Workshop: Predictive NanoQSAR models
1) Walkey, C. D., Olsen, J. B., Song, F., Liu, R., Guo, H., Olsen, D. W. H., Chan, W. C. W. (2014). Protein corona fingerprinting predicts the cellular interaction of gold and silver nanoparticles. ACS Nano, 8:3.
2) Hardy, B. et al, Collaborative Development of Predictive Toxicology Applications, Journal of Cheminformatics 2010, 2:7.
3) NTUA OpenTox Web services, available at http://opentox.ntua.gr:8080/ , described at http://opentox.ntua.gr/
ABSTRACT CONTENT / DETAILS:
This workshop will offer hands-on work on the development of NanoQSAR models, focusing on the use case of predicting cell association of Gold and Silver Nanoparticles, based on experimental data, publically available in the publication of Walkey et. al. (1).
The workshop will include the preparation of a data-file properly structured, so that they can be explored using the OpenTox infrastructure (2), the development of a PMML (Predictive Model Markup Language) XML-based file for defining transformation of descriptors, and the creation of predictive models using statistical and machine learning algorithms.
We will also explore how descriptors can be derived from nanoparticle images, using the Fiji/ImageJ open source software. Finally, R tools combined with protein corona data and Gene Ontology will be used to derive useful information on the mechanisms of action.
OpenTox web services are built on the RESTful API, a contemporary web technology characterized by flexibility and modularity, together with implementation and development independence.
Popular statistical and machine learning methodologies (such as Multiple Linear Regression and Support Vector Machines) as well as algorithms developed in-house (like Fast Training RBF Neural Networks) have been implemented in our group and are deployed as REST web services within the OpenTox framework. Security of services is controlled with the design and implementation of an SSO-based authorization and authentication API.