Nowadays, the interest in the search for new nanomaterials with improved electrical, optical, catalytic and biological properties has increased. Despite the potential benefits that can be gathered from the use of nanoparticles, only little attention has been paid to their possible toxic effects that may affect human health. In this context, several assays have been carried out to evaluate the cytotoxicity of nanoparticles in mammalian cells. Owing to the cost in both resources and time involved in such toxicological assays, there has been a considerable increase in the interest towards alternative computational methods, like the application of quantitative structure–activity/toxicity relationship (QSAR/QSTR) models for risk assessment of nanoparticles. However, most QSAR/QSTR models developed so far have predicted cytotoxicity against only one cell line, and they did not provide information regarding the influence of important factors rather than composition or size. This work reports a QSTR-perturbation model aiming at simultaneously predicting the cytotoxicity of different nanoparticles against several mammalian cell lines, and also considering different times of exposure of the cell lines, as well as the chemical composition of nanoparticles, size, conditions under which the size was measured, and shape. The derived QSTR-perturbation model, using a dataset of 1681 cases (nanoparticle–nanoparticle pairs), exhibited an accuracy higher than 93% for both training and prediction sets. In order to demonstrate the practical applicability of our model, the cytotoxicity of different silica (SiO₂), nickel (Ni), and nickel(II) oxide (NiO) nanoparticles were predicted and found to be in very good agreement with experimental reports. To the best of our knowledge, this is the first attempt to simultaneously predict the cytotoxicity of nanoparticles under multiple experimental conditions by applying a single unique QSTR model.

Nanotechnology has brought great advances to many fields of modern science. A manifold of applications of nanoparticles have been found due to their interesting optical, electrical, and biological/chemical properties. However, the potential toxic effects of nanoparticles to different ecosystems are of special concern nowadays. Despite the efforts of the scientific community, the mechanisms of toxicity of nanoparticles are still poorly understood. Quantitative-structure activity/toxicity relationships (QSAR/QSTR) models have just started being useful computational tools for the assessment of toxic effects of nanomaterials. But most QSAR/QSTR models have been applied so far to predict ecotoxicity against only one organism/bio-indicator such as Daphnia magna. This prevents having a deeper knowledge about the real ecotoxic effects of nanoparticles, and consequently, there is no possibility to establish an efficient risk assessment of nanomaterials in the environment. In this work, a perturbation model for nano-QSAR problems is introduced with the aim of simultaneously predicting the ecotoxicity of different nanoparticles against several assay organisms (bio-indicators), by considering also multiple measures of ecotoxicity, as well as the chemical compositions, sizes, conditions under which the sizes were measured, shapes, and the time during which the diverse assay organisms were exposed to nanoparticles. The QSAR-perturbation model was derived from a database containing 5520 cases (nanoparticle-nanoparticle pairs), and it was shown to exhibit accuracies of ca. 99% in both training and prediction sets. In order to demonstrate the practical applicability of our model, three different nickel-based nanoparticles (Ni) with experimental values reported in the literature were predicted. The predictions were found to be in very good agreement with the experimental evidences, confirming that Ni-nanoparticles are not ecotoxic when compared with other nanoparticles. The results of this study thus provide a single valuable tool toward an efficient prediction of the ecotoxicity of nanoparticles under multiple experimental conditions.

Azithromycin (AZY) is among the antibiotics with the highest concentrations in aqueous matrices.

Understanding protein-protein interactions is a key challenge in biochemistry. In this work, we describe a more accurate methodology to predict Hot-Spots (HS) in protein-protein interfaces from their native complex structure compared to previous published Machine Learning (ML) techniques. Our model is trained on a large number of complexes and on a significantly larger number of different structural- and evolutionary sequence-based features. In particular, we added interface size, type of interaction between residues at the interface of the complex, number of different types of residues at the interface and the Position-Specific Scoring Matrix (PSSM), for a total of 79 features. We used twenty-seven algorithms from a simple linear-based function to support-vector machine models with different cost functions. The best model was achieved by the use of the conditional inference random forest (c-forest) algorithm with a dataset pre-processed by the normalization of features and with up-sampling of the minor class. The method has an overall accuracy of 0.80, an F1-score of 0.73, a sensitivity of 0.76 and a specificity of 0.82 for the independent test set.