Design over nature

For each shipped cell pair: the designed sequences and the natural promoters they were bred from, side by side. A promoter is the stretch of DNA in front of a gene that decides how strongly it is switched on. Its margin is how much more strongly the model predicts it fires in the target cell type than in the cell type it is supposed to leave alone. Design over nature divides one by the other: a design’s margin over the margin of the best natural promoter — both measured in the same genomic window, in the same run, under the same two context strings.

Why a ratio, and not the margin itself. A margin belongs to the window it was measured in as much as to the sequence. The identical 600 bp albumin promoter scores -0.012 in the FOXB2 window and 4.844 in FAM47C — same DNA, same cell type, same checkpoint, and the whole difference belongs to the window rather than to the sequence. Divide by a natural promoter measured in that same window and the window largely cancels; quote the raw margin and it does not. That is why this ratio is the one number allowed to cross between the two databases, and why every margin on this site is quoted with its window named.

The two databases are stacked here, never merged. A natural promoter’s rank is partly recall of the model’s own training data — it was trained on essentially every gene’s promoter — and a generated design’s rank cannot be, because the sequence exists in no genome. Pooling them into one ranking would put those two kinds of rank in one column, so the tables below stay separate.

What makes the comparison legitimate. The natural promoters below are not quoted from some other experiment: they were members of the generation run’s own starting population, scored in the same window, in the same run, under the same two verbatim context strings, on the same checkpoint. That is the only condition under which a design’s margin and a natural promoter’s margin are the same quantity — a margin from any other run or window does not meet it, and cannot be substituted here.