Health and environmental safety of nanomaterials: O Data, Where Art Thou?

Advances in neuroimaging, genomic, motion tracking, eye-tracking and many other technology-based data collection methods have led to a torrent of high dimensional datasets, which commonly have a small number of samples because of the intrinsic high cost of data collection involving human participants. High dimensional data with a small number of samples is of critical importance for identifying biomarkers and conducting feasibility and pilot work, however it can lead to biased machine learning (ML) performance estimates. Our review of studies which have applied ML to predict autistic from non-autistic individuals showed that small sample size is associated with higher reported classification accuracy. Thus, we have investigated whether this bias could be caused by the use of validation methods which do not sufficiently control overfitting. Our simulations show that K-fold Cross-Validation (CV) produces strongly biased performance estimates with small sample sizes, and the bias is still evident with sample size of 1000. Nested CV and train/test split approaches produce robust and unbiased performance estimates regardless of sample size. We also show that feature selection if performed on pooled training and testing data is contributing to bias considerably more than parameter tuning. In addition, the contribution to bias by data dimensionality, hyper-parameter space and number of CV folds was explored, and validation methods were compared with discriminable data. The results suggest how to design robust testing methodologies when working with small datasets and how to interpret the results of other studies based on what validation method was used.

As the production of nanoparticles of ZnO, TiO2 and CuO is increasing, their (eco)toxicity to bacteria Vibrio fischeri and crustaceans Daphnia magna and Thamnocephalus platyurus was studied with a special emphasis on product formulations (nano or bulk oxides) and solubilization of particles. Our innovative approach based on the combination of traditional ecotoxicology methods and metal-specific recombinant biosensors allowed to clearly differentiate the toxic effects of metal oxides per se and solubilized metal ions. Suspensions of nano and bulk TiO2 were not toxic even at 20 g l(-1). All Zn formulations were very toxic: L(E)C50 (mg l(-1)) for bulk ZnO, nanoZnO and ZnSO4.7H2O: 1.8, 1.9, 1.1 (V. fischeri); 8.8, 3.2, 6.1 (D. magna) and 0.24, 0.18, 0.98 (T. platyurus), respectively. The toxicity was due to solubilized Zn ions as proved with recombinant Zn-sensor bacteria. Differently from Zn compounds, Cu compounds had different toxicities: L(E)C50 (mg l(-1)) for bulk CuO, nano CuO and CuSO4: 3811, 79, 1.6 (V. fischeri), 165, 3.2, 0,17 (D. magna) and 95, 2.1, 0.11 (T. platyurus), respectively. Cu-sensor bacteria showed that toxicity to V. fischeri and T. platyurus was largely explained by soluble Cu ions. However, for Daphnia magna, nano and bulk CuO proved less bioavailable than for bacterial Cu-sensor. This is the first evaluation of ZnO, CuO and TiO2 toxicity to V. fischeri and T. platyurus. For nano ZnO and nano CuO this is also a first study for D. magna.

Using quantitative models to predict the biological interactions of nanoparticles will accelerate the translation of nanotechnology. Here, we characterized the serum protein corona 'fingerprint' formed around a library of 105 surface-modified gold nanoparticles. Applying a bioinformatics-inspired approach, we developed a multivariate model that uses the protein corona fingerprint to predict cell association 50% more accurately than a model that uses parameters describing nanoparticle size, aggregation state, and surface charge. Our model implicates a set of hyaluronan-binding proteins as mediators of nanoparticle-cell interactions. This study establishes a framework for developing a comprehensive database of protein corona fingerprints and biological responses for multiple nanoparticle types. Such a database can be used to develop quantitative relationships that predict the biological responses to nanoparticles and will aid in uncovering the fundamental mechanisms of nano-bio interactions.