Journal: Developmental neurobiology

Article Title: Large-scale waves of activity in the neonatal mouse brain in vivo occur almost exclusively during sleep cycles

doi: 10.1002/dneu.22901

Figure Lengend Snippet: Non-negative Matrix Factorization of a typical record from a P1 animal. A Grayscale reference image followed by the 13 features required to capture 90% of the variance of the original data, displayed in descending order of relative prominence in sleep vs wake. Note the smaller total number of features required capture the same fraction of variance of the data compared to P6-P8 and the more even weighting of those features between sleep (8) and wake (5), reflecting the more evenly distributed complexity of activity between the two states at P1. Scale bar represents 1mm. B: Time vectors and records as in Fig. 5. C: Relationship between number of features (NMF Rank) and fraction of variance captures for P6-P8 vs P1 animals. On average, fewer features are required to capture equal variances at P1. D-F: Bivariate distributions of Feature Area vs. Sleep Score at P6-P8 (D) and P1 (E), and the Sleep-Wake difference plot (F). Development during the first postnatal week is characterized by added features with wider spread (20–40% area coverage) in sleep (Sleep Score 0.5–1.2). P-values determined using permutation test.

Article Snippet: The resulting matrix was factored into a series of paired spatial and temporal components, using the MatLab implementation of the alternating least squares algorithm for NMF.

Techniques: Activity Assay

Journal: Nature Biotechnology

Article Title: Mostly natural sequencing-by-synthesis for scRNA-seq using Ultima sequencing

doi: 10.1038/s41587-022-01452-6

Figure Lengend Snippet: ( a ) NMF applied to either Illumina, Ultima, or Illumina data with permuted genes, to extract the cell loadings, and test how well they fit either the Illumina (left) or Ultima (right) data, measured by Mean Squared Error (MSE; lower MSE is better). Dots show analysis repeated ten times with different seeds. ( b ) NMF applied to either Illumina, Ultima, or Illumina data with permuted genes, to extract the gene loadings, and test how well those loadings fit either the Illumina (left) or Ultima (right) data, measured by MSE. Dots as in ( a ). ( c ) Correlation between cell level loadings shown after applying cNMF on Illumina and Ultima data with 15 factors. ( d ) Correlation between gene level loadings shown after applying cNMF on Illumina and Ultima data with 15 factors. ( e ) Correlation between cell level loadings of both runs after applying cNMF on the same Illumina data twice with 15 factors. ( f ) Correlation between gene level loadings of both runs after applying cNMF on the same Illumina data twice with 15 factors. ( g ) Correlation between cell level loadings in the Illumina data after applying cNMF. ( h ) Correlation between gene level loadings in the Illumina data after applying cNMF. ( i ) Correlation between cell level loadings after performing cNMF on the PBMC Illumina and Perturb-Seq Illumina data with 15 factors and projecting the Perturb-Seq gene loadings onto the PBMC data to get cell loadings. ( j ) Correlation of gene level loadings after performing cNMF on PBMC Illumina and Perturb-Seq Illumina data with 15 factors. Feature plots of the cell level loading cNMF factors for Ultima ( k ) and Illumina ( l ) in the joint UMAP space. All correlations here are Pearson correlations.

Article Snippet: Because NMF runs are not identical even when re-run on the same data, to compare NMF models from Ultima and Illumina data, we fit NMF on Ultima data, Illumina data and a null of randomly permuted Illumina expression values ( ) and then measured how well cell or gene loadings fit each dataset.

Techniques:

Journal: Sheng Wu Yi Xue Gong Cheng Xue Za Zhi = Journal of Biomedical Engineering

Article Title: 单细胞数据的整合方法综述

doi: 10.7507/1001-5515.202104073

Figure Lengend Snippet: Summary of integration methods for multimodal single-cell data 单细胞多模态数据整合方法的概览

Article Snippet: Coupled NMF , 转录组+染色质可及性 , 非负矩阵分解、耦合聚类 , Matlab , [ 34 ].

Techniques: