Integrating Explainability into Graph Neural Network Models for the Prediction of X-ray Absorption Spectra.

The use of sophisticated machine learning (ML) models, such as graph neural networks (GNNs), to predict complex molecular properties or all kinds of spectra has grown rapidly. However, ensuring the interpretability of these models' predictions remains a challenge. For example, a rigorous understanding of the predicted X-ray absorption spectrum (XAS) generated by such ML models requires an in-depth investigation of the respective black-box ML model used. Here, this is done for different GNNs based on a comprehensive, custom-generated XAS data set for small organic molecules. We show that a thorough analysis of the different ML models with respect to the local and global environments considered in each ML model is essential for the selection of an appropriate ML model that allows a robust XAS prediction. Moreover, we employ feature attribution to determine the respective contributions of various atoms in the molecules to the peaks observed in the XAS spectrum. By comparing this peak assignment to the core and virtual orbitals from the quantum chemical calculations underlying our data set, we demonstrate that it is possible to relate the atomic contributions via these orbitals to the XAS spectrum.

SELENOF Controls Proliferation and Cell Death in Breast-Derived Immortalized and Cancer Cells.

SELENOF expression is significantly lower in aggressive breast tumors compared to normal tissue, indicating that its reduction or loss may drive breast tumorigenesis. Deletion of SELENOF in non-tumorigenic immortalized breast epithelial MCF-10A cells resulted in enhanced proliferation, both in adherent culture and matrix-assisted three-dimmensional (3D) growth. Modulation of SELENOF in vitro through deletion or overexpression corresponded to changes in the cell-cycle regulators p21 and p27, which is consistent with breast tumor expression data from the METABRIC patient database. Together, these findings indicate that SELENOF affects both proliferation and cell death in normal epithelial and breast cancer cells, largely through the regulation of p21 and p27. In glandular cancers like breast cancer, the filling of luminal space is one of the hallmarks of early tumorigenesis. Loss of SELENOF abrogated apoptosis and autophagy, which are required for the formation of hollow acini in MCF-10A cells in matrix-assisted 3D growth, resulting in luminal filling. Conversely, overexpression of SELENOF induced cell death via apoptosis and autophagy. In conclusion, these findings are consistent with the notion that SELENOF is a breast tumor suppressor, and its loss contributes to breast cancer etiology.

A low cost, low power sap flux device for distributed and intensive monitoring of tree transpiration.

Accurate estimation of transpiration in individual trees is important for understanding plant responses to environmental drivers, closing the water balance in forest stands and catchments, and calibrating earth system models, among other applications. However, the cost and power consumption of commercial systems based on sap flow methods still limit their usage. We developed and tested a cost-effective (<$150), simple to construct, and energy efficient sap flux device based on the heat pulse method. Energy savings were achieved by reducing the voltage of heat pulses and using an internal clock to completely shut down the device between pulses. Device accuracy was confirmed by laboratory estimates of sap flow made on excised branches of Acer saccharum and Tsuga canadensis (adjusted R2 = 0.96). In a 174-d field installation of 12 devices, batteries (eight rechargeable Ni-MH AA) needed to be replaced every 14 days. Sap flux measurements in the field tracked expected variations in vapor pressure deficit and tree phenology. The low cost, compact design, reliability, and power consumption of this device enable sap flux studies to operate with more replication and in more diverse ecological settings than has been practical in the past.

Graph Neural Networks for Learning Molecular Excitation Spectra.

Machine learning (ML) approaches have demonstrated the ability to predict molecular spectra at a fraction of the computational cost of traditional theoretical chemistry methods while maintaining high accuracy. Graph neural networks (GNNs) are particularly promising in this regard, but different types of GNNs have not yet been systematically compared. In this work, we benchmark and analyze five different GNNs for the prediction of excitation spectra from the QM9 dataset of organic molecules. We compare the GNN performance in the obvious runtime measurements, prediction accuracy, and analysis of outliers in the test set. Moreover, through TMAP clustering and statistical analysis, we are able to highlight clear hotspots of high prediction errors as well as optimal spectra prediction for molecules with certain functional groups. This in-depth benchmarking and subsequent analysis protocol lays down a recipe for comparing different ML methods and evaluating dataset quality.

Roadside ditch macroplastic and other litter dataset in the Finger lakes region across land uses and COVID-19 pandemic.

Litter was collected from 12 roadside ditches in the Finger lakes Region of New York State over two sampling periods: pre-COVID-19 pandemic and during COVID-19 pandemic. Litter pieces were washed in DI water, oven dried, massed and plastic-type visually determined. Macroplastic data was analysed to assess the impact of land use, traffic, and COVID-19 variables on macroplastic accumulation on a piece, mass, and plastic-type basis. These data are all litter pieces collected, including both plastics categorized as 1 through 7 in the RIC resin classification codes as well as non-plastic litter. These data have wide-ranging reuse potential, as terrestrial microplastic accumulation is not well documented. These data could be compared with other litter accumulation across regions, specifically to assess total environmental macroplastic loading and enable contaminant mitigation strategies. These data also have direct application to modelling and transport of macroplastics into surface water bodies as a result of road ditch sampling locations. Macroplastic accumulation data across varying land uses, traffic, and COVID-19 conditions has been published [1].

Macroplastic accumulation in roadside ditches of New York State's Finger Lakes region (USA) across land uses and the COVID-19 pandemic.

Macroplastics are a ubiquitous and growing environmental contaminant with impacts in both marine and terrestrial systems. Marine sampling has dominated research in this field, despite the terrestrial origins of most plastic debris. Due to the high surface water connectivity facilitated by roadside ditches, these landscape features provide a unique sampling location linking terrestrial and surface water systems. We collected and analyzed macroplastic accumulation by number of pieces, mass, and polymer type in roadside ditches across four land uses, before and during the COVID-19 pandemic in the Finger Lakes Region of New York State. Commercial land use plastic accumulation rate was highest, while forested land use accumulation rates were lowest on a piece basis. Pre-COVID-19 piece accumulation rates were significantly higher than COVID-19 piece accumulation rates across all land uses. Mass accumulation rates followed similar patterns observed in piece accumulation, but the patterns were not always statistically significant. Plastic type 4 (i.e. thin plastic films), especially plastic bags and wrappers, was the most frequently collected type of macroplastic by piece across all land uses within the 1-7 Resin Identification Codes. By mass, the data were distributed less consistently across land uses. Cigarette filters, containing the polymer cellulose acetate, were the most frequently found roadside plastic, but are not within the 1-7 classification system. Our results suggest that policies in place limiting plastic bag usage could substantially reduce roadside plastics but other plastics, such as food wrappers and other single use plastic films, which comprised a large proportion of the plastic debris collected, should also be regulated to further decrease macroplastic pollution.

A Self-Propelled Mechanism to Increase Range of Bistable Operation of a Piezoelectric Cantilever-Based Vibration Energy Harvester.

In our previous works, we presented a method to increase the harvested energy from vibrations using a piezoelectric cantilever and to increase the frequency range of operation by introducing bistability with the use of magnetic repulsion. However, for small excitations, the cantilever may not be able to overcome the magnetic repulsive force but vibrate at one of its two equilibrium positions with reduced amplitude. This work introduces a method of increasing the range of excitations over which the operation remains bistable. This is achieved by spring loading one of the magnets, previously on a fixed support, allowing motion in one dimension only, toward and away from the cantilever in the horizontal plane. Configured so, as the cantilever moves toward this magnet, the repulsion due to the cantilever-mounted magnet pushes the spring-loaded magnet away, increasing distance, and thus, reducing magnetic force between them, required to be overcome by external excitations for bistable operation. Similarly, as the cantilever moves away, the spring pushes the magnet closer to the cantilever-mounted magnet, increasing vibration amplitude. Thus, the spring introduces a negative feedback which favors bistable operation over an increased range of excitations. This completely mechanical method requires no additional energy cost. Peak power gains of up to 90 and a decrease in excitation voltage of up to 60% were observed over the fixed magnet.