molecule, limonene. For limonene, the SMILES format is obtained from the PubChem entry with ID 22311 and entered into the SwissTargetPrediction web server.

Figure 4. Information for the Limonene Molecule in the PubChem Database.

Figure 5. Input of the Limonene Molecule into the SwissTargetPrediction Web Server

Step 5: The results for the limonene molecule are examined.

Figure 6. Results Page for the Limonene Molecule

When the results for the limonene molecule are examined, it is observed that Peroxisome proliferator-activated receptor alpha (PPARA) and Cannabinoid receptor 2 (CNR2) proteins show similar likelihood scores, indicating that although limonene has a relatively low probability, it could still potentially interact with these two proteins.

In conclusion, the SwissTargetPrediction web tool provides a powerful and accessible in silico approach for predicting the potential protein targets of small molecules. By employing statistical methods based on experimental data, it enables researchers to perform preliminary evaluations in processes such as drug discovery, drug repurposing, and mechanistic analysis. Thus, protein target identification can be achieved on a faster and more rational basis during the pre-laboratory stage.

References

1.Daina, A., Michielin, O., & Zoete, V. (2019). SwissTargetPrediction: updated data and new features for efficient prediction of protein targets of small molecules. Nucleic acids research, 47(W1), W357–W364. https://doi.org/10.1093/nar/gkz382

2.Vane, J. R., & Botting, R. M. (2003). The mechanism of action of aspirin. Thrombosis research, 110(5–6), 255–258. https://doi.org/10.1016/s0049-3848(03)00379-7