Popular repositories Loading
-
enzyme-kinetics
enzyme-kinetics PublicEnzyme kinetics with honest uncertainty: Lineweaver-Burk's Km spread is 6086% against 18.2% for nonlinear fitting under constant error, and an i.i.d. standard error is 12× too small against a trace…
Python
-
variant-effect-prediction
variant-effect-prediction PublicZero-shot ESM-2 variant effect prediction against ProteinGym deep mutational scanning: a better training target ranks better in 32 of 32 replicate pairs, but does not pick the top ten significantly…
Python
-
single-cell-de
single-cell-de PublicTreating cells as replicates: a per-cell Wilcoxon test called 73% of genes significant on data where nothing changed; pseudobulk called none — and loses no power. Simulation and real data, with thi…
Python
-
mlip-fidelity
mlip-fidelity PublicDo ML interatomic potentials preserve chemical decisions? 16,089 quantum conformer pairs: ORB-omol 0.900, MACE-OFF23 0.760, MMFF94 0.601 — with the disclosure that two of the three models trained o…
Python
-
epistasis-ceiling
epistasis-ceiling PublicAdditivity ceiling on ProteinGym double mutants: adding single-mutant effects ranks 813,014 of them well (median AUC 0.884), but the apparent bias is a centring artefact — and none of the tested pr…
Python
-
admet-split-truth
admet-split-truth PublicHow far each ADMET validation split mis-states the error on the next batch of compounds: random 0.76, scaffold 0.84, time-calibrated 1.02, measured against a real time-forward split.
Python
If the problem persists, check the GitHub status page or contact support.