Zhen Fang

We demonstrate that single-atom catalysts can be rationally engineered to couple the electrocatalytic oxygen reduction reaction with a catalase-like process, enabling a favorable H ₂ O ₂ to-O ₂ conversion pathway.

Protein subcellular localization prediction is critical for understanding protein function, disease mechanisms, and drug target discovery. Existing computational methods face significant limitations: traditional approaches rely on hand-crafted features and cannot capture long-range dependencies, while recent deep learning methods either use single-scale convolutional filters that fail to capture the multi-scale nature of sorting signals or depend on large pre-trained protein language models with high computational costs. We propose MSA-Loc (Multi-Scale Attention Network for Subcellular Localization), a lightweight framework designed around the hypothesis that protein sorting signals occur at heterogeneous length scales and require adaptive scale selection with sequence-level contextual modeling. MSA-Loc employs parallel convolutional branches with kernel sizes {3, 5, 9, 15, 21} to extract local patterns at different scales, bidirectional LSTMs to capture long-range contextual dependencies, and multi-head self-attention to automatically identify critical sorting signals at the residue level. Evaluated on the DeepLoc 1.0 dataset containing 13,858 single-label protein sequences across 10 subcellular locations, MSA-Loc achieves an accuracy of 0.6325, macro-F1 of 0.4908, and MCC of 0.5580, outperforming seven baseline methods including traditional machine learning approaches, single-scale deep learning models, and attention-based architectures. With only 6.2M parameters, MSA-Loc provides an efficient and interpretable alternative when computational resources are constrained.

Electro‐oxidation (EO) is a rapid and effective method for the elimination of organic and inorganic pollutants from wastewater. In this study, the titanium‐based mesh electrodes Ti/TiO ₂ –RuO ₂ were used for the removal of rhodamine B (RhB) azo dye in simulated water by the treated EO and EO‐treated biodegradation (EO and BD) with 20 mA/cm ⁻² direct current applied for the EO system, and the color removal was measured by UV‐spectrophotometer. The chemical oxygen demand (COD) was greatly reduced to 78% in the EO‐treated and 93% in the EO‐treated biodegradation sample, respectively, and the decolorization was accomplished in 30 min in 10 ppm. The biodegradation data verify that the bacterial strain used both organic and inorganic materials. TOC (total organic carbon) in EO and EO with biodegradation were 57.2% and 86.7%, respectively. It confirms that the EO with biodegradation effectively removes RhB and aids in complete mineralization. Fourier transform infrared spectroscopy (FTIR) confirms the functional groups cleaved during the process HRMS concluded that the intermediate products NH, C–N, and C=O were generated during EO with biodegradation. The calculated toxicity of LC50 and EC50 values of RhB was observed in RhB 6 (100<) in water for fish, daphnid, and green algae. The toxicity study confirms that the majority of intermediates from EO and BD were scarcely harmful. Overall, this study concludes that RhB was effectively degraded by the EO and mineralized by the BD, and this combined treatment plays a vital role in the remediation of pollutants in the wastewater effluent treatment plant.

Bioelectrochemical sensors (BES) are promising to specifically detect nitrate or nitrite but never realize simultaneous detection in a single system. In this study, a novel biosensor using self-assembled hydrogel bioelectrodes with reduced graphene oxide (rGO) and genetically engineered electroactive Shewanella species was designed for the simultaneous detection of nitrate and nitrite. The highly conductive rGO rendered a drastically improved reverse electron transfer from the electrode to Shewanella, which achieved a sensitive and quantitative response to nitrite/nitrate. Meanwhile, the genetically engineered Shewanella enabled efficient differentiation of nitrate and nitrite detection in a single system (a sensitivity of 883.48 μA mM^-1 cm^-2 to nitrate with a limit of detection of 0.92 μM and a sensitivity of 888.48 μA mM^-1 cm^-2 to nitrite with a limit of detection of 0.72 μM). The whole-cell biohydrogel based BES also showed excellent anti-interference and long-term storability for plug-and-play application. Based on these properties, this work demonstrated the power of genetically engineered electroactive bacteria in rGO biohydrogel for dual-analyte detection and also provided a new strategy for developing high-performance whole-cell BES for practical environmental monitoring.