Release notes: Major release
- pangolin 3.0 comes with additional functionality and requires an environment update as extra dependencies now include UShER and scorpio.
- Requires pangoLEARN data >= 2021-05-27
Lineage assignment updates
- PANGO assignment uses a sequence hash from all currently designated sequences to assign lineages.
- PLEARN (pangoLEARN) assignment using a machine-learning model to assign the most likely lineage. Current model is decision tree model.
- PUSHER (pangolin-UShER, pangUSHlin, pangUShER) assignment uses fast parsimony placement of a query sequence into a protobuf file based on currently designated sequences and infers most pasimonious lineage based on this placement (thanks to @AngieHinrichs and the rest of the UShER team).
- Default
--max-ambig
value changed to 0.3.
Constellation assignment
- **NEW FEATURE** pangolin now runs scorpio a tool by @rmcolq and @benjamincjackson to assign the most likely constellation to a given sequence. Full set of possible constellations assigned can be found at: https://github.com/cov-lineages/constellations/tree/main/constellations/definitions
- This replaces the post-hoc tests pangolin previously ran to curate the variant of concern calls. The output of scorpio, which is explict SNP-threshold based assignment, and pangolin, which can now use either the pangoLEARN or UShER infererence engines, are displayed side-by-side in the output report.
Input updates
- Based on @ArtPoon's PR, pangolin is now gzip and lzma compatible.
Output updates
- NOTE: Output file columns have changed, details can be found at https://cov-lineages.org/pangolin_docs/output.html.