HOME

Regarding toxicological safety, our group is dedicated to the development of tools for the assessment and evaluation of the potential risks associated with exposure to chemical substances, drugs, pollutants, or other agents that may be harmful to human health or the environment. With the aim of determining safe exposure levels we produce more sustainable, ethical and effective toxicological evaluations without resourcing to in vivo assays.

The expertise of our group members spans a broad range of scientific areas including pharmacology, toxicology, safety policy development, statistics, and data science. We combine our solid knowledge with several years of experience to provide support in different projects and develop high quality applications supporting standard toxicological assessment methods.

Multilevel machine learning approaches for toxicological predictions

Multilevel machine learning approaches in toxicological predictions involve the integration of different data sources and models to improve the accuracy and reliability of toxicity assessments. These approaches aim to capture toxicity-related information at various levels to provide a comprehensive understanding of the potential adverse effects of chemicals or substances.

Some key components and methods we apply are: Data Integration, Multilevel Modeling and Model Fusion.

By integrating data from multiple levels, these approaches offer a more holistic understanding of toxicity mechanisms and can aid in prioritizing chemicals for further testing or assessing potential risks associated with exposure.

Through machine learning algorithms, multilevel modeling methods and with the latest digital tools and scientific software, which some of them are also developed within our group, we work so as to obtain statistical approaches to determine whether a substance is considered toxic or not.

Related projects: eTRANSAFE, RiskHunt3r.

In silico methods for drug safety assessment

Drug safety assessment is one of the most important bottlenecks in drug development. In this context, computational methodologies play a crucial role in various stages of the drug development process, adding value to the discovery, design, optimization, and evaluation of drug candidates.

They also offer a very interesting alternative to other experimental methods; they are faster, cheaper and require no amount of valuable compounds to obtain predictions that often have comparable quality with those obtained using in vitro or even in vivo methods.

In this regard, we develop innovative computational methods and models for toxicological assessments. Moreover, we have developed software platforms which facilitates data exchange and integration of toxicity studies and other endpoints with visualization and prediction services.

Related projects we have coordinated: eTOX and eTRANSAFE.

Support of NAMs development for NGRA

Science and society demand a paradigm shift towards animal-free chemical safety and risk assessment. This is why alternatives to animal testing are constantly gaining in importance due to their human relevance, mechanistic character as well as other ethical and time related benefits.

With the aim of facilitating the integration of this novel testing paradigm, we actively contribute to the Next Generation Risk Assessment (NGRA) by developing innovative, mechanistic in silico New Approach Methodologies (NAMs) characterized by high human relevance.

On the one hand, our activities focus on in silico assisted hazard identification. We also model and predict information used for quantitative MIE and KE parameterization of quantitative Adverse Outcome Pathway (qAOP) networks. On the other hand, we are working to develop a dashboard that will offer a professional data repository management, a test method and protocol repository, an interactive compound dashboard and a case study management module. The main objective is to make it reproducible in a way that enables reuse in the future.

Related projects: Risk Hunt3r

Scientific software development

Within our group we develop scientific software including not only the backend but also the frontend, gathering and integrating multiple data sources and allowing the access to such data through a single application programming interface (API).

We use scientific libraries in our usual developments like Scikit-learn, NumPy, RDKit (Open-Source Cheminformatic Software) among other.

Our main developments are open-source codes, which are available for the entire scientific community.

For more details, repositories and examples of our developments, please see section Software