Comparison of RNA velocity methods for droplet scRNA-seq data

Method ID Agreement Strand-aware Counting of multi-gene reads Comments
alevin_sep_gtr good yes yes
alevin_coll_gtr good yes yes
kallisto|bus_sep_excl good no no Reads in ambiguous regions typically discarded
starsolo good yes no
alevin_coll_decoy_gtr variable yes yes
velocyto variable yes no
kallisto|bus_coll_excl variable no no
alevin_sep_decoy_gtr poor yes yes Reads in ambiguous regions double-counted.
alevin_spliced_unspliced_gtr variable yes yes 3’ tag protocols does not provide enough information for spliced/unspliced resolution.
dropest poor no partially Insufficient UMI collapsing
kallisto|bus_coll_incl poor no no
kallisto|bus_sep_incl poor no no Reads in ambiguous regions double-counted.
starsolo_diff variable yes partially Non-zero unspliced count possible even without introns; nominally negative unspliced count possible.

Reference: Soneson C, Srivastava A, Patro R, Stadler MB. Preprocessing choices affect RNA velocity results for droplet scRNA-seq data. PLoS Comput Biol. 2021 Jan 11;17(1):e1008585. doi: 10.1371/journal.pcbi.1008585. PMID: 33428615; PMCID: PMC7822509.