Diverse nanostructures of our in-house database
Nanoscience and technology have made a big impact on modern medicine, energy and other areas. However, experimental testing of nanomaterials for important physic-chemical properties, bioactivities and nanotoxicities is costly and time consuming. Informatics related studies are promising to answer this challenge by virtually designing new biocompatible nanomaterials. In Zhu lab, we want to create novel nanoinformatics tools for data curation, model developments and data/model sharing purposes. Novel nano-bio interaction models, including complex nanotoxicity models, were developed using modern machine learning approaches. This project will lead to the design of novel biocompatible nanomaterials with low toxicity and desired bioactivities.
Based on the annotated nanostructure, novel geometrical nanodescriptors were developed by employing Delaunay tessellation and atomic properties. The nanodescriptors digitalized nanomaterials by quantifying nanostructures. The nanodescriptors can be used to develop quantitative nanostructure activity relationship models.
ViNAS-Pro (Virtual Nanomaterial Simulation Professional) is a data-driven nanoinformatics platform. It comprises six core components that enable data profiling, data preprocessing and visualization, machine learning modeling, virtual nanomaterials prediction, as well as nanostructure and nanodescriptor calculation.
ViNAS-Pro maintains two machine readable databases: the Structure database and the Assay database. The Structure database provides structural information for 13 types of NMs, while the Assay database offers data on the experimentally assessed properties and biological activities of these NMs across 25 different assays. The Descriptor toolkit provides users with modules for data visualization and preprocessing, ensuring the structure diversity of the training data in the machine learning (ML) modeling procedure. The Model toolkit includes two modules: NanoPredictor and AutoNanoML. The NanoPredictor module maintains pre-developed ML models, enabling users to predict specific endpoints for new NMs. The AutoNanoML module provides an interface that allows users to develop their own ML models for various prediction purposes. The Library component provides data analysis, structure data, and endpoint predictions for virtual NMs. ViNAS-Pro provides services for data deposit, nanostructure construction, and nanodescriptor calculation through the Service component.
Related Research Articles: