18 Integrating QSAR, Read-Across, and Screening Tools …
377
other tools are thought of for the USA’s regulations; therefore, the thresholds used
are different compared to the European ones.
Furthermore, previous models applied one single model for each endpoint, typically. Within JANUS, 48 models run together and provide the uncertainty of the final
score based on multiple results for the same endpoint.
18.4 Discussion
The VEGAHUB platform represents an example of a mature system, which goes
beyond the predictions, and tries to extract from the individual tools all elements
useful for the evaluation of the chemical substances. We moved our perspective from
the point of view of the developers to the point of view of the user. This affected a
number of our choices.
The software is freely downloadable, and this has been done in order to increase
its use. The payment of a fee or a license is a barrier for some users, in particular
public bodies, academia and small industries. Furthermore, industry and regulators
do not like to disclose structures of the substances they are studying, and for this
reason, we developed a tool that can be downloaded and used internally.
We are aware that multiple models exist or may be developed for the same endpoint. This increases the confidence of the user. Consequently, we tried to develop
more than one model for the same endpoint. We also incorporated models from other
systems and established collaborations with a series of developers. Indeed, VEGA
is promoting links with other systems.
A peculiar feature provided by VEGA is the ADI. This is not available within
other popular systems, such as ToxTree and EPISuite. Indeed, the evaluation of the
AD within these systems is not possible, or is prohibitively time-consuming, since
it has to be done manually on a set of parameters.
The ADI within VEGA is composed of several components, which relate to the
three components of any QSAR model: chemical, computational, and property. Many
models evaluate the AD considering the chemical component only and evaluate how
similar the target compound is, compared to the population of substances within the
training set of the model. However, this is only one part of the AD. Since at the
basis of any QSAR model there are also aspects related to the algorithm and the
biological/toxicological/environmental property, these aspects may surely affect the
model and thus the AD. For this reason, we developed a sophisticated ADI.
Multiple factors are at the basis of the evaluation of substances, and we progressively integrate tools to assist the users facilitating this evaluation. ToxWeight is a
clear example. The program automatically runs QSAR models, compares the results
of the different models considering their individual reliability, and in parallel evaluates SA within a read-across perspective, which is merged with the QSAR results.
This strategy has a clear reference with the Guidance on Weight-of-Evidence published by EFSA [8]. The evaluation of the reliability and consistency of the individual
lines of evidence are indeed at the basis of this Guidance.
377
other tools are thought of for the USA’s regulations; therefore, the thresholds used
are different compared to the European ones.
Furthermore, previous models applied one single model for each endpoint, typically. Within JANUS, 48 models run together and provide the uncertainty of the final
score based on multiple results for the same endpoint.
18.4 Discussion
The VEGAHUB platform represents an example of a mature system, which goes
beyond the predictions, and tries to extract from the individual tools all elements
useful for the evaluation of the chemical substances. We moved our perspective from
the point of view of the developers to the point of view of the user. This affected a
number of our choices.
The software is freely downloadable, and this has been done in order to increase
its use. The payment of a fee or a license is a barrier for some users, in particular
public bodies, academia and small industries. Furthermore, industry and regulators
do not like to disclose structures of the substances they are studying, and for this
reason, we developed a tool that can be downloaded and used internally.
We are aware that multiple models exist or may be developed for the same endpoint. This increases the confidence of the user. Consequently, we tried to develop
more than one model for the same endpoint. We also incorporated models from other
systems and established collaborations with a series of developers. Indeed, VEGA
is promoting links with other systems.
A peculiar feature provided by VEGA is the ADI. This is not available within
other popular systems, such as ToxTree and EPISuite. Indeed, the evaluation of the
AD within these systems is not possible, or is prohibitively time-consuming, since
it has to be done manually on a set of parameters.
The ADI within VEGA is composed of several components, which relate to the
three components of any QSAR model: chemical, computational, and property. Many
models evaluate the AD considering the chemical component only and evaluate how
similar the target compound is, compared to the population of substances within the
training set of the model. However, this is only one part of the AD. Since at the
basis of any QSAR model there are also aspects related to the algorithm and the
biological/toxicological/environmental property, these aspects may surely affect the
model and thus the AD. For this reason, we developed a sophisticated ADI.
Multiple factors are at the basis of the evaluation of substances, and we progressively integrate tools to assist the users facilitating this evaluation. ToxWeight is a
clear example. The program automatically runs QSAR models, compares the results
of the different models considering their individual reliability, and in parallel evaluates SA within a read-across perspective, which is merged with the QSAR results.
This strategy has a clear reference with the Guidance on Weight-of-Evidence published by EFSA [8]. The evaluation of the reliability and consistency of the individual
lines of evidence are indeed at the basis of this Guidance.