2.6. RNA Velocity

The RNA velocity analysis of naïve and inflamed samples was performed as previously described [32]. In a first step, spliced and unspliced counts were calculated from sorted aligned bam files of the 559 single cells using the velocyto.py run-smartseq2 function with default settings. From here, we followed the velocyto.R manual (http://velocyto.org/, accessed on 30 April 2021) and used spliced (emat) and unspliced (nmat) to estimate the RNA velocity of naïve and inflamed data combined. We restricted our analysis to protein-coding genes to avoid nuclear or cytoplasmic RNA retention biases. Using the cell-type annotation defined in Section 2.5, we performed gene filtering using the min.max.cluster.average parameter set to 5 and 1 for emat and nmat, respectively. RNA velocity was then estimated with the function gene.relative.velocity.estimates with default parameters except for kCells = 5 and fit.quantile = 0.5. To select differentially “speeded” (DS) genes, we first calculated the RNA velocity values using the formula

where β is equal to 1, γg is the estimated gamma coefficient ($gamma), and Sg(tc) and Ug(tc) are the normalized emat ($conv.emat.norm) and nmat ($conv.nmat.norm) values of a gene g in a cell tc, respectively. After plotting the mean RNA velocity per gene between naïve and inflamed conditions, we calculated the perpendicular distance between each point and the slope = 1. Finally, we selected DS genes having a slope distance higher than 0.1. Gene ontology for biological processes and pathway analyses of DS were performed as described above, using protein-coding genes expressed in both conditions as background.