Drop the table your screen produced. This measures how much of its
order is explained by how the gene sets were built — chiefly their size —
with no reference to what any gene does.
Drop your results file hereor click to choose · CSV, TSV or TXT
MAGeCK · BAGEL2 · drugZ · g:Profiler · DAVID · clusterProfiler · Enrichr · fgsea · GSEA
Nothing to hand? — the 50 Hallmark programs from the study behind this tool. It grades us
at 46%, which is worse than nine in ten published screens.
Your file never leaves this browser. There is no server and no
upload: the Python that runs the check is downloaded to your machine and runs there.
You can disconnect after the page loads and it still works.
The verdict
Against the field
What moves when the correction is applied
Does a model beat size alone?
This table has a column that could be a model's predicted
score. If it is, the check below reports what a predictor that sees
only how big each set is scores on your own evaluation, next to
yours. The size-only baseline is leave-one-out, so it never saw the row it
predicts.
The metric is asked for, never inferred — a
baseline scored with a different metric than yours is not a comparison.
Look up a published screen's floor
The atlas carries the no-biology floor for 1,272 published human
CRISPR screens — the share of each one's gene-set ranking that set size alone predicts.
Enter a BioGRID ORCS screen id to get the
number, the method, and the string to cite. Looked up, never recomputed, so your copy
and everyone else's agree.