search

SACGF / variantgrid

VariantGrid public repo
Other
23 stars 2 forks source link

Gene / Disease prioritisation #442

Open davmlaw opened 3 years ago

davmlaw commented 3 years ago

See also #154 - Variant Ranking / Prioritisation Sarah raised an issue https://github.com/SACGF/variantgrid_com/issues/12

Background In some cases it is not possible to reduce variants for investigation to a reasonable number. In these cases it would be extremely helpful to be able to rank variants for investigation.

Describe the solution you'd like Suggestion 1: Implementation of an existing ranking tool, although I haven't found a suitable option as yet. Suggestion 2: In-house ranking tool, something along the lines of:

Variant rank = scaled population frequency score (0-1) x prediction score (0-1) x HPO based gene relevance score (0-1) x scaled publication score (0-1)

Where scores are calculated as:

  1. Population gnomad highest sub-pop maf frequency absent = 1, scaled to 0 at a certain % (20%?, scale based on probability score?)

    Pathogenicity prediction Maximum of: LOF no LOFTEE tag = 1 Splice = Highest score of SpliceAI. dbscSNV RF, dbscSNA Ada, MAXEntScan %diff (scaled 0-1), Missense = highest of REVEL or CADD (0-1 scale) Intron = ?? which score. Conservation? TFBS?

    Gene relevance from HPO inputs hiPHIVE or phen2gene 2

    Scaled mastermind publications (0-1)

Additional context

Potential master's project?
Only raising for discussion - testing & modelling required prior to implementation. Having said that, a rough priority score (say just taking into account pop & pathogenicity) would be extremely helpful at this stage.
davmlaw commented 3 years ago

See ClinGen CC

SKS says super hard to deal with gene specific rules now (eg table of thresholds, LOF rules etc)

ClinGen might be soon releasing webpage of this info

EmmaTudini commented 3 years ago

@davmlaw They've released the clingene gene specific rules in a repo. Can export as JSON. See: https://cspec.genome.network/cspec/ui/svi/