Skip to main content

# Saturation genome editing

2 related articles

Subscribe to updates

bioRxiv Nov 9, 2026
MELD learns the wild-type-to-mutant difference directly, lifting ClinVar AUROC from 0.800 to 0.986 with 1.9 million parameters
MELD puts the score difference between wild-type and mutant sequence directly into the supervised objective, instead of training on absolute signal and comparing two sequences after the fact. On 1.57 million ClinVar variants it raised AUROC from 0.800 to 0.986 against a same-data, same-backbone post-hoc baseline, and on a shared benchmark its 1.9 million parameters outperformed frozen foundation models with 3 to 1,300 times more parameters.
MELD puts the score difference between wild-type and mutant sequence directly into the supervised objective, instead of training on absolute signal and comparing two sequences after the fact. On 1.57 million ClinVar variants it raised AUROC from 0.800 to 0.986 against a same-data, same-backbone post-hoc baseline, and on a shared benchmark its 1.9 million parameters outperformed frozen foundation models with 3 to 1,300 times more parameters.
Qatar University University of Doha for Science and Technology
Aliaa Alajji（First author）Mohamed Mokhtar Elshrif（Corresponding author）
MELD Variant Effect Prediction Variant Pathogenicity Classification Saturation genome editing
发表出处待核验 Sep 14, 2026
functional-standard-atlas: an attenuation-corrected, territory-resolved benchmark of variant effect predictors against saturation genome editing
This work builds a frozen, content-hashed data asset and a uniform scoring harness that maps seven MaveDB saturation genome editing (SGE) score sets to GRCh38, harmonises orientation and freezes them into immutable matrices, then evaluates nineteen variant effect predictors across sixteen strata using per-gene Spearman rho pooled by DerSimonian-Laird random-effects meta-analysis with per-stratum measurement-reliability estimates, attenuation correction, paired dependent-correlation tests and leave-one-gene-out validation, covering 64,178 variants and seven cancer susceptibility genes.
This work builds a frozen, content-hashed data asset and a uniform scoring harness that maps seven MaveDB saturation genome editing (SGE) score sets to GRCh38, harmonises orientation and freezes them into immutable matrices, then evaluates nineteen variant effect predictors across sixteen strata using per-gene Spearman rho pooled by DerSimonian-Laird random-effects meta-analysis with per-stratum measurement-reliability estimates, attenuation correction, paired dependent-correlation tests and leave-one-gene-out validation, covering 64,178 variants and seven cancer susceptibility genes.
National University of Singapore
Ningyi Zhang（First author, Corresponding author）
functional-standard-atlas Saturation genome editing Attenuation correction

You’re all caught up