Automated machine learning in nanotoxicity assessment: A comparative study of predictive model performance.

Computational modeling has earned significant interest as an alternative to animal testing of toxicity assessment. However, the process of selecting an appropriate algorithm and fine-tuning hyperparameters for the developing of optimized models takes considerable time, expertise, and an intensive search. The recent emergence of automated machine learning (autoML) approaches, available as user-friendly platforms, has proven beneficial for individuals with limited knowledge in ML-based predictive model development. These autoML platforms automate crucial steps in model development, including data preprocessing, algorithm selection, and hyperparameter tuning. In this study, we used seven previously published and publicly available datasets for oxides and metals to develop nanotoxicity prediction models. AutoML platforms, namely Vertex AI, Azure, and Dataiku, were employed and performance measures such as accuracy, F1 score, precision, and recall for these autoML-based models were then compared with those of conventional ML-based models. The results demonstrated clearly that the autoML platforms produced more reliable nanotoxicity prediction models, outperforming those built with conventional ML algorithms. While none of the three autoML platforms significantly outperformed the others, distinctions exist among them in terms of the available options for choosing technical features throughout the model development steps. This allows users to select an autoML platform that aligns with their knowledge of predictive model development and its technical features. Additionally, prediction models constructed from datasets with better data quality displayed, enhanced performance than those built from datasets with lower data quality, indicating that future studies with high-quality datasets can further improve the performance of those autoML-based prediction models.

Developing random forest based QSAR models for predicting the mixture toxicity of TiO2 based nano-mixtures to Daphnia magna.

During emission, TiO2 nanoparticles (NPs) might meet various chemicals, including metal ions and organic compounds in aquatic environments (e.g., surface water, sediments). At environmentally safe concentrations, combinations of both TiO2 NPs and those chemicals might cause cocktail effects (i.e., mixture toxicity) to aquatic organisms. Previous models such as concentration addition and independent action require dose-response curves of single components in the mixtures to predict the mixture toxicity. Structure-activity relationship (QSAR) models might predict the toxicity of nano-mixtures without dose-response curves of single components in the mixtures. However, current quantitative structure-activity relationship (QSAR) models are mainly focused on predicting cytotoxicity (i.e., cell viability) of heterogeneous metallic TiO2 nanoparticles (NPs) or mixtures of TiO2 NPs and four metal ions (Cu2+, Cd2+, Ni2+, and Zn2+). To minimize the experimental cost of nano-mixture risk assessment, in this study, we developed novel nano-mixture QSAR models to predict i) EC50 of 76 nano-mixtures containing TiO2 NPs and one of eight inorganic/organic compounds (i.e., AgNO3, Cd(NO3)2, Cu(NO3)2, CuSO4, Na2HAsO4, NaAsO2, Benzylparaben and Benzophenone-3), to Daphnia magna(D. magna), and ii) immobilization of D. magna exposed to one of 98 mixtures containing TiO2 NPs and one of eleven inorganic/organic compounds (i.e., AgNO3, Cd(NO3)2, Cu(NO3)2, CuSO4, Na2HAsO4, NaAsO2, Benzylparaben Benzophenone-3, Pirimicarb, Pentabromodiphenyl Ether and Triton X-100). The nano-mixture QSAR models were developed with mixture descriptors (Dmix) combing quantum descriptors of mixture components (e.g., TiO2 NPs and its partners) by using different machine learning techniques (i.e., random forest, neural network, support vector machine, and multiple linear regression). Nano-mixture QSAR models built with the random forest algorithm and proposed mixture descriptors exhibited good performance for predicting logEC50 (Adj.R2test = 0.955 ± 0.003, RMSEtest = 0.016 ± 0.002, and MAEtest = 0.008 ± 0.001) and immobilization (Adj.R2test = 0.888 ± 0.011, RMSEtest = 11.327 ± 0.730, and MAEtest = 5.933 ± 0.442). The models developed in this study were implemented in a user-friendly application for assessing the aquatic toxicity of TiO2 based nano-mixtures.

Evaluation of the Performance of a ZnO-Nanoparticle-Coated Hydrocolloid Patch in Wound Healing.

Hydrocolloid dressings are an important method for accelerating wound healing. A combination of a hydrocolloid and nanoparticles (NPs), such as gold (Au), improves the wound healing rate, but Au-NPs are expensive and unable to block ultraviolet (UV) light. Herein, we combined zinc oxide nanoparticles (ZnO-NPs) with hydrocolloids for a less expensive and more effective UV-blocking treatment of wounds. Using Sprague-Dawley rat models, we showed that, during 10-day treatment, a hydrocolloid patch covered with ZnO-NPs (ZnO-NPs-HC) macroscopically and microscopically stimulated the wound healing rate and improved wound healing in the inflammation phase as shown by reducing of pro-inflammatory cytokines (CD68, IL-8, TNF-α, MCP-1, IL-6, IL-1ß, and M1) up to 50%. The results from the in vitro models (RAW264.7 cells) also supported these in vivo results: ZnO-NPs-HCs improved wound healing in the inflammation phase by expressing a similar level of pro-inflammatory mediators (TNF-α and IL-6) as the negative control group. ZnO-NPs-HCs also encouraged the proliferation phase of the healing process, which was displayed by increasing expression of fibroblast biomarkers (α-SMA, TGF-ß3, vimentin, collagen, and M2) up to 60%. This study provides a comprehensive analysis of wound healing by measuring the biomarkers in each phase and suggests a cheaper method for wound dressing.

Scanning transmission X-ray microscopy study of subcellular granules in human platelets at the carbon K- and calcium L2,3-edges.

Platelets and their subcellular components (e.g., dense granules) are essential components in hemostasis. Understanding their chemical heterogeneities at the sub-micrometer scale, particularly their activation during hemostasis and production of platelet-derived extracellular vesicles, may provide important insights into their mechanisms; however, this has rarely been investigated, mainly owing to the lack of appropriate chemical characterization tools at nanometer scale. Here, the use of scanning transmission X-ray microscopy (STXM) combined with X-ray absorption near edge structure (XANES) to characterize human platelets and their subcellular components at the carbon K-edge and calcium L2,3-edge, is reported. STXM images can identify not only the spatial distribution of subcellular components in human platelets, such as dense granules (DGs) with sizes of ~200 nm, but also their granule-to-granule chemical heterogeneities on the sub-micrometer scale, based on their XANES spectra. The calcium distribution map as well as the principal component analysis of the STXM image stacks clearly identified the numbers and locations of the calcium-rich DGs within human platelets. Deconvolution of the carbon K-edge XANES spectra, extracted from various locations in the platelets, showed that amide carbonyl and carboxylic acid functional groups were mainly found in the cytoplasm, while ketone-phenol-nitrile-imine, aliphatic, and carbonate functional groups were dominant in the platelet DGs. These observations suggest that platelet DGs are most likely composed of calcium polyphosphate associated with adenosine triphosphate (ATP) and adenosine diphosphate (ADP), with significant granule-to-granule variations in their compositions, while the cytoplasm regions of platelets contain significant amounts of proteins.

Status Quo in Data Availability and Predictive Models of Nano-Mixture Toxicity.

Co-exposure of nanomaterials and chemicals can cause mixture toxicity effects to living organisms. Predictive models might help to reduce the intensive laboratory experiments required for determining the toxicity of the mixtures. Previously, concentration addition (CA), independent action (IA), and quantitative structure-activity relationship (QSAR)-based models were successfully applied to mixtures of organic chemicals. However, there were few studies concerning predictive models for toxicity of nano-mixtures before June 2020. Previous reviews provided comprehensive knowledge of computational models and mechanisms for chemical mixture toxicity. There is a gap in the reviewing of datasets and predictive models, which might cause obstacles in the toxicity assessment of nano-mixtures by using in silico approach. In this review, we collected 183 studies of nano-mixture toxicity and curated data to investigate the current data and model availability and gap and to derive research challenges to facilitate further experimental studies for data gap filling and the development of predictive models.

Identification of Ca-rich dense granules in human platelets using scanning transmission X-ray microscopy.

Whole-mount (WM) platelet preparation followed by transmission electron microscopy (TEM) observation is the standard method currently used to assess dense granule (DG) deficiency (DGD). However, due to the electron-density-based contrast mechanism in TEM, other granules such as α-granules might cause false DG detection. Here, scanning transmission X-ray microscopy (STXM) was used to identify DGs and minimize false DG detection of human platelets. STXM image stacks of human platelets were collected at the calcium (Ca) L2,3 absorption edge and then converted to optical density maps. Ca distribution maps, obtained by subtracting the optical density maps at the pre-edge region from those at the post-edge region, were used to identify DGs based on the Ca richness. DGs were successfully detected using this STXM method without false detection, based on Ca maps for four human platelets. Spectral analysis of granules in human platelets confirmed that DGs contain a richer Ca content than other granules. The Ca distribution maps facilitated more effective DG identification than TEM which might falsely detect DGs. Correct identification of DGs would be important to assess the status of platelets and DG-related diseases. Therefore, this STXM method is proposed as a promising approach for better DG identification and diagnosis, as a complementary tool to the current WM TEM approach.

Implementation of Complementary Model using Optimal Combination of Hematological Parameters for Sepsis Screening in Patients with Fever.

The early detection and timely treatment are the most important factors for improving the outcome of patients with sepsis. Sepsis-related clinical score, such as SIRS, SOFA and LODS, were defined to identify patients with suspected infection and to predict severity and mortality. A few hematological parameters associated with organ dysfunction and infection were included in the score although various clinical pathology parameters (hematology, serum chemistry and plasma coagulation) in blood sample have been found to be associated with outcome in patients with sepsis. The investigation of the parameters facilitates the implementation of a complementary model for screening sepsis to existing sepsis clinical criteria and other laboratory signs. In this study, statistical analysis on the multiple clinical pathology parameters obtained from two groups, patients with sepsis and patients with fever, was performed and the complementary model was elaborated by stepwise parameter selection and machine learning. The complementary model showed statistically better performance (AUC 0.86 vs. 0.74-0.51) than models built up with specific hematology parameters involved in each existing sepsis-related clinical score. Our study presents the complementary model based on the optimal combination of hematological parameters for sepsis screening in patients with fever.

Fatal avian influenza A(H5N1) infection in a 36-week pregnant woman survived by her newborn in Sóc Trang Province, Vietnam, 2012.

BACKGROUND: Reports of pregnant women infected with avian influenza are rare. Studies showed that A/H5N1 virus can penetrate the placental barrier and infect the fetus. Of six documented cases, four died and two survivors had a spontaneous abortion. OBJECTIVES: We report a clinical, outcome and epidemiological characteristics of a 36-week pregnant woman infected with A/H5N1 and her newborn in Soc Trang province of Vietnam in 2012. METHODS: Epidemiological and laboratory investigations were conducted. Clinical manifestations, progress, treatment and outcome of the case-patient and her newborn were collected. Human tracheal aspirate, throat swab and serum specimens were tested for influenza A/H5N1, A/H3N1, A/H1N1pdm09 and B by real-time RT-PCR and genome sequencing. Poultry throat and rectal swabs were tested by PCR and virus isolation. RESULTS: Case-patient hospitalized with high fever and cough, and died after onset 6 days. She continuously slaughtered sick poultry 5 days before illness onset. Clinical manifestation showed rapid progressive severe pneumonia. Her tracheal aspirate sample was positive influenza A/H5N1 virus. Her new-born was delivered by caesarean section with low birth weight and early onset pneumonia, however fully recovered after 16 days treatment. Neonate's throat swabs and paired serum samples tested negative for influenza A/H5N1. Clade 1.1 A/H5N1 virus was detected in poultry samples, was same clade and highly homogenous with the virus was detected in the mother. CONCLUSIONS: This was the first documented a live birth from a pregnant woman infected with influenza A/H5N1 virus. Intensive studies are needed to better understand mother-to-child transmission of influenza A/H5N1 virus.

Quasi-QSAR for predicting the cell viability of human lung and skin cells exposed to different metal oxide nanomaterials.

A quasi-QSAR model was developed to predict the cell viability of human lung (BEAS-2B) and skin (HaCaT) cells exposed to 21 types of metal oxide nanomaterials. A wide range of toxicity datasets obtained from the S2NANO (www.s2nano.org) database was used. The data of descriptors representing the physicochemical properties and experimental conditions were coded to quasi-SMILES. In particular, hierarchical cluster analysis (HCA) and min-max normalization method were respectively used in assigning alphanumeric codes for numerical descriptors (e.g., core size, hydrodynamic size, surface charge, and dose) and then quasi-QSAR model performances for both methods were compared. The quasi-QSAR models were developed using CORAL software (www.insilico.eu/coral). Quasi-QSAR model built using quasi-SMILES generated by means of HCA showed better performance than the min-max normalization method. The model showed satisfactory statistical results (Radj2 for the training dataset: 0.71-0.73; Radj2 for the calibration dataset: 0.74-0.82; and Radj2 for the validation dataset: 0.70-0.76).