ProTox 3.0: Next Generation Toxicity Prediction for Chemical Compounds

In silico toxicity models aim to support existing in vitro toxicity methods for predicting the toxic effects of chemicals. In this way, the need for animal testing, time loss and costs are minimized. These models have a comprehensive knowledge base consisting of toxicology, biostatistics, systems biology, computer science and many other disciplines (1). The toxicity level of a chemical can be evaluated by considering factors such as mutagenicity and carcinogenicity. In addition, it can be measured by quantitative methods such as LD50 (dose that causes death of 50% of the tested population) or qualitative methods such as active/inactive, based on analyses for specific cell types or specific indication areas such as cytotoxicity, immunotoxicity and hepatotoxicity (2). As a significant development in the field of in silico toxicity, the rodent acute toxicity prediction platform was developed to be offered to a wide range of users, especially experimental researchers and computational toxicologists, and ProTox: Web Server for rodent oral toxicity prediction was released in 2014 (3).

ProTox 3.0 web server offers several advantages over existing computational models. ProTox web server includes components such as chemical, molecular target information, metabolism, and Adverse Consequence Pathways (AOPs). One of the innovations of the ProTox web server is the classification of the prediction scheme according to different toxicity levels. These levels include oral toxicity, organ toxicity (hepatotoxicity, neurotoxicity, respiratory toxicity, cardiotoxicity, and nephrotoxicity), toxicological endpoints (mutagenicity, carcinogenicity, cytotoxicity, immunotoxicity, BBB permeability, ecotoxicity, clinical toxicity, and nutritional toxicity), 12 toxicological pathways (AOPs), 15 toxicity targets, 14 molecular initiating event (MIE) targets, and 6 metabolism-related molecular targets. This provides insight into the possible molecular mechanisms that cause toxic responses. The latest version, ProTox 3.0, has been developed for the prediction of various toxicity endpoints, including molecular similarity, pharmacophore-based approaches, fragment propensities, and machine learning models. ProTox 3.0, which includes 61 models, is a freely accessible computational toxicity prediction web server that allows the prediction of the most comprehensive toxicological endpoints (4). ProTox 3.0 web server It can be accessed via the website https://tox.charite.de/protox3/ .

In our example application, toxicity prediction will be made using the molecules albuterol and aflatoxin B1. Albuterol (or salbutamol, as it is known outside the United States) is the most widely used β2-agonist for asthma and has an efficacy, safety, and selectivity profile that makes it a preferred rescue treatment for acute asthma exacerbations and symptoms (5). Aflatoxin B1 (AFB1) is among the most common and harmful toxins. Its carcinogenicity and immunosuppressive capacity have been widely reported in various animal species, including chickens, trout, cattle, and mice (6).

Step 1: Go to the home page of ProTox 3.0 web server from the site mentioned above and click on the 'Tox Prediction' tab at the top of the page.

Figure 1. ProTox 3.0 web server homepage.

As input for the analysis; the PubChem names of the molecules, the Canonical SMILES format or the 2D structure of the molecule can be drawn and the molecule structure can be introduced to the server. In the application, we will process the Canonical SMILES formats for both molecules.

Step 2: PubChem site can be used to obtain the Canonical SMILES formats of the molecules. PubChem site can be accessed from https://pubchem.ncbi.nlm.nih.gov/ . First, “albuterol” is written in the search field and the molecule with the PubChem CID number 2083 is selected.

Figure 2. Information of the albuterol molecule in PubChem database.

In the Canonical SMILES section under the Names and Identifiers section on the page, the format CN1C2CCC1CC(C2)OC(=O)C(CO)C3=CC=CC=C3 is copied.

Figure 3. Canonical SMILES format of the albuterol molecule.

Step 3: The Canonical SMILES format of the copied albuterol molecule is pasted into the input field on ProTox 3.0 web server and the 'smiles' button is clicked. Then, the toxicity parameters and pathways to be estimated are selected in the lower tab. The 'All' option can be used to examine all parameters. Since we will estimate all parameters for each molecule in this application, the 'All' option is clicked.

Figure 4. Results screen for the Albuterol molecule.

Step 4: Then, click on the Start Tox-Prediction button and wait for the results. (The analysis takes approximately 1 minute)

Figure 5. Estimation results of LD50, toxicity class and molecular properties of the albuterol molecule.

The upper part of the results page includes parameters such as the 2D representation of the molecule, the estimated dose of the LD50 value, the estimated toxicity class, molecular weight, the number of hydrogen bond acceptors and donors, the number of atoms, the number of bonds, the number of rotatable bonds, the molecular refractivity value, the topological polar surface area value and the LogP value. The estimated LD50 value for the albuterol molecule is 660 mg/kg and the toxicity class is 4. The damaging potential decreases as the toxicity class value goes from 1 to 6.

Figure 7. Estimation results of toxicity parameters of the albuterol molecule.

At the bottom of the results page, toxicity estimates are given. In the results, if the values ​​are “Inactive”, it does not cause toxicity, and “Active”, it does. In addition, the closer the Probability value is to 1, the closer the estimate is to the correct one. In this result for the albuterol molecule, it is seen that it is inactive for most toxicity parameters, which is what we want. And the estimates are generally close to 1.

Step 5: The same step is done for the aflatoxin B1 molecule. The molecule with the PubChem CID number 186907 is used for aflatoxin B1. When the Canonical SMILES format is obtained, the same steps are performed and the analyses are performed, the following results are obtained.

Figure 8. Estimation results of LD50, toxicity class and molecular properties of the Aflatoxin B1 molecule.

The estimated LD50 value of the aflatoxin B1 molecule is 3 mg/kg and its toxicity class is 1. This value shows that aflatoxin B1 belongs to a very dangerous and toxic molecule class.

Figure 9. Estimation results of toxicity parameters of the Aflatoxin B1 molecule.

When the toxicity parameters of the Aflatoxin B1 molecule are estimated; it is seen that it is nephrotoxic, toxic for respiratory tract, carcinogenic, immunotoxic and mutagenic. In addition, its properties such as not being able to pass through the blood-brain barrier (BBB) ​​and nutritional toxicity have been actively determined.

ProTox 3.0 web server offers a powerful tool for evaluating the possible toxic effects of chemical compounds with its comprehensive toxicity prediction models and machine learning-based approaches. Thanks to its advanced algorithms including molecular similarity, pharmacophore-based analyses and biological pathways associated with toxicity, it provides researchers with more reliable and detailed predictions. The ability to analyze from a wide perspective, from organ and system toxicities to metabolic pathways, offers a critical advantage for drug safety and environmental toxicity assessments. With its user-friendly interface and free access, ProTox 3.0 continues to be an important resource in drug discovery, chemical safety assessment and toxicology research.

References

1. Zhang, L., McHale, C. M., Greene, N., Snyder, R. D., Rich, I. N., Aardema, M. J., Roy, S., Pfuhler, S., & Venkatactahalam, S. (2014). Emerging approaches in predictive toxicology. Environmental and molecular mutagenesis, 55(9), 679–688. https://doi.org/10.1002/em.21885

2. Raies, A. B., & Bajic, V. B. (2016). In silico toxicology: computational methods for the prediction of chemical toxicity. Wiley interdisciplinary reviews. Computational molecular science, 6(2), 147–172. https://doi.org/10.1002/wcms.1240

3. Drwal, M. N., Banerjee, P., Dunkel, M., Wettig, M. R., & Preissner, R. (2014). ProTox: a web server for the in silico prediction of rodent oral toxicity. Nucleic acids research, 42(Web Server issue), W53–W58. https://doi.org/10.1093/nar/gku401

4. Banerjee, P., Kemmler, E., Dunkel, M., & Preissner, R. (2024). ProTox 3.0: a webserver for the prediction of toxicity of chemicals. Nucleic acids research, 52(W1), W513–W520. https://doi.org/10.1093/nar/gkae303 

5. Ameredes, B. T., & Calhoun, W. J. (2009). Levalbuterol versus albuterol. Current allergy and asthma reports, 9(5), 401–409. https://doi.org/10.1007/s11882-009-0058-6

6. Marchese, S., Polo, A., Ariano, A., Velotto, S., Costantini, S., & Severino, L. (2018). Aflatoxin B1 and M1: Biological Properties and Their Involvement in Cancer Development. Toxins, 10(6), 214. https://doi.org/10.3390/toxins10060214