Periodic Reporting for period 1 - QUANTUM-TOX (QUANTUM-TOX - Revolutionizing Computational Toxicology with Electronic Structure Descriptors andArtificial Intelligence)
Período documentado: 2024-02-01 hasta 2025-01-31
Computational toxicology plays a crucial role in assessing the safety of chemicals, pharmaceuticals, and environmental pollutants while reducing the reliance on costly and time-consuming experimental studies. By leveraging in silico models, it is possible to predict the toxicity of compounds early in the development pipeline, improving risk assessment and regulatory decision-making. However, existing computational approaches face significant challenges, particularly in their ability to cover the vast chemical space. Typically, toxicity prediction models rely on molecular fingerprints based on structural similarity, which may fail to generalize to novel or underrepresented chemical classes. Additionally, these models are often very complex, requiring thousands of descriptors.
A desirable computational toxicology framework should offer broader coverage of chemical space while using a small number of interpretable descriptors. The QUANTUM-TOX consortium is addressing these limitations by developing a next-generation fingerprint that moves beyond conventional structure-based predictions. Instead of relying solely on molecular structures, the new approach leverages electronic structure information to provide a more fundamental and transferable representation of chemical behavior. By incorporating quantum-mechanical descriptors, this fingerprint aims to enhance the accuracy of toxicity predictions and overcome the limitations of existing models, ultimately improving the reliability and applicability of computational toxicology.
A desirable computational toxicology framework should offer broader coverage of chemical space while using a small number of interpretable descriptors. The QUANTUM-TOX consortium is addressing these limitations by developing a next-generation fingerprint that moves beyond conventional structure-based predictions. Instead of relying solely on molecular structures, the new approach leverages electronic structure information to provide a more fundamental and transferable representation of chemical behavior. By incorporating quantum-mechanical descriptors, this fingerprint aims to enhance the accuracy of toxicity predictions and overcome the limitations of existing models, ultimately improving the reliability and applicability of computational toxicology.
During the EIC project, the consortium has worked towards completing its technical, organizational, and strategic goals, including but not limited to:
- Finalized chemicals for skin sensitization
- Initial geometries obtained and reviewed by UNIMAN through scans to conduct prelim testing. This has aided in WP2 specific goals in developing the quantum chemical descriptors and methodologies.
- Perturbations and Descriptors of Covalent Binding completed. Reactivity analysis, CDFT analysis, and initial validation using multiple linear regression were done by FCC to ensure the performance of the computational methodologies .
- An AI system to analyze discrete-valued data, linking features of electronic signatures (ESigns) to toxicity experimental endpoints using least-squares regression was done by FCC.
- Work has started on development and use of QSAR models to compare results with QM by LJMU.
All the participants are working towards technical development of the project. Apart from this, regular communication and dissemination work is being done regularly to exploit the results and gain market visibility.
- Finalized chemicals for skin sensitization
- Initial geometries obtained and reviewed by UNIMAN through scans to conduct prelim testing. This has aided in WP2 specific goals in developing the quantum chemical descriptors and methodologies.
- Perturbations and Descriptors of Covalent Binding completed. Reactivity analysis, CDFT analysis, and initial validation using multiple linear regression were done by FCC to ensure the performance of the computational methodologies .
- An AI system to analyze discrete-valued data, linking features of electronic signatures (ESigns) to toxicity experimental endpoints using least-squares regression was done by FCC.
- Work has started on development and use of QSAR models to compare results with QM by LJMU.
All the participants are working towards technical development of the project. Apart from this, regular communication and dissemination work is being done regularly to exploit the results and gain market visibility.
The advancements made through the QUANTUM-TOX project will transform research in drug discovery, general toxicity prediction, and molecular modeling by integrating innovative fingerprints with AI models. These developments will be driven by a groundbreaking electronic structure-based fingerprint - ESign, offering a more accurate and fundamental approach to predicting chemical behavior.
In addition to its planned initiatives, the consortium actively engages in outreach efforts with key stakeholders, including the pharmaceutical industry and organizations specializing in in vitro toxicity research. Partnering with pharmaceutical companies is important for gaining access to proprietary data and ensuring that the project’s software aligns with the companies’ protocols and workflows. Collaboration with in vitro research institutions helps develop standardized in silico/in vitro methods that improve predictive accuracy and streamline risk assessment processes. These interactions strengthen the project’s applicability and ensure its tools meet real-world scientific and regulatory needs.
In addition to its planned initiatives, the consortium actively engages in outreach efforts with key stakeholders, including the pharmaceutical industry and organizations specializing in in vitro toxicity research. Partnering with pharmaceutical companies is important for gaining access to proprietary data and ensuring that the project’s software aligns with the companies’ protocols and workflows. Collaboration with in vitro research institutions helps develop standardized in silico/in vitro methods that improve predictive accuracy and streamline risk assessment processes. These interactions strengthen the project’s applicability and ensure its tools meet real-world scientific and regulatory needs.