GeneTrajectory is a method for inferring gene trajectories in single-cell data using optimal transport metrics. It identifies gene-level trajectories rather than cell-level trajectories, enabling the extraction of gene programs and their pseudotemporal order. This approach is particularly effective in resolving gene dynamics in complex biological processes where multiple concurrent processes occur. The method calculates optimal transport distances between gene distributions across a cell graph to define gene trajectories, allowing for the deconvolution of gene programs that are otherwise obscured by cell pseudotime approaches. In simulations and real-world applications, GeneTrajectory accurately captures gene dynamics in processes such as myeloid lineage maturation and dermal condensate genesis. It outperforms traditional cell trajectory methods in recovering gene order in both cyclic and linear processes, demonstrating robustness to variations in cell numbers and data sparsity. GeneTrajectory can identify gene trajectories without requiring prior knowledge of cell states, making it applicable to scenarios where cells do not form clear lineages. The method was tested against five cell trajectory methods, including Monocle2, Monocle3, Slingshot, PAGA, and CellRank, showing superior performance in recovering gene order in concurrent processes. It also successfully deconvolves gene processes in the Wntless mutant, revealing defects in DC differentiation. GeneTrajectory provides a framework for understanding the transcriptional dynamics of biological processes by identifying gene trajectories that reflect the progression of gene activity. It can be applied to various single-cell modalities, including scATAC-seq and spatial transcriptomics, and has potential applications in integrating multi-omics data. The method's ability to detect and disentangle multiple gene programs makes it a valuable tool for analyzing complex biological systems.GeneTrajectory is a method for inferring gene trajectories in single-cell data using optimal transport metrics. It identifies gene-level trajectories rather than cell-level trajectories, enabling the extraction of gene programs and their pseudotemporal order. This approach is particularly effective in resolving gene dynamics in complex biological processes where multiple concurrent processes occur. The method calculates optimal transport distances between gene distributions across a cell graph to define gene trajectories, allowing for the deconvolution of gene programs that are otherwise obscured by cell pseudotime approaches. In simulations and real-world applications, GeneTrajectory accurately captures gene dynamics in processes such as myeloid lineage maturation and dermal condensate genesis. It outperforms traditional cell trajectory methods in recovering gene order in both cyclic and linear processes, demonstrating robustness to variations in cell numbers and data sparsity. GeneTrajectory can identify gene trajectories without requiring prior knowledge of cell states, making it applicable to scenarios where cells do not form clear lineages. The method was tested against five cell trajectory methods, including Monocle2, Monocle3, Slingshot, PAGA, and CellRank, showing superior performance in recovering gene order in concurrent processes. It also successfully deconvolves gene processes in the Wntless mutant, revealing defects in DC differentiation. GeneTrajectory provides a framework for understanding the transcriptional dynamics of biological processes by identifying gene trajectories that reflect the progression of gene activity. It can be applied to various single-cell modalities, including scATAC-seq and spatial transcriptomics, and has potential applications in integrating multi-omics data. The method's ability to detect and disentangle multiple gene programs makes it a valuable tool for analyzing complex biological systems.