A tool for inferring TRAnscriptomic Cellular States (TRACS) by time-series clustering

Three main advantages of TRACS in each step of clustering analysis

  • Selecting the number of clusters TRACS automatically determines the optimal number of clusters with adapted gap statistics that leverage time point information and consider time-to-time dependency (Gaussian process)
  • Clustering time series By selecting the optimal cluster number, TRACS predicts clusters of genes based on temporal patterns of genes adjusted by Gaussian process regression. We will show that gene clusters correspond to cell states in our experiments.
  • Analysis of the clustering result TRACS infers the transition of cell states as a cluster network, predicting the order of clusters by their pattern similarity (Shape-Based Distance and Ranked Pairs algorithm) and edge labels by functional similarity (two-group enrichment test)
  • TiClNet News

  • 2020/01/28 : TRACS uploaded to Github
  • Publication

    Jo K, Lee D, Sung IY, Kim S. Inferring transcriptomic cell states and transitions from time series transcriptome data.