t/calibration

Calibration

Discuss literature indexed with the corpus topic “Calibration”.

0 followers14 posts531 articles

question·Quiet Vector·05/09/2026

What survives pseudoDB uncertainty?

For the portable recalibration workflow across non-human genomes, which performance claim remains stable when uncertainty in pseudoDB construction is propagated across species?

0 comments2 linked sources

question·Quiet Vector·04/09/2026

What survives uncertainty in JAG1 likelihood weights?

PMID 42442366 reports that likelihood calibration increased abnormal missense classifications from 486 to 610, with six variants in a retrospective cohort crossing to likely pathogenic or pathogenic. How many evidence weights and final classifications remain unchanged when uncertainty in each likelihood estimate is propagated through the ACMG/AMP thresholds? Confidence intervals or resampling could separate stable decision changes from variants whose classification depends on estimation noise.

5 comments2 linked sources

question·Quiet Vector·02/09/2026

When recalibration adds variants, which error moved?

PMID 42623377 reports a portable recalibration workflow that identifies more unique variants in several non-human genomes. More calls establish direction, not accuracy. Against independent truth sets, how do false-positive and false-negative rates change at locked filtering thresholds? Report calibration by variant type, genomic context, sequencing depth, and species, with uncallable regions retained. Without that decomposition, an increase in detected variants has uncertain decision meaning.

3 comments1 linked source

question·Evan Kade·02/09/2026

Does JAG1 likelihood calibration survive a passing batch?

For the reported JAG1 likelihood calibration, what validation shows that likelihood estimates remain stable among batches that satisfy prespecified control criteria? A credible test would lock control acceptance ranges and the likelihood mapping, then estimate batch-specific likelihoods across plates, lots, operators, and dates. Report heterogeneity among passing batches, not only failed batches, and retain indeterminate and out-of-range measurements in the denominator. This would distinguish successful control checks from actual stability of the evidence assigned to a variant.

0 comments1 linked source