Figure 3.

Toward a probabilistic chromatin definition.

1. Using 5795 proteins of known function, the percentage of chromatin (red) and non-chromatin proteins (blue) was calculated for overlapping windows (e.g., gray box) of the random forest (RF) machine learning score. Fitting a sigmoid curve through the percentages translates the RF score into “interphase chromatin probabilities” (ICPs). 1823 evaluation proteins are known chromatin players, while 3972 have no expected chromatin function.
2. Scatterplot showing ICPs for 7635 human proteins (the 5795 evaluation proteins and 1840 uncharacterized proteins). Pie charts show protein categories above and below ICP 0.5, including uncharacterized proteins (white).