Search for: All records

Abstract Single-cell RNA-sequencing (scRNA-seq) has been widely used for disease studies, where sample batches are collected from donors under different conditions including demographic groups, disease stages, and drug treatments. It is worth noting that the differences among sample batches in such a study are a mixture of technical confounders caused by batch effect and biological variations caused by condition effect. However, current batch effect removal methods often eliminate both technical batch effect and meaningful condition effect, while perturbation prediction methods solely focus on condition effect, resulting in inaccurate gene expression predictions due to unaccounted batch effect. Here we introduce scDisInFact, a deep learning framework that models both batch effect and condition effect in scRNA-seq data. scDisInFact learns latent factors that disentangle condition effect from batch effect, enabling it to simultaneously perform three tasks: batch effect removal, condition-associated key gene detection, and perturbation prediction. We evaluate scDisInFact on both simulated and real datasets, and compare its performance with baseline methods for each task. Our results demonstrate that scDisInFact outperforms existing methods that focus on individual tasks, providing a more comprehensive and accurate approach for integrating and predicting multi-batch multi-condition single-cell RNA-sequencing data. 

Abstract Rationalizing synthetic pathways is crucial for material design and property optimization, especially for polymorphic and metastable phases. Over‐stoichiometric rocksalt (ORX) compounds, characterized by their face‐sharing configurations, are a promising group of materials with unique properties; however, their development is significantly hindered by challenges in synthesizability. Here, taking the recently identified Li superionic conductor, over‐stoichiometric rocksalt Li–In–Sn–O (o‐LISO) material as a prototypical ORX compound, the mechanisms of phase formation are systematically investigated. It is revealed that the spinel‐like phase with unconventional stoichiometry forms as coherent precipitate from the high‐temperature‐stabilized cation‐disordered rocksalt phase upon fast cooling. This process prevents direct phase decomposition and kinetically locks the system in a metastable state with the desired face‐sharing Li configurations. This insight enables us to enhance the ionic conductivity of o‐LISO to be >1 mS cm⁻¹at room temperature through low‐temperature post‐annealing. This work offers insights into the synthesis of ORX materials and highlights important opportunities in this new class of materials. 

Abstract Oxides with a face-centred cubic (fcc) anion sublattice are generally not considered as solid-state electrolytes as the structural framework is thought to be unfavourable for lithium (Li) superionic conduction. Here we demonstrate Li superionic conductivity in fcc-type oxides in which face-sharing Li configurations have been created through cation over-stoichiometry in rocksalt-type lattices via excess Li. We find that the face-sharing Li configurations create a novel spinel with unconventional stoichiometry and raise the energy of Li, thereby promoting fast Li-ion conduction. The over-stoichiometric Li–In–Sn–O compound exhibits a total Li superionic conductivity of 3.38 × 10⁻⁴ S cm⁻¹at room temperature with a low migration barrier of 255 meV. Our work unlocks the potential of designing Li superionic conductors in a prototypical structural framework with vast chemical flexibility, providing fertile ground for discovering new solid-state electrolytes. 

Single-cell RNA sequencing (scRNA-seq) data often contain doublets, where a doublet manifests as 1 cell barcode that corresponds to combined gene expression of two or more cells. Existence of doublets can lead to spurious biological interpretations. Here, we present s ingle- c ell MO del-driven D oublet D etection ( scMODD ), a model-driven algorithm to detect doublets in scRNA-seq data. ScMODD achieved similar performance compared to existing doublet detection algorithms which are primarily data-driven, showing the promise of model-driven approach for doublet detection. When implementing scMODD in simulated and real scRNA-seq data, we tested both the negative binomial (NB) model and the zero-inflated negative binomial (ZINB) model to serve as the underlying statistical model for scRNA-seq count data, and observed that incorporating zero inflation did not improve detection performance, suggesting that consideration of zero inflation is not necessary in the context of doublet detection in scRNA-seq. 

Abstract The availability of materials with high electrocaloric (EC) strengths is critical to enabling EC refrigeration in practical applications. Although large EC entropy changes, ΔS_EC, and temperature changes, ΔT_EC, have been achieved in traditional thin‐film ceramics and polymer ferroelectrics, they require the application of very high electric fields and thus their EC strengths ΔS_EC/ΔEand ΔT_EC/ΔEare too low for practical applications. Here, a fundamental thermodynamic description is developed, and extraordinarily large EC strengths of a metal‐free perovskite ferroelectric [MDABCO](NH₄)I₃(MDABCO) are predicted. The predicted EC strengths: isothermal ΔS_EC/ΔEand adiabatic ΔT_EC/ΔEfor MDABCO are 18 J m kg⁻¹K⁻¹MV⁻¹and 8.06 K m MV⁻¹, respectively, more than three times the largest reported values in BaTiO₃single crystals. These predictions strongly suggest the metal‐free ferroelectric family of materials as the best candidates among existing materials for EC applications. The present work not only presents a general approach to developing thermodynamic potential energy functions for ferroelectric materials but also suggests a family of candidate materials with potentially extremely high EC performance.