MAY 2019 | Wouter Saelens, Robrecht Cannoodt, Helena Todorov, Yvan Saeys
This study provides a comprehensive evaluation of 45 trajectory inference (TI) methods, which are used to analyze single-cell omics data and infer developmental trajectories. The evaluation covers 110 real and 229 synthetic datasets, assessing the methods' performance on cellular ordering, topology, scalability, and usability. The results highlight the complementarity of existing tools and suggest that the choice of method should depend on the dataset dimensions and trajectory topology. The authors develop practical guidelines to help users select the most suitable method for their specific dataset. They also emphasize the need for standardized input and output interfaces to enhance the broad applicability of TI methods. The study identifies several ongoing challenges, such as improving the inference of complex topologies and enhancing code assurance and documentation. The freely available data and evaluation pipeline will aid in the development of improved tools for analyzing large and complex single-cell datasets.This study provides a comprehensive evaluation of 45 trajectory inference (TI) methods, which are used to analyze single-cell omics data and infer developmental trajectories. The evaluation covers 110 real and 229 synthetic datasets, assessing the methods' performance on cellular ordering, topology, scalability, and usability. The results highlight the complementarity of existing tools and suggest that the choice of method should depend on the dataset dimensions and trajectory topology. The authors develop practical guidelines to help users select the most suitable method for their specific dataset. They also emphasize the need for standardized input and output interfaces to enhance the broad applicability of TI methods. The study identifies several ongoing challenges, such as improving the inference of complex topologies and enhancing code assurance and documentation. The freely available data and evaluation pipeline will aid in the development of improved tools for analyzing large and complex single-cell datasets.