Viridian 0.9 checklist

[iqbal-lab]

Unchanged

• command line invocation , main output files (consensus, VCF, json)

New features that people are likely to care about:

1. Coverage on SNP (not indel) alleles reported in VCF
2. A new summary block added to the output JSON which should include everything anyone needs to look at results and QC (separating this from internal/debug output in the JSON). Intention is users should generally only need to look here.
3. Output JSON records state of how far through the process the workflow has gone.

Subtle changes in quality of assembly sequence

1. Slightly fewer Ns called in primers due to improved primer-aware QC
2. Consensus will include the very first and last primers in the genome
3. Lengths of runs of Ns now fixed to match the length of associated sequence in the reference (ie not intended to imply indels)

Improved robustness

• Additional robustness against pathological cases with ARTIC4.1 (which we have not actually ever seen in real life)

Performance/speed

• High depth samples are subsampled to 1000x (per amplicon) thus controlling runtime and RAM. Will control runtime for the crazy-big samples in the ENA .

New debug output

• new TSV file produced which shows coverage of different bases at each position

Internal changes

• Large amount of code refactoring

Test process

1. Unit tests improved
2. Tests on simulated genomes with specific error-modes in primers
3. Tests on nanopore/illumina ARTIC/midnight truth samples using covid-truth-eval framework
4. Robustness check running on tens of thousands of genomes from ENA.