SL mitochondrial granularity triple

IN_PROGRESS PIPELINE

Species: human, mouse, yeast, SCHPO, rat, ARTAN

The mitochondrial granularity triple

A controlled comparison. Three UniProt locations, one organelle, one pipeline, one set of
curators — differing only in how specific the location is.

SL GO term Reviewed Issues Rate
SL-0171 GO:0031966 mitochondrial membrane 13 4 31%
SL-0168 GO:0005743 mitochondrial inner membrane 19 2 11%
SL-0170 GO:0005759 mitochondrial matrix 14 1 7%

The under-specified location is flagged three to four times as often as its precise siblings.
Because the organelle, the pipeline, the mapping mechanism, and the reviewer pool are all held
constant, the difference is attributable to specificity itself rather than to anything about
mitochondria.

The SL-0171 flags are all the same complaint

All four resolve to "should be the inner membrane":

The specific siblings fail differently, and more interestingly

The three flags on SL-0168 and SL-0170 are not granularity complaints. They are substantive
localization errors that a precise term makes visible:

What this implies

Under-specification and mis-assignment are different diseases with different treatments.
Making an SL location more specific does not reduce the number of errors; it converts
un-checkable vagueness into checkable claims, some of which then turn out to be wrong. That is
an improvement — a 7% error rate on precise terms is more useful than a 31% vagueness rate on
imprecise ones — but it should not be sold as error reduction.

No annotations were changed under this subproject; all four SL-0171 cases had already been
flagged by prior reviews.