
Score a target's safety liability for inhibition or knockout.
Source:R/targetLiability.R
targetLiability.RdtargetLiability() is a convenience wrapper that combines the
evidence returned by geneticConstraintQuery,
safetyQuery and (optionally) depMapQuery into a
single, interpretable liability score with a written rationale. It is
intended as a quick triage step when prioritising drug targets: genes that
are highly constrained, essential in cell lines, or already associated with
safety events are flagged as risky for full inhibition, while genes lacking
these signals may tolerate a partial or tissue-restricted approach.
Value
A one-row tibble with the input identifiers, the individual
component scores and flags, the combined liability_score, a
liability_category ("Low", "Moderate" or "High"), a
recommendation, and a human-readable rationale. Returns
NULL if none of the underlying queries return data.
Details
The overall liability_score is a weighted average (range 0-1, higher
means greater liability) of up to three normalised components: genetic
constraint (weight 0.40), known safety liabilities (0.35) and DepMap
essentiality (0.25). Components without data are dropped and the remaining
weights are renormalised. See Details for the mapping of each
component.
Genetic constraint uses the loss-of-function (LoF) decile bin from
geneticConstraintQuery() (gnomAD LOEUF, where bin 0 is the most
constrained and bin 9 the least). The most constrained bin scores 1 and the
least constrained scores 0.
Safety liabilities count the distinct events reported by
safetyQuery(); the component saturates at five or more events.
Essentiality uses the median CRISPR gene-effect score across DepMap screens; a score of 0 (non-essential) maps to 0 and -1 (strongly essential) maps to 1.
This is a heuristic aid, not a validated clinical safety measure; always review the underlying evidence before acting on the score.