You are viewing v1. See latest version (v2) →
TrajectoryInferenceEngine: Pseudotime Ordering, RNA Velocity, Cell Fate Probability, and Lineage Branch Point Detection
Trajectory inference methods reconstruct developmental and differentiation trajectories from single-cell RNA-seq data. We present TrajectoryInferenceEngine, a pure-Python pipeline for trajectory analysis. The engine implements pseudotime ordering (diffusion map + principal curve), RNA velocity simulation (spliced/unspliced ratio dynamics), cell fate probability (Markov chain transition matrix), branch point detection (entropy peak), and gene expression dynamics along pseudotime. Applied to 2000 cells × 500 genes with 3 lineages, the pipeline identifies branch point pseudotime=0.939, max entropy=1.099, spliced/unspliced ratio=1.033, and reconstructs stem→progenitor→terminal A/B/C trajectories.
Introduction
Single-cell RNA-seq enables reconstruction of developmental trajectories. Pseudotime orders cells along a developmental axis. RNA velocity uses spliced/unspliced mRNA ratios to infer transcriptional change direction. Cell fate probabilities quantify likelihood of reaching each terminal state.
Methods
Pseudotime
Diffusion map from cell-cell similarity. Principal curve fitted to first two diffusion components.
RNA Velocity
Velocity = ds/dt = β×u - γ×s, where u=unspliced, s=spliced.
Cell Fate
Markov chain T_ij = exp(-||v_i - (x_j - x_i)||²/σ²). Fate = absorption probability.
Results
Branch point=0.939. Max entropy=1.099. Spliced/unspliced=1.033. 3 lineages.
Code Availability
Discussion (0)
to join the discussion.
No comments yet. Be the first to discuss this paper.