scMetaTraj maps metabolic shifts in single cells

Researchers have developed scMetaTraj, an R package that reconstructs single-cell trajectories from metabolic-program activity rather than whole-transcriptome similarity. The peer-reviewed Correspondence, published in Med Research, describes how the framework builds a metabolic state space, identifies metabolic subclusters, and orders cells along metabolic pseudotime (mPT).

The author team includes Jinwei Dai and corresponding author Zhaoxin Qian at Xiangya Hospital, Central South University, with collaborators from the National Clinical Research Center of Geriatric Diseases (Xiangya Hospital), Xiangya Hospital's Department of Spine Surgery, and University Medical Center Groningen.

Single-cell analyses commonly define cell identities and lineages from transcriptomic profiles and examine metabolism afterward. That design can make gradual metabolic changes harder to resolve when cells with similar transcriptomic identities occupy different metabolic states. scMetaTraj instead uses scores for curated metabolic pathways and modules as the primary analysis space. It is intended to complement, not replace, transcriptome-centered trajectory methods.

The software accepts Seurat objects or gene-by-cell expression matrices. It calculates module scores from log-normalized expression, performs principal component analysis on the cell-by-module matrix, constructs a weighted k-nearest-neighbor graph, and identifies subclusters by Louvain community detection. mPT is then calculated as a rescaled Dijkstra shortest-path distance from a selected root cell. UMAP is used only to display the inferred structure.

Because metabolic programs may change gradually during immune activation, stress adaptation, tissue repair, and tumor development, a continuous ordering can help researchers locate regions where several programs change together. scMetaTraj provides metabolic state maps, subcluster profiles, module trends, and candidate transition regions. These outputs are descriptive and hypothesis-generating; they do not by themselves establish a biological lineage or causal direction.

In the 10x PBMC 3k reference dataset, 2,700 cells remained after preprocessing. scMetaTraj identified 10 metabolic subclusters across a continuous manifold. Transcriptome-derived labels were distributed across multiple metabolic neighborhoods rather than mapping one-to-one onto the metabolic clusters. This distribution supports the interpretation that metabolic organization was not identical to canonical PBMC identity.

The authors then analyzed an independent dataset, GSE226488, containing 14,771 PBMCs from resting and 4-hour lipopolysaccharide-stimulated conditions. In the monocyte-lineage subset, median mPT increased from 0.530 in resting cells to 0.661 after stimulation (Wilcoxon P = 0.0006). Individual modules, including glycolysis, oxidative phosphorylation, the tricarboxylic acid cycle, and lactate metabolism, showed directionally consistent but modest mean shifts. The authors interpret the mPT result as a multivariate redistribution across the metabolic manifold rather than a change driven by one dominant pathway.

Monocle 3 pseudotime showed weak rank concordance with scMetaTraj mPT (Spearman's rho = -0.14) in this subset. The paper presents this as an initial reference comparison, not a systematic benchmark, because the two methods organize cells in different feature spaces.

The authors caution that mPT is inferred from transcript-derived module scores and does not directly measure metabolite abundance, enzyme activity, or metabolic flux. Detailed ordering can also depend on the selected root cell, and external validation currently includes one perturbation dataset. scMetaTraj is therefore positioned as an exploratory framework for describing metabolic-state organization and generating hypotheses for validation with metabolomic, proteomic, or flux-based measurements.

The article was first published online on 10 May 2026. The R package, gene sets, and supporting materials are available through the project resources linked in the article.

Expert Contact: Zhaoxin Qian, Department of Critical Care Medicine, Xiangya Hospital, Central South University, Changsha, Hunan, China; Email: xyqzx@csu.edu.cn.

Custom Tweet: scMetaTraj maps transcript-derived metabolic programs along pseudotime in single-cell data.