Global Property Prediction: A Benchmark Study on Open-Source, Perovskite-like Datasets.

ACS Omega

Department of Electrical and Computer Engineering, Technical University of Munich, Munich 80333, Germany.

Published: May 2021


Category Ranking

98%

Total Visits

921

Avg Visit Duration

2 minutes

Citations

20

Article Abstract

Screening combinatorial space for novel materials, such as perovskite-like ones for photovoltaics, has resulted in a high amount of simulated high-throughput data and analysis thereof. This study proposes a comprehensive comparison of structural fingerprint-based machine learning models on seven open-source databases of perovskite-like materials to predict band gaps and energies. It shows that none of the given methods, including graph neural networks, are able to capture arbitrary databases evenly, while underlining that commonly used metrics are highly database-dependent in typical workflows. In addition, the applicability of variance selection and autoencoders to significantly reduce fingerprint size indicates that models built with common fingerprints only rely on a submanifold of the available fingerprint space.

Download full-text PDF

Source
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC8154242PMC
http://dx.doi.org/10.1021/acsomega.1c00991DOI Listing

Publication Analysis

Top Keywords

global property
4
property prediction
4
prediction benchmark
4
benchmark study
4
study open-source
4
open-source perovskite-like
4
perovskite-like datasets
4
datasets screening
4
screening combinatorial
4
combinatorial space
4

Similar Publications

Salmonella enterica serovar Typhi, the etiological agent of Typhoid fever, remains a critical public health concern associated with high morbidity in many developing countries. The widespread emergence of multidrug-resistant (MDR) Salmonella Typhi strains against the fluoroquinolone group of antibiotics, particularly ciprofloxacin, poses a significant global therapeutic challenge with underlying resistance due to mutations in quinolone-resistance determining region (QRDR) of gyrA gene, encoding DNA gyrase subunit A (GyrA). In pursuit of alternative therapeutic candidates, the present study was designed to evaluate ciprofloxacin analogues against prevalent GyrA mutations (S83F, D87G, and D87N) to overcome fluoroquinolone resistance through machine learning (ML)-based approach.

View Article and Find Full Text PDF

Ginger, a globally cultivated spice and medicinal herb, is renowned for its health benefits and distinctive flavor. As ginger's main pungent and bioactive components, 6-gingerol and 6-shogaol share similar physicochemical properties and can be obtained by extraction from ginger or chemical synthesis. After oral ingestion, the biological fate of 6-gingerol and 6-shogaol are influenced by processes including absorption, biotransformation, distribution, and excretion.

View Article and Find Full Text PDF

Chronic liver disease (CLD) is a leading cause of global morbidity and mortality, necessitating effective preventive strategies. Growing evidence is linking coffee consumption with reduced risk of disease progression in various CLDs, including metabolic dysfunction associated steatotic liver disease (MASLD), alcoholic liver disease, hepatitis B and C, autoimmune hepatitis, and a reduction in the risk of hepatocellular carcinoma development. Coffee, a globally consumed beverage, contains bioactive compounds like caffeine, chlorogenic acids, diterpenes, and polyphenols, which may offer hepatoprotective benefits through anti-inflammatory, antioxidant, and metabolic regulatory effects.

View Article and Find Full Text PDF

Global challenges posed by freshwater scarcity and the water-energy nexus drive demand for novel macromolecular design of tailored nanostructures endowed with a variety of hydrophilic and hydrophobic features. Offering potential to meet this demand, metal-organic framework (MOF) materials are synthesized from coordinated formations that create versatile reticular structures with variable water adsorption affinities. However, advances in the fundamental understanding of water interactions within these structures are impeded by the failure of classical analyses to identify mechanisms of interaction, connect fundamental isotherm types, and provide appropriate benchmarks for assessment.

View Article and Find Full Text PDF

Buildings are increasingly being conceived as dynamic systems that interact with their surroundings to optimize energy performance and enhance occupant comfort. This evolution in architectural thinking draws inspiration from biological systems, where the building envelope functions like a thermally responsive "skin" that can autonomously adjust its optical and thermal properties in response to environmental temperature changes. Among the many approaches developed for smart building envelopes, passive thermoresponsive spectral modulation systems have attracted growing interest due to their structural simplicity and low energy demand.

View Article and Find Full Text PDF