S1: Engineered Nanomaterials

Predictive Modelling of Biological Effects of Engineered Nanomaterials

Dr Georgios Drakakis


National Technical University of Athens


Prof Haralambos Sarimveis, School of Chemical Engineering, National Technical University of Athens


Dr. Georgios Drakakis received his Diploma in Computer Science from the Aristotle University of Thessaloniki in 2007. He then proceeded to obtain two MSc degrees in Computer Science and Advanced Biological Sciences/Bioinformatics from Trinity College Dublin and the University of Liverpool respectively.

In January of 2015 he completed his PhD at the University of Cambridge in the area of Computational Chemistry/Cheminformatics and joined the National Technical University of Athens (ca. 2 years). He is currently a Post-Doctoral Research Associate at the Unit of Process Control and Informatics working on algorithm design and nano-QSAR modelling.

His research interests are in the areas of machine learning, data mining, cheminformatics, computational chemistry and computer vision. Over the past 4 years he has co-authored 10 original research publications in these fields and contributed to 2 book chapters.


Concerns about the safety of engineered nanomaterials (ENMs) in connection with the REACH direction to reduce animal testing of all substances has given rise to computational nanotoxicology.

Non-testing data can be generated by three main approaches:

a) grouping approaches, which include read-across and chemical category formation;
b) (quantitative) structure-activity relationships ((Q)SARs); and
c) expert systems.

The development and application of all three non-testing methods is based on the similarity principle, i.e. hypothesis that similar compounds should have similar biological activities. Contrary to pure chemicals where similarity is basically defined in terms of structural similarity, the complexity of ENMs leads to multiple perspective approaches for defining similarity, which may include routes of exposure, material types (e.g. fullerenes, carbon nanotubes, metal oxides etc.), physicochemical characteristics (e.g. size, shape, surface area, solubility etc.), biophysical interaction and biological impact (e.g. protein and lipid corona formation, gene expressions, cellular and organ responses) and bio-kinetics properties.

Computation of similarity and generation of predictive models often requires processing raw data, such as microscopic images and omics data. Obviously, a common framework that harmonizes and unifies all diverse data needed for describing and characterizing the complexity of ENMs would be extremely useful for researchers and practitioners in the nanosafety area.

The eNanoMapper project is addressing this need by developing an agreed ontology, i.e. a common language for the characterization of ENMs and a database infrastructure for data storage, sharing and searching. This presentation will focus on the JaqPot modelling infrastructure which has been developed in the context of the eNanoMapper project.

JaqPot is fully compatible with the eNanoMapper database and allows the automatic extraction and integration of diverse data for the creation of predictive nanoQSAR and read-across models and for optimal design of additional experiments. We will also demonstrate that by considering both physicochemical and biological similarity and taking into account information from open biological databases such as the Gene Ontology, we can boost the performance of predictive models.