A: A Quantitative Structure-Activity Relationship (QSAR) model is a predictive model that relates the structure of a chemical (encoded in a series of numerical descriptors) to one of its properties. QSAR models use pre-existing experimental data, analysed through machine-learning algorithms, to establish this relationship. These models are statistically validated with separate training and validation datasets to ensure predictability.

A: There is extensive literature on the technique. More information can be found on our website. Additionally, in-house courses on developing and applying QSAR and other computational techniques can be found at ChemoPredictionAcademy.

A: Yes. Most regulatory frameworks not only accept QSAR models but also encourage their use as an alternative to animal-based tests.

A: No, except for ProtoNANO, the models are developed to work with organic discrete chemicals. Thus, any chemical containing metals or other elements not common in organic molecules is not accepted. Additionally, mixtures and salts are not directly usable in ProtoPRED^®. In such cases, the user can predict the properties separately for individual ions or components.

A: Some QSAR models are available across multiple modules, as they fall under different classifications. For example, boiling point is a REACH requirement but also a physico-chemical property; therefore, the corresponding model is included in both the ProtoREACH and ProtoPHYSCHEM modules. The QSAR model is the same, and the results should be the same, but some information included in the PDF reports may be slightly different, as those are tailored for the particular application and/or include combined reports which consider additional information.

A: ProtoPRED^® is a professional platform with commercial models. However, registration is completely free of charge. Once registered, users receive an initial number of free tokens, which allows access to all platform features.

A: ProtoPRED^® tokens are units used by the server to manage the credit. Each time a prediction is made or a PDF report is generated, a specific number of tokens is deducted from the user's account.

A: Once the free tokens are used up, additional token packages can be purchased directly from the online store. Simply visit our website and select the package that best suits your needs. For academic use or higher-volume prediction requirements, please contact us at bd@protoqsar.com.

A: ProtoPRED^® stores information of the user in 4 different forms:

• Personal information provided during the registration. This information is stored by ProtoQSAR, but it is not used for any other purpose than provide the service (unless the permission is explicitly given during the registration process).
• Historic of the movements. All the movements done in the server that affect the number of tokens (such as predictions, generation of reports, credit purchases) can be consulted in the user page, to ensure proper record of the use of the server. By default, this historic includes data such as the SMILES and the predicted value, to facilitate user revision, but this can be changed in the user section.
• Last prediction. Complete results of the last prediction of each module are also stored internally, to allow the user to recover the data in case of accidental closure, connectivity issues or any other interruption. This can be also changed in the user section to ensure data privacy.
• QPRF data. To simplify the filling of the author data section in the QPRF (obligatory), default values are offered for the user name, address, phone and email. Those fields can be edited in the user page. For each PDF request, the current values can be stored as new defaults.

A: In the user page, there is a Reports data section to modify the QPRF author data. This form includes a check box for Data Privacy. If this is selected, all the information about the results will be removed, except for the model used and the date. Note that the data is really erased and cannot be recovered. This also will deactivate the option to recover the results of the last prediction for each module.

A: Yes. For queries about personal data, ProtoQSAR data representatives can be contacted at gdpr@protoqsar.com.

A: ProtoPRED^® produces four different types of output in the results page: a summary table with all the predictions performed; a model-specific table with additional details for each different model; a downloadable Excel spreadsheet (with more information such as the values in the internal model units); and downloadable standardised PDF reports (this could have an additional cost) such as QPRFs or specific combined reports, such as in ProtoICH and GenoITS.

A: ProtoPRED^® uses calculated molecular descriptors to characterise your substance. In some cases, these descriptors cannot be calculated due to numerical issues or incompatibilities between the structure and the descriptors. When this occurs, ProtoPRED^® uses imputation to estimate the values from our database. However, this is only permitted for a limited number of descriptors. If an excessive number of descriptors require imputation, the prediction will be considered unreliable and will not be provided. In such cases, neither the prediction nor the QPRF would be charged.

A: QSAR models are trained using a database of molecules with experimental values, so their validity is restricted to the chemical space defined by this dataset. ProtoPRED^® uses four different techniques to evaluate if a target molecule falls inside the chemical space of a model. Although a prediction will be provided, those outside the applicability domain are considered unreliable. In such cases, neither the prediction nor the QPRF would be charged.

A: Assessing the reliability of a calculation implies factors such as the overall predictivity of the model, the relationship of the target molecule with the training set, or the behaviour of the model with similar substances. An explanation of the parameters used and their values are provided in the QPRF. Note that this score is presented as a guidance and does not substitute a tailored expert assessment considering the details of each factor included in the QPRF.

A: No. The information on analogues is provided to assess the prediction. Analogues do not contribute directly to the prediction, but they do influence the reliability score of the prediction. However, they do influence the reliability score of the prediction.

A: Content for various sections of the QPRF can be provided, such as those regarding personal data, the identity of the substance, and additional comments in different sections. However, it is currently not possible to modify the existing content directly.

A: Retrying the operation may solve the issue. If it persists, please contact the support team at protopred@protoqsar.com.

A: Yes. Using the Retrieve the results from the last calculation button of each module, the results can be recovered (located above the INPUT MOLECULE(S) section). Note that this option should not work if the Data privacy option is selected in the reports data section of the user page.

A: The tokens should be added to the account name provided in the purchase form within 24 hours. A confirmation email should be received once the account has been updated. If the tokens do not appear, contact protopred@protoqsar.com with a copy of the purchase confirmation.

A: The ICH M7 guideline requires an integrated computational approach, where a statistical and an expert-rule method are combined. Additionally, a profiler is applied to detect the species considered part of the carcinogenicity cohort of concern, including aflatoxins, N-nitrosamines and azoxy compounds.

A: ProtoPRED^® includes buttons to automatically select the endpoints required for each of the Annexes in REACH, based on the substance's tonnage. However, this serves as general guidance, as the exact requirements may vary depending on existing data for these parameters or other properties.

A: An Integrated Testing Strategy (ITS) is a substance assessment approach that involves the selection and prioritisation of assays. This strategy aims to reduce or replace unnecessary experiments by making use of all available information during the assessment process.

A: No. GenoITS does not provide a prediction of genotoxicity, but rather a conclusion based on the results of the five studies required under REACH: in bacteria mutagenicity, in vitro and in vivo mutagenicity, and in vitro and in vivo cytogenicity. This information may be supplied by the user or derived from computational models, such as the QSAR models available through the ProtoPRED® platform.

A: Yes. The QPRF is the format recommended under REACH for reporting QSAR predictions, and it carries regulatory validity. GenoITS generates QPRFs for all models used within the module. Additionally, GenoITS provides a supplementary 'GenoITS report', which does not hold regulatory validity on its own.

A: Yes. The genotoxicity ITS recommended under REACH is annex-dependent; therefore, specifying the annex is mandatory in order to run the GenoITS module and obtain a conclusion.

A: In some cases, the compound under evaluation may fall outside the applicability domain of one or more of the models used, preventing a conclusion from being reached. In such instances, the report indicates which alternative study should be conducted by the user, along with the potential final conclusion depending on the outcome of that study.

A: The cost for performing the genotoxicity assessment of a compound is fixed at 20 tokens, regardless of the REACH annex selected. Additionally, the results and QPRFs for all five models included in the module are provided, irrespective of whether they are used to reach the final conclusion.