Nano-Lazar: Read across Predictions for Nanoparticle Toxicities with Calculated and Measured Properties

Abstract

The lazar framework for read across predictions was expanded for the prediction of nanoparticle toxicities, and a new methodology for calculating nanoparticle descriptors from core and coating structures was implemented. Nano-lazar provides a flexible and reproducible framework for downloading data and ontologies from the open eNanoMapper infrastructure, developing and validating nanoparticle read across models, open-source code and a free graphical interface for nanoparticle read-across predictions. In this study we compare different nanoparticle descriptor sets and local regression algorithms. Sixty independent crossvalidation experiments were performed for the Net Cell Association endpoint of the Protein Corona dataset. The best RMSE and r2 results originated from models with protein corona descriptors and the weighted random forest algorithm, but their 95% prediction interval is significantly less accurate than for models with simpler descriptor sets (measured and calculated nanoparticle properties). The most accurate prediction intervals were obtained with measured nanoparticle properties (no statistical significant difference (p < 0.05) of RMSE and r2 values compared to protein corona descriptors). Calculated descriptors are interesting for cheap and fast high-throughput screening purposes. RMSE and prediction intervals of random forest models are comparable to protein corona models, but r2 values are significantly lower.

Practices and Trends of Machine Learning Application in Nanotoxicology

“…Six studies, out of the 86 gathered, used Genetic Algorithm (GA) for feature selection [30][31][32][33][34][35]. Five of them used Pearson correlation coefficients between pairs of variables to identify those that correlate with the endpoint or correlations among variables to avoid inter-correlations [36][37][38][39]. Few of the studies applied more than one feature selection technique.…”

“…RF has been demonstrated to be ideal for rigorous meta-analysis of complex and heterogeneous data [64]. Helma et al [36] note in their study that, with the exclusion of p-chem/proteomics descriptors, the RF model performed better than PLS and weighted average models. They showed excellent predictivity with small or large datasets, which performed well even with missing values.…”

Perspectives from the NanoSafety Modelling Cluster on the validation criteria for (Q)SAR models used in nanotechnology

“…Reproducibility of the models, including easy transfer and exchange across different platforms, is an important issue in QSAR modelling (Tetko et al, 2017), which also applies to the case of nano-QSAR analysis (Helma et al, 2017). Indeed, the reproducibility of computational science more generally has been a key concern in the recent scientific literature (Editorial, 2014).…”

“…In addition to descriptive documentation of the modelling workflow, full compliance with principle 2 (discussed in section 2.2), may require the relevant source code and, where relevant, random number generator seeds and other computational details to be documented in order to fully reproduce the models, and/or for the structure of the models to be encoded in the Predictive Model Markup Language (PMML) along with other necessary information required to calculate the descriptors etc. (Editorial, 2014;Helma et al, 2017;Tetko et al, 2017). (The Òother necessary informationÓ might be documented using the QMRF or QSARdb formats (Tetko et al, 2017), possibly with adaptations as discussed in section 2.2.)…”

Read-across predictions of nanoparticle hazard endpoints: a mathematical optimization approach

“…k-nearest neighbor, partial least squares, random forests). 5,19 In the categorical approach, the ENM samples are organized into groups of similar compounds. Groups are formed considering structural similarities between samples, and it is assumed that due to these similarities, the biological or toxic activity of the ENMs within a group follows a regular pattern.…”