E. Aladyeva

TL;DR: It is reported that about 14% CR for 2 years in healthy humans improved thymopoiesis and was correlated with mobilization of intrathymic ectopic lipid andexpression of the gene encoding platelet activating factor acetylhydrolase (PLA2G7) was decreased in humans undergoing caloric restriction.

TL;DR: Novel insights into healthy human aging are provided and a concerted age-associated increase in type 2/interleukin (IL)4-expressing memory subpopulations across CD4+ and CD8+ T cell compartments is found, suggesting a systematic functional shift in immune homeostasis with age.

TL;DR: A single-cell atlas of healthy human blood reveals age-related changes in immune cell populations, including a loss of NKG2C+GZMB-CD8+ memory T cells and accumulation of type 2 memory T cells, across 12 subpopulations between 25-85 years.

Abstract: Non-negative matrix factorization is a powerful linear algebra tool used in multiple areas of data analysis, including computational biology. Despite numerous optimization methods devised for non-negative matrix factorization, our understanding of the inherent topological structure within factorizable matrices remains limited. This study reveals the topological properties of linear mixture data, leading to a remarkable reduction of the non-negative matrix factorization optimization problem to a search for K(K-1) variables, where K represents the number of pure components, regardless of the initial matrix size. This is achieved by revealing complementary simplex structures existing in both feature and sample spaces and leveraging the Sinkhorn transformation to find the relationship between these simplexes. We validate this approach in the context of an unconstrained mixed images scenario and achieve a significant improvement in decomposition accuracy. Furthermore, we successfully applied the proposed approach in the biological context of bulk RNA-seq gene expression deconvolution. The Dual Simplex unified analytical framework improves robustness to noise and enhances optimization stability, enabling accurate recovery of component proportions and expression profiles. Importantly, the framework naturally accommodates both reference-free and marker-based deconvolution settings, providing a general and efficient solution for analyzing complex biological mixtures such as bulk RNA-seq and single-cell derived data.