Retrospective Reanalysis Framework

[MattWellie]

Run AIP repeatedly in a way which simulates incremental analyses over time.

Start: 01-01-2020 Increments: 3 Months

Static: Joint Call, Consequence Annotation, AIP version, Cohort Participants, HPO Metadata, MOI-per-gene (latest data) Changing: Gene List (reflect the PanelApp state at each time point, see this), ClinVar data (reflect the latest ClinVar submissions up to each time point, see this)

This will run only on the Acute-Care dataset.

Exome & Genome 'Types' run in parallel.

Familial and Singleton report generated for each Type & time point (4 reports per time point)

No hard filtering of previously seen variants, data will be presented in the report with dates, for interactive filtering.

---

N.B. these ideas were considered, but ultimately were not actioned.

• Do we want to alter the Clinvar (Cat. 1) exemptions for various population frequency filters
• Do we want to recognise ClinVar 3/4 (Expert Review & Practice guidelines) over + above the current 1* for Cat. 1
• Do we want to enable a variant whitelist (known common pathogenics) so that we can enforce stricter filtering more broadly, without worrying about missing obvious variants
• Can we wrap some of this in mild automation/scripting so we don't need to manually trigger and wait for 40 AIP runs

[MattWellie]

For context relating to the ClinVar Gold Stars:

Stars	All Subs	Pathogenic only
Any	1.03M	186k
1	1.02M	178k
3	12570	8673
4	13	13

These results relate to the privately re-summarised data, excluding all submissions from MCRI/VCGS

[MattWellie]

Index file linking to the reports created based on this battle plan https://test-web.populationgenomics.org.au/acute-care/retro_analysis/retro_index.html