Developmental trajectory analysis using Palantir

Palantir²⁸ is a high-resolution algorithm, which allows computing a continuous probabilistic process to model cell fate choice by applying multiple diffusion components. Here, Palantir was applied to NK subsets in our data and to the integrated dataset. Basically, we used the CCA-aligned subspaces generated from Seurat to replace the low-dimensional principal components subspaces to reduce the batch effects. Twenty diffusion components were selected and computed with default parameters in Palantir. Diffusion components scaled by an Eigen gap were used as inputs and perplexity was set to 200 to generate the t-SNE maps. To accurately define the initial cell state, we imputed a pseudo cell as the start cell by calculating the average gene expression of our identified dNKp cells. A waypoints = 1200 value was applied, and the parameter k was set to 50 for datasets.