Predictive models
QSAR/Read-across
Data Driven Modelling
A large collection of models for the prediction of properties and toxicity end-points are provided as ready-to-use web applications through the NanoSolveIT platform. The models have been developed using various approaches, including machine learning methods, computational read-across techniques and prediction networks.
Available Services
Cytotoxicity (Cell Viability) Prediction for Metal Oxide NPs
A read across predictive model, validated according to OECD principles, with a well defined domain of applicability, for the prediction of the cytotoxicity of Metal Oxide Nanoparticles. Computational, atomistic and experimental nanodescriptors were used. The machine learning model is available as a web application.
Read-Across Model for Zeta Potential Prediction
A read across predictive model, validated according to OECD principles, with a well defined domain of applicability, for the prediction of the zetal potential of nanoparticles employing image descriptors from the NanoXtract tool. The machine learning model is available as a web application.
MS³bD NMs Zeta Potential Predictive Μodel
A read across predictive model, validated according to OECD principles, with a well defined domain of applicability, for the prediction of the zeta potential of nanomaterials in deionised water. The model employs a set of molecular and physicochemical descriptors. The machine learning model is available as a web application.
Vythos (Gold NM cell association, CNT absorption coefficients)
Vythos is a web application that includes read-across models developed using a Mixed Integer Linear Programming (MILP) groping and the development of linear models within each group.
NanoPot (logP, Zeta Potential in water, Cellular Uptake of A549 of Gold ENM)
This web application hosts three trained models to predict the logP, ZP and Cellular Uptake of A549 of Gold Nanoparticles. The models are trained on a set of geometrical descriptors (Delaunay tessellation) that contain information both on the geometry of the NPs and on some physicochemical characteristics of the involved atoms. The user selects the required model, uploads a PDB representation of the NP and the service calculates the descriptors and returns a prediction on the selected endpoint.
Cytotoxicity (Cell Viability) Classification Model
This web application is an online classification of cytotoxic metal oxide nanoparticles (NPs) . The model is trained on a small set of indicated properties, namely the energy of the MexOy conduction band, the average length of force vector surface tangent component for all metal atoms, the average length of force vector surface tangent component for metal atoms in surface region, the concentration of the exposure’s dose and the surface area of the metal oxide. The user may provide the input values either manually or by uploading a .csv file.
SAPNet: predict the toxicity of nanoparticle towards the CHO-K1 cell line
This web tool is an implementation of the Structure–Activity Prediction Network (SAPNet) model, which predicts the toxicity of nanoparticle towards the CHO-K1 cell line. An important advantage of the SAPNet concept is that it defines the nanoparticle physicochemical and structural properties, which are essential and can be adjusted to alter the target properties.
Metal oxide cytotoxicity prediction utilising facet-based electronic, and periodic table derived properties
This web tool is an implementation of a model predicting the toxicity of metal oxide nanomaterials towards bronchial epithelial (BEAS-2B), Murine myeloid (RAW 264.7) and E. coli cell lines. The model is validated according to the OECD guidelines and employs a combination of physicochemical, structural and periodic-table derived properties to perform predictions.
Drug Release: QSAR model for predicting the release of Doxorubicin from Doxorubicin-conjugated nanoformulations
This application is an online tool for predicting the % drug release rate of Doxorubicin from various Iron Oxide Nanoparticles (IONPs). The QSAR model is built using 14 experimental variables. Drug release was found to depend strongly on specific physicochemical properties of the nanoformulations (size, charge & saturation magnetization) and is largely influenced by external conditions (pH, temperature, release duration, AMF).
Ecotoxicological Read-Across Models for predicting acute toxicity
This weis web tool is an implementation of a model to predict the toxicological effects on Daphnia magna of freshly dispersed and 2-year aged nanomaterials. The model is validated according to the OECD guidelines and employs a set of nanomaterials’ physicochemical and media characteristics to predict whether a nanomaterial is toxic or non-toxic, based on an EC40 population reduction threshold.
Model for zeta potential prediction
Tree-based QSAR regression model for predicting the zeta potential of nanomaterials (NMs). The model uses a combination of physicochemical and molecular properties of the NMs, described by a set of 7 most relevant nanodescriptors.
Hamaker constant calculation model
Web application for calculating the Hamaker constant of specific materials in different chemical environments. The user can select among Universal ForceFields (UFF), Dreiding, OPLS, and CHARMM FFs and automatically calculate the Hamaker constant.
DoxViability: QSAR model for predicting the cytotoxicity of Doxorubicin-conjugated iron oxide nanoparticles
Web application for predicting the cytotoxic effects of Doxorubicin-conjugated Iron Oxide Nanoparticles (IONPs) against various cancer cells lines, expressed as % cell viability after treatment. The model uses 24 descriptors, both experimental and theoretical. Cytotoxic effects were found to depend strongly on the concentration of Dox delivered to the target, the ability of the nanoformulations to fully release the drug in the acidic environment of the cancer cells and on external conditions (magnetic field etc.). Our findings suggest that, apart from experimental descriptors mostly related to the Dox-IONPs as a whole, certain properties of the outermost surface ligands – like molecular weight & complexity, charged surface area and hydrophilicity – are factors influencing cytotoxicity.